ArticlePDF Available

Abstract and Figures

In recent years, blogging has become an exploding passion among Internet communities. By combining the grassroots blogging with the richness of expression available in video, videoblogs (vlogs for short) will be a powerful new media adjunct to our existing televised news sources. Vlogs have gained much attention worldwide, especially with Google's acquisition of YouTube. This article presents a comprehensive survey of videoblogging (vlogging for short) as a new technological trend. We first summarize the technological challenges for vlogging as four key issues that need to be answered. Along with their respective possibilities, we give a review of the currently available techniques and tools supporting vlogging, and envision emerging technological directions for future vlogging. Several multimedia technologies are introduced to empower vlogging technology with better scalability, interactivity, searchability, and accessability, and to potentially reduce the legal, economic, and moral risks of vlogging applications. We also make an in-depth investigation of various vlog mining topics from a research perspective and present several incentive applications such as user-targeted video advertising and collective intelligence gaming. We believe that vlogging and its applications will bring new opportunities and drives to the research in related fields.
Content may be subject to copyright.
15
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web
WEN GAO, YONGHONG TIAN, and TIEJUN HUANG
Peking University
and
QIANG YANG
Hong Kong University of Science and Technology
In recent years, blogging has become an exploding passion among Internet communities. By combining the
grassroots blogging with the richness of expression available in video, videoblogs (vlogs for short) will be
a powerful new media adjunct to our existing televised news sources. Vlogs have gained much attention
worldwide, especially with Google’s acquisition of YouTube. This article presents a comprehensive survey of
videoblogging (vlogging for short) as a new technological trend. We first summarize the technological chal-
lenges for vlogging as four key issues that need to be answered. Along with their respective possibilities,
we give a review of the currently available techniques and tools supporting vlogging, and envision emerg-
ing technological directions for future vlogging. Several multimedia technologies are introduced to empower
vlogging technology with better scalability, interactivity, searchability, and accessability, and to potentially
reduce the legal, economic, and moral risks of vlogging applications. We also make an in-depth investiga-
tion of various vlog mining topics from a research perspective and present several incentive applications
such as user-targeted video advertising and collective intelligence gaming. We believe that vlogging and its
applications will bring new opportunities and drives to the research in related fields.
Categories and Subject Descriptors: F.1.2 [Computation by Abstract Devices]: Modes of Computation—
Online computation; H.3.4 [Information Storage and Retrieval]: Online Information Services—Web-
based services;H.4[Information Systems Applications]: Communications Applications—Computer con-
ferencing, teleconferencing, and videoconferencing
General Terms: Design, Human Factors, Management
Additional Key Words and Phrases: Survey, vlogs, vlogging, multimedia computing, vlog mining
ACM Reference Format:
Gao, W., Tian, Y., Huang, T., and Yang, Q. 2010. Vlogging: A survey of videoblogging technology on the Web.
ACM Comput. Surv. 42, 4, Article 15 (June 2010), 57 pages.
DOI =10.1145/1749603.1749606, http://doi.acm.org/10.1145/1749603.1749606
The authors are supported by grants from National Basic Research Program of China under contract No.
2009CB320906, Chinese National Science Foundation under contract No. 60605020 and 90820003, and Na-
tional Hi-Tech R&D Program (863) of China under contract No. 2006AA01Z320 and 2006AA010105. Dr.
Qiang is supported by Hong Kong CERG grant 621307.
Authors’ addresses: Y. Tian, the Institute of Digital Media, School of EE & CS, Peking University, Beijing,
100871, China; email: yhtian@pku.edu.cn.
Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted
without fee provided that copies are not made or distributed for profit or commercial advantage and that
copies show this notice on the first page or initial screen of a display along with the full citation. Copyrights
for components of this work owned by others than ACM must be honored. Abstracting with credit is per-
mitted. To copy otherwise, to republish, to post on servers, to redistribute to lists, or to use any component
of this work in other works requires prior specific permission and/or a fee. Permissions may be requested
from Publications Dept., ACM, Inc., 2 Penn Plaza, Suite 701, New York, NY 10121-0701 USA, fax +1 (212)
869-0481, or permissions@acm.org.
c
2010 ACM 0360-0300/2010/06-ART15 $10.00
DOI 10.1145/1749603.1749606 http://doi.acm.org/10.1145/1749603.1749606
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:2 W. Gao et a l .
1. INTRODUCTION
When the Web was invented, practically all forms of online media were still very much
a one-way street. However, with the rapid development and proliferation of the World
Wide Web, particularly with the emergence of Web 2.0 [O’Reilly 2005] and beyond, we
now have new means by which to express our opinions and gain access to informa-
tion instantly. Weblogs (blogs for short) are playing an increasingly important role in
realizing this goal. Like instant messaging, email, cell phones, and Web pages, blogs
are a new form of mainstream personal communication, allowing millions of people
to publish and exchange knowledge/information, and to establish networks or build
relationships in the blog world [Rosenbloom 2004]. Blogs began as a textual genre
of personal publishing, but within this genre visual expression, such as photoblogs,
and the adaptation of sound and video [Hoem 2005] were developed. By combining
the ubiquitous, grassroots blogging with the richness of expression available in video,
videoblogs (vlogs for short) will be an important force in a future world of Web-based
journalism and a powerful new media adjunct to our existing televised news sources
[Parker and Pfeiffer 2005].
As defined by Wikipedia [2008], videoblogging, shortened as vlogging in this arti-
cle, is a form of blogging for which the medium is video. Vlog entries are made regu-
larly and often combine embedded video or a video link with supporting text, images,
and other metadata. In recent years, vlogging has gained much attention worldwide,
especially with Google’s acquisition of YouTube in November 2006 for $1.65 billion.
YouTube is a popular vlog site that lets users upload, tag, and share video clips—
thereby making them known to the world. Since a video can show a lot more than text,
vlogs provide a much more expressive medium for vloggers than text-blogs in which
to communicate with the outer world. This has particular appeal to younger audience,
who are typically equipped with popular mobile devices such as personal digital assis-
tants (PDAs) and camera phones.
Vlogs have brought about a new revolution in multimedia usage. Nevertheless, a
quick look at the current vlogs reveals that not many of the results from multimedia
research, such as media content analysis, semantic content classification and annota-
tion, structured multimedia authoring, or digital-right management, found their way
into vlogging techniques. For example, a simple shot-segmentation technique could
provide YouTube users with small bookmarks for each video and allow them to eas-
ily jump into different video scenes, instead of watching it all or using the slider [Boll
2007]. Hence, there are many opportunities for multimedia researchers to provide vlog-
gers with better vlogging techniques for a more powerful experience in designing and
using vlogs.
This article outlines the current state of vlogging technology, covers the probable
evolution, and highlights what newcomers leading the charge are poised to start. Our
article is partly motivated by Parker and Pfeiffer [2005], which for the first time pro-
vided an introductory view of vlogging technology before 2005. Since then, vlogging has
become a huge success and is a very hot online application with high general interest.
We found that it is worthwhile to revisit the current vlogging technology and explore
the potential benefits and challenges that the multimedia community offers to vlog-
ging, and vice versa. Our idea is consistent with the panelists’ opinions in the panel
titled “Multimedia and Web 2.0: Hype, Challenge, Synergy” that was held half a year
after the time of this writing (April 2006), at the ACM 2006 Multimedia Conference in
Santa Barbara, California.
A possible organization of the various facets of vlogging technology is shown in
Figure 1 (see Section 2.3 for a more detailed description). Our article follows a sim-
ilar structure. For the purpose of completeness and readability, we first review the
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:3
Fig. 1. A view of the different facets of vlogging technology, reflected in the structure of this article.
blog phenomenon of recent years and then present an overview of current vlogs in
Section 2. We also summarize in this section the key issues of vlogging technology that
need to be answered. The rest of this article is arranged as follows: In Section 3, we
give a review of the infrastructure and tools of current vlogging technology. Then, we
present a new vision for future vlogging and introduce some techniques from multime-
dia to help make this vision possible in Section 4. Since the vlogging phenomenon also
poses many new computational opportunities for researchers, we pay special attention
to these issues in Section 5. The prevalence of vlogging and the further combination
of vlogging and multimedia technologies will in turn give birth to attractive applica-
tions such as vlog-based advertising and gaming, thus, in Section 6, we discuss some
samples of these applications. Finally, we conclude in Section 7.
2. VLOGS AND VLOGGING: A GLOBAL PICTURE
2.1. The Evolution of Blogs: From Text to Audio-Video
In or around 1997, blogging became a quickly spreading passion among Internet liter-
ates. A weblog,orblog is a “frequently updated Web page with dated entries in reverse
chronological order, usually containing links with commentary” [Blood 2002]. The term
originated from “WeB log,” and was promoted further by www.blogger.com as a blog.
Blog can also be used as a verb, meaning to maintain or add content to a blog. The act
itself is often referred to as blogging. According to Gill [2004], several primary charac-
teristics of a blog include regular chronological entries; links to related news articles,
documents, or blogs (referred to as blogrolling); archived entries such that old con-
tent remains accessible via static links (referred to as permalinks); ease of syndication
with RSS (Rich Site Summary, see Section 3.2.2) or XML feed. Blogs contribute to Web
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:4 W. Gao et a l .
content by linking and filtering evolving content in a structured way. All blogs and
their interconnections are often called the blogosphere [Rosenbloom 2004]. It is the
perception that blogs exist together as a connected community (or as a collection of
connected communities) or as a social network.
Many blogs provide commentary or news on a particular subject; others function
as more personal online diaries. Five main blogging motivations were identified in
Nardi et al. [2004]: documenting one’s life; providing commentary and opinions; work-
ing out emotional issues; thinking by writing; and promoting conversation and com-
munity. Blogs have become an increasingly important way of learning about news and
opinions not found in mainstream media, and blogging has become a popular social
activity for establishing and maintaining online communities. The number of blogs is
growing exponentially—the famous blog search engine, Technorati, reported that there
were about 4.2 million blogs worldwide in October 2004 [Rosenbloom 2004], and up to
112.8 million blogs by December 2007. It is estimated that there are about 175,000
new blogs a day, and about 11% (or about 50 million) of Internet users are regular blog
readers.
Traditionally, blogging is a textual activity, which is limited because text is only one
aspect of the diverse skills needed in order to understand and manage different aspects
of modern communication [Hoem 2005]. In recent years, the amount of digital multi-
media distributed over the Web has increased tremendously because almost anyone
can follow the production line of digital multimedia content. As a result, many differ-
ent types of blogs have emerged gradually, such as artlog (i.e., a form of art-sharing and
publishing in the format of a blog); photoblog (i.e., a blog containing photos); sketchblog
(i.e., a blog containing a portfolio of sketches); and in particular audioblog (or podcast)
and videoblog (vlog for short). They differ not only in the type of content, but also in
the way that content is delivered or written.
Aiming to present a richer story for bloggers, audioblogs and vlogs are twins; they ex-
hibit many similarities in terms of their content and production. Simply speaking, an
audioblog is a service that provides bloggers with the ability to attach audio to their
blogs with a microphone at any time from anywhere. Note that audioblogs are also
known as podcasts, since their pioneers are the iPod users; although podcasting is in a
stricter sense merely one form of audioblogging. On the other hand, even before audio
media became a full-blown component of blogging, video was already being experi-
mented with [King 2003]. Vlogging features many of the same characteristics found in
audioblogging, providing vloggers with a more compelling and expressive medium.
Vlogging has experienced tremendous growth over the past several years. The
Yahoo! Videoblogging Group saw its membership increase dramatically in 2005. The
most popular video-sharing site to date, YouTube, was publicly launched between
August and November 2005 and acquired by Google in November 2006. By March
17th, 2008, there were up to 78.3 million videos published on YouTube, with 150,000
to over 200,000 videos published everyday. According to the data collected by Mefee-
dia, the number of vlogs was only 617 in January 2005 and 8,739 in January 2006,
but up to 20,193 in January 2007 [Mefeedia 2007]. Driven by the prevalence of digital
cameras and mobile phones and the near-ubiquitous availability of broadband net-
work connections, vlogging has surged to an unprecedented level, and gained much
attention worldwide.
2.2. Vlogs: Concept and Taxonomy
According to Wikipedia [2008], a videoblog, which is shortened to vlog in this article,
is a blog that uses video as the primary content. Vlog entries are made regularly, and
often combine embedded video or a video link with supporting text, images, and other
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:5
metadata. These videos may be embedded and watched on the viewer’s Web browser,
or downloaded to the viewer’s machine or a portable device for later viewing. Like
textblogs, vlogs also often take advantage of Web syndication to allow for the distri-
bution of video over the Internet using either the RSS or Atom syndication formats,
for automatic aggregation and playback on mobile devices and PCs [Wikipedia 2008].
Apart from the term vlogs, videoblogs are also known as v-logs, vid-blogs, movie blogs,
vblogs, vidcasts, videocasts, vcasts, v-casts, episodic video, Web shows or online TV,
and so on. Vlogs are created by videobloggers or vloggers, while the act itself is re-
ferred to as videoblogging or vlogging. As a derivative of blogosphere, vlogosphere is
the collective term encompassing all vlogs. Vlogs exist together as a community or a
social network.
Roughly, the life cycle of a vlog is defined in this article as consisting of three stages:
Producing: In this stage, the vlogger creates and edits a video, and uploads it to the
hosting site and then a new vlog is generated. Sometimes, vloggers can even edit
videos online, without a locally installed software.
Posting: The new vlog is distributed online, and starts to get recognized in the vlogo-
sphere and often get linked by other vlogs. In this stage, the vlog can be viewed and
commented by other vloggers. Some vlogs even become authorities in a certain field.
Rather than always having to remember to visit a site, a vlog reader can configure
his or her RSS browser software to automatically subscribe to the timely updates.
Archiving: When the vlog becomes out of date or loses usefulness, it will be archived
or even deleted.
To further characterize a vlog, we have to distinguish among several significantly dif-
ferent genres that claim to be vlogs, ranging from simply uploading unedited video
files via play-lists to edited sequences, sometimes with complex interactivity [Hoem
2005]:
Vogs: Vogs are made of pre-edited sequences which normally include interactive
elements. Typically, they are made with different kinds of software and then posted
to individual blog sites. A typical example of vogs is the vlog of B. Obama in the 2008
US presidential election, at http://www.youtube.com/user/BarackObama-dotcom (see
Figure 2(a)).
Moblogs: Mobile blog (moblog for short) is a form of blog in which the user pub-
lishes blog entries directly to the Web from a mobile phone or other mobile device
[Wikipedia 2008]. Entities in moblogs normally contain uploaded pictures, audio-video
clips and additional text, not edited sequences. Moblogging is popular among peo-
ple with camera-enabled cell phones which allow them to either e-mail photos and
videos that then appear as entries on a Web site, or use mobile blogging software to
directly publish content to a Web server. Here moblogs are used to denote vlogs con-
taining relatively short, autonomous video clips. Examples of moblogs can be found at
http://moblog.net/ (see Figure 2(b)), which is a well-known free hosting site for moblogs.
Playlists: Playlists are collections of references to video files on different servers,
which may even provide a level of interactivity without manipulating the content in
these files. One way to achieve this is by using SMIL (synchronized multimedia inte-
gration language). Strictly speaking, playlists are the filter-style vlogs. Examples of
playlists can be found at http://www.mefeedia.com/collections/ (see Figure 2(c)), which
contains 4,114 playlists, gathering 16,182 videos by April 2008.
It should be noted that the distinction between vogs and moblogs is clear [Hoem 2005]:
the videos in vogs are edited, and may offer quite complex interactivity, while moblogs
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:6 W. Gao et a l .
Fig. 2. Screenshot of exemplary vlogs: (a) vog at http://www.youtube.com/user/BarackObama-dotcom;
(b) moblog at http://moblog.net/; (c) playlist at http://www.mefeedia.com/collections/; and (d) diary-style vlog
at http://crule.typepad.com/.
are generally easy to use because moblogging is just a matter of simply uploading a
video file to a dedicated vlog site.
Vlogs can also be grouped into two categories according to their presentation styles:
diary or Web-TV show. The vlogs of the first category take the online video diary
style. Such an example is Charlene’s vlog at http://crule.typepad.com/ (as shown in
Figure 2(d)), where video is included through hyperlinks. Currently, most vlogs take
the second category of presentation formats, i.e., the Web-TV show-like style. Such an
example is Obama’s vlog (as shown in Figure 2(a)), which has inline video files.
Since a video can show a lot more than pure text, pictures, or audio, there are a
lot more things that a vlog can cover compared to a typical textblog, photoblog, or au-
dioblog. For example, if one runs a real-estate business and he or she wishes to provide
regular virtual tours of the properties for sale, a vlog that features regular Web tours
would be very appropriate. In some sense, video is easier to produce, as it only needs to
record some realistic scenes and can be informative even with little or no editing. The
affordability and portability of most consumer-level video camcorders and some mobile
phone handsets mean that anyone can easily express themselves in a documentary or
narrative-style film on a daily basis and then post it for viewer consumption [King
2003]. Compared to text- or audioblogs, vlogs can be used in a much wider range of ap-
plications such as online education and learning, online gaming, products marketing,
and news reporting.
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:7
Tab le I. Differences Among IPTV, Internet Video and Vlogging Services
IPTV Internet Video Vlogging
Content TV programs, movies Movies, TV programs, news, Mainly user-generated vi-
partially user-generated video deo in talk-radio format
Length Long clips or live stream Long clips or live stream, Short clips (typically 3–5
and partially short clips minutes)
Video MPEG 2/4, H.264 (with Windows Media, Real, Quick Mostly Flash (often with
Format high-definition) Time, Flash and others low-definition)
Video “Broadcast” TV Quality, Best effort quality, QoS not Often low quality, no
Quality Controlled QoS guaranteed QoS
Content EPGs by timeline or Often by topic category, by As dated entry in reverse
Organization topic category order view times/upload date etc. chronological order
Users Known customers with Any users (generally un- Any users (generally un-
known IP and locations known) known)
Footprint Local (limited operator Potentially supranational or Potentially supranational
coverage) worldwide or worldwide
Receiver Set-box with a TV dis- PC and mobile devices Mobile devices and PC
Device play
Producing Professional filmmaking Advanced editing tools for Simple-to-use editing
tools for filmmakers service providers tools for common users
Transmission Broadband network with Internet access, often using Internet access, possibly
CDN support P2P transmission using P2P transmission
Delivery Video-on-demand, live Video-on-demand, partially Video-on-demand
Form broadcast, delayed live live broadcast
Interactivity Program selection, few (Sometimes) voting or rank- Commenting or hyper-
interaction ing, with a few interaction linking, social interaction
Searchability Metadata/EPG search Metadata/caption-based, par- Metadata search, XML-
tially content-based search based aggregation
RSS Support No Partially yes Yes
Reliability Stable Subject to connection Subject to connection
Security Users are authenticated Unsafe Unsafe
and protected
Copyright Protected video Often unprotected Mostly unprotected
Other EPGs, onsite support Low-quality on-demand ser- Generally no support
Services vices, generally no support
Some comparison items between IPTV and Internet video are quoted from Martinsson [2006].
∗∗CDN (content delivery network) is a system of computers networked together across the Internet that
cooperate transparently to deliver content (especially large media content) to end users.
∗∗∗EPG means electronic program guide, an on-screen guide to scheduled programs, typically allowing a
viewer to navigate, select, and discover content by time, title, channel, genre, and so on.
2.3. Vlogging: Key Technological Challenges
Vlogging is a form of blogging for which the medium is video. However, online delivery
of video content is not a new activity, either by direct downloading or via streaming.
It is useful to make a comparison among three closely related services of online video
delivery (see Table I): IPTV (Internet Protocol Television), Internet video, and vlog-
ging. From the content perspective, the indexed video content in three services have
some differences. Generally speaking, IPTV often provides high-quality TV programs
and movies with guaranteed quality of services (QoS); a general online video service
often supports a wide range of videos, from low-quality videos generated by nonpro-
fessional users, to high-quality TV clips or movie files, either in live broadcast or via
downloaded file; while videos in vlogs are often generated by nonprofessional users in
a talk-radio format and with fairly short length, encoded with relatively low resolu-
tion and archived in reverse-chronological order. Different video types and dissimilar
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:8 W. Gao et a l .
service targets then characterize their different supporting technologies, as shown in
Table I.
Like text-only blogging, we need supporting techniques to facilitate effective vlog-
ging. However, it is clear that a simple combination of textblogging tools and online
video delivery does not meet the ever-growing requirements of vloggers [Parker and
Pfeiffer 2005]. Thus, with the explosive growth of vlogs worldwide, several challenges
are posed for vlogging technology, which are summarized as follows:
Basic supporting issue: The first challenge mainly addresses the basic support-
ing infrastructure and techniques for vlogging. Network bandwidth and media stor-
age can be considered as two major hardware infrastructures for vlogging, while
Web 2.0 and streaming media technologies are two software infrastructures. To fa-
cilitate vloggers designing and using vlogs, possible solutions range from simply in-
tegrating textblogging tools with online delivery of video, to flexibly combining differ-
ent multimedia systems and services on all stages of vlogging. Even in the simplest
situation, the supporting platform also needs some blogging softwares and systems
(e.g., RSS/Syndication, content management systems, hosting and directory services),
and necessary video production and distribution tools (e.g., video creation and coding
software, streaming servers). Given these matured systems, tools and services, vlog-
ging technology will go further by “mashing-up” them into more powerful supporting
platforms.
Value-added issue: Vlogging is a typical application of Web 2.0. As pointed out by
Boll [2007], Web 2.0 and multimedia can benefit each other, and we should integrate
each other’s valuable results and best practices. Thus the second challenge is mostly
about what multimedia technology could give to vlogging (not to the much broader
field, Web 2.0) and vice versa. In this sense, we refer to this challenge as a value-added
issue, compared with the “basic supporting issue”.
Incumbent issue: The ease with which anyone can view and upload videos to
YouTube and similar vlog hosting sites also poses potential copyright, moral, and le-
gal issues [Meisel 2008]. For example, Viacom, a large corporation whose holdings
include MTV and Comedy Central, sued YouTube in March 2007 for over $1 billion for
“massive intentional copyright infringement” [Peets and Patchen 2007]. The “incum-
bent techniques” for distributing content can react aggressively to this new source of
competition and pursue technological strategies to combat the abuse of unauthorized
content by vloggers.
Incentive application issue: In the development process of vlogging, YouTube’s
success is undoubtedly one of the major milestones, mainly because it created an in-
centive application platform that can provide a better user experience around sharing
video clips and publishing vlogs online. This shows us that the spawning of incentive
applications may be one important issue for future vlogging technology. Some promis-
ing examples are vlog-based advertising and gaming. These incentive applications may
drive the further development of vlogging.
These four challenges cover most of the important problems and aspects related to the
current and future vlogging technology. We believe that embarking on the research
and development of vlogging technology should consider and tackle these challenges.
Along with their respective possibilities, we can comprehensively survey, analyze, and
quantify the current progress and future prospects of vlogging technology. Towards
this end, different facets will be addressed one by one in the following sections, as
illustrated in Figure 1.
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:9
Fig. 3. An overview of sketchy roadmaps on video services.
3. VLOGGING TECHNOLOGY IN THE REAL WORLD
By the nature of its task, current vlogging technology boils down to two intrinsic prob-
lems: (a) from a system perspective, what software, hardware, and network infrastruc-
tures are need to support online vlogging applications and services; and (b) from a
user perspective, how to provide vloggers with technological supports for all stages of
vlogging. Thus this section is dedicated to understanding the current status of vlog-
ging infrastructures, techniques, and tools in the real world from the two perspectives
above.
3.1. Vlogging Infrastructure
As far as technological advances are concerned, growth in video services has unques-
tionably been rapid. Figure 3 shows an overview of sketchy roadmaps on video services,
which are roughly divided into four phases: broadcast TV services (e.g., traditional
TV); on-demand services (e.g., VOD); interactive services (e.g., IPTV); and social ser-
vices (e.g., vlogs). We can see that this evolving progress is advanced by several key
technologies. For example, high-capacity, low-cost storage and streaming media tech-
nologies enable on-demand video services, and IPTV is often provided in conjunction
with VOD and broadband network technologies. Similarly, vlogging services emerge
when these technologies and the new generation of the Web, Web 2.0, meet. Compared
with interactive video services such as IPTV, what makes vlogs new is their social par-
ticipation and embeddedness. From this perspective, we refer to these kinds of new
video services as socially-interactive video services,orsimplysocial services. It should
also be noted that for social services such as vlogs, these technologies can serve as the
enabling infrastructures. Among them, broadband network and media storage can be
considered two major hardware infrastructures for vlogging technology. Meanwhile,
vlogging is closely related to Web 2.0 and streaming media technologies (referred to as
the software infrastructures).
3.1.1. Broadband Network and Media Storage.
The advances in computer networking
combined with high-capacity, low-cost storage systems made online delivery of video
practical, either by direct downloading or via streaming. Compared to text-, photo-
or even audioblogs, vlogs obviously place many more requirements on network band-
width and media storage capacity. Take YouTube as an example. In August 2006, the
Wall Street Journal published an article revealing that YouTube was hosting about
6.1 million videos and had more than 65,000 videos uploaded everyday, consequently
requiring about 45 terabytes of storage space and as much as 5 to 6 million dollars per
month in running costs (specifically the required bandwidth). These figures were up
to 78.3 million videos hosted totally and 150,000 videos uploaded daily in March 2008.
We can speculate that the required storage space and corresponding running costs are
exponential in growth.
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:10 W. Gao et a l .
From a user perspective, the limited network bandwidth also poses a significant
challenge for efficient video transmission in the current network environment where
sufficiently high bandwidth is not widely available. To reduce network congestion,
peer-to-peer (P2P) transmission technology is widely used. The P2P network refers
to a network that primarily relies on participants in the network, rather than on a few
dedicated servers for its service. This is a fundamentally different paradigm compared
with the client-server architecture, and brings a number of unique advantages such as
scalability, resilience, and effectiveness to cope with dynamics and heterogeneity [Liu
et al. 2008]. BitTorrent is a protocol designed for transferring files in a P2P network,
which has been exploited by many vlogs and related tools, such as FireANT, to improve
the transmission efficiency.
Discussion. From the perspective of vlogging service providers, video storage and
transmission costs are still significant; from the perspective of vloggers, however, de-
spite the network, bandwidth is still a concern for the vloggers who do not have suffi-
ciently high bandwidth available; however, the problem will ventually solve itself with
the advent of near-ubiquitous broadband/wireless Internet access. Therefore, more de-
tailed analysis of broadband network and media storage is outside the scope of this
article.
3.1.2. Web 2.0.
Web 2.0 refers to an updated and improved version of WWW that al-
lows the users to communicate, collaborate, and share information online in completely
new ways. Web 2.0 has numerous definitions. As pointed out by O’Reilly [2005], there
are huge numbers of disagreements on what Web 2.0 means. However, the central
point of Web 2.0 is the user’s stronger involvement and participation in the Web, form-
ing social networks and virtual communities in a global network. Generally speaking,
Web 1.0 was marked by isolated information silos and static Web pages where the
user used search engines or surfed from one Web site to another; Web 2.0, on the other
hand, is characterized by a more living, dynamic, and interactive Web that is based on
social networks, user-generated content, and that is an architecture of participation
[Best 2006; Jensen 2007].
One of the key drives in the development of Web 2.0 is the emergence of a new gen-
eration of Web-related technologies and standards [Anderson 2007]. Technologically,
Web 2.0 is mainly considered [Boll 2007] to represent: (1) the combination of matured
implementation techniques for dynamic Web applications; (2) the mashing of existing
Web services into value-added applications; and (3) an emphasis on community and
collaboration, as well as the establishment of direct distribution platforms for user-
generated content. It should be noted that the technological infrastructure of Web 2.0
is still evolving.
One important feature of Web 2.0 is that it puts less emphasis on the software and
far more on the application that provides a service. Key Web 2.0 applications include
blogging (including audioblogging and vlogging), wiki, tagging and social bookmark-
ing, multimedia sharing, and content syndication [Anderson 2007]. Some well-known
commercial examples that count as Web 2.0 applications are Flickr, YouTube, Upcom-
ing, Wikipedia, Facebook, and Google Earch.
Discussion. Web 2.0 provides simple, but interesting, working services and is better
at involving and integrating users so that they can participate more. Given the strong
focus on media in Web 2.0, however, not many of the valuable results from multimedia
found their way into Web 2.0 [Boll 2007]. It can be argued that vlogging is exactly one
of the best fields in which they could meet.
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:11
3.1.3. Streaming Media.
From a user perspective, streaming media are viewable im-
mediately after downloading starts [Lawton 2000]. The real-time content is streamed
through the user datagram protocol (UDP). In terms of transport protocols, a
streaming-media server uses internet protocol (IP) multicast, the real-time streaming
protocol (RTSP) incorporating UDP, or the transmission control protocol (TCP), or TCP
alone—a hybrid strategy with real-time streaming via HTTP when a firewall cannot
pass UDP data [Pieper et al. 2001]. Currently, most vlogs can support the streaming
media preview mode. This means that the user can check out an audio or video before
deciding to download it.
During the past two decades, there have been tremendous efforts and many techni-
cal innovations made in support of real-time video streaming. IP multicast represents
an earlier attempt to tackle this problem [Deering and Cheriton 1990]. It forwards
IP datagrams to a group of interested receivers by delivering the messages over each
link of the network only once, and creating copies only when the links to the destina-
tions split. As a loosely coupled model that reflects the basic design principles of the
Internet, IP multicast retains the IP interface and requires routers to maintain a per-
group state. Unfortunately, today’s IP multicast deployment remains largely limited
to reach and scope due to concerns regarding scalability, deployment, and support for
higher level functionality [Liu et al. 2008]. To address many of the problems associated
with IP multicast, several researchers have advocated moving multicast functionality
away from routers towards end systems [Chu et al. 2000]. As such a representative
nonrouter-based architecture, P2P-based broadcast has emerged as a promising tech-
nique that supports real-time video streaming, but is also cost effective and easy to
deploy [Li and Yin 2007; Liu et al. 2008]. Thus, many P2P video streaming solutions
and systems are proposed.
Discussion. With more and more vlogs taking the presentation form of online Web-
TV, streaming media technology is expected to be more cost-effective, more scalable,
and easier to deploy. However, although P2P applications such as file download and
voice-over-IP have gained tremendous popularity, P2P-based video delivery is still in
its early stage, and its full potential remains to be seen. We will explore this issue
further in Section 4.1.1.
3.2. Existing Supporting Techniques for Vlogging
Essentially, existing vlogging technology is a simple mashing of textblogging tech-
niques with online video delivery. Figure 4 shows a paradigm of existing supporting
techniques for vlogging; specifically, they focus mainly on the following issues:
—How to support efficient online video presentation and delivery;
—How to publish the machine-readable items containing metadata (such as the au-
thor’s name and publication date) and content for videos;
—How to add comments to vlogs, or how to interact with vlogs through hyperlinks;
—How to automatically aggregate updates from widely distributed vlogs;
—How to collaboratively create and manage tags for annotating and categorizing video
content;
—How to effectively index and search the regularly updated vlogs (and videos).
Below, we review each of these issues.
3.2.1. Online Video Presentation and Delivery.
The broadband-powered ability to view
and download online video content is just one part of vlogging; the other part is ap-
propriate video presentation and delivery interfaces that allow vloggers to quickly and
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:12 W. Gao et a l .
Fig. 4. A paradigm of existing supporting techniques for vlogging.
intuitively access vlogs in a Web browser or even use a mobile handset. The continuous
improvement in video delivery solutions, including video compression/coding, online
video player, and streaming video delivery, has played an important role in advancing
the prospects of online vlogging services.
Video compression and coding for streaming media delivery on the Internet has been
studied for more than ten years [Conklin et al. 2001], and is still an active research
field. Generally speaking, video compression is used to (often lossy) reduce the size
of the video data, followed by codecs (i.e., coding and decoding devices or softwares)
to compactly but accurately represent the video content. The present generation of
codecs, such as Windows Media Video (WMV), VC-1, AVC, and DivX, are now capable
of compressing rather large files to much smaller while maintaining picture resolution.
For the video stream to be useful in stored or transmitted forms, it must be encapsu-
lated together in a container format. A container format is a computer file format that
can contain various types of data, compressed by means of standardized audio/video
codecs. The most popular containers for streaming video are AVI, ASF, QuickTime,
RealMedia, DivX, and MP4. Finally, coded videos can be played in video players such
as Apple’s QuickTime, Microsoft’s Media Player, and RealNetworks’ RealPlayer.
However, compared with RealMedia, QuickTime or Windows Media, Flash video
(FLV) has much greater presence in the real-world applications of online video de-
livery. FLV videos can be delivered in the following ways: (1) using embedded video
within SWF files; (2) using progressive download FLV files; and (3) streaming video
from own Flash Media Server or from a hosted server using Flash Video Streaming
Services. Table II summarizes the characteristics of FLV video delivery techniques
(quoted from Adobe [2009] with necessary simplification).
The current crop of online video and vlog Web sites such as MySpace, YouTube,
and Google Video use FLV codec as the codec of choice, because it is highly efficient,
playable and viewable on most operating systems via the widely available Adobe Flash
Player and Web browser plug-in [Amel and Cryan 2007]. In contrast to video in other
formats, often played in a dedicated external player, Flash player is directly embed-
ded into a Web page without requiring the user to launch a separate application. FLV
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:13
Table II. Summary of Video Delivery Techniques in FLV Video
Embedded Video Progressive Download Streaming Delivery
Encoding VP6 video codec for FLV Plug-in or stand-alone FLV Same as progressive down-
Player 8+,orbeencoded Encoder, also support load, additionally with
by other codecs and then MPEG-4 formats encoded bandwidth detection capa-
packaged by Flash using H.264 codec bilities in streaming
File Size AV streams and Flash in- SWF and FLV files are se- Same as progressive
terface contained in a sin- parate, resulting in a smaller download
gle large file SWF file size
Timeline Individual keyframes are Be played back at runtime, Same as progressive
Access visible on Flash Timeline keyframes are not visible download
Publishing Need to republish the en- Not require referencing the Same as progressive
tire video file video file directly download
Frame Rate Video’s and SWF’s frame Video’s and SWF’s frame Same as progressive
rates are the same rates may be different download
ActionScript Control SWF content’s Use the netStream object for Same as progressive
Access playback on Flash Time- video playback download, and can also use
line for video playback server-side ActionScript
Components No video-specific compo- Media, FLVPlayback compo- Media, FLVPlayback, sever
nents nents comm. components
Web SWF file is progressively Video files are progressively Video files are streamed
Delivery downloaded downloaded, cached for play from server, played locally
Performance AV sync. is limited after Bigger and longer video, with Best performance, with
120s of video reliable AV sync. and best optimal bitrate and limited
image quality image quality
Usage Short video clips (<1m, Long video clips (>720×480 Very long video clips, live
<320×240 and <12fps) and <30fps) and multiway streaming
This table is taken from Table 2 in Adobe [2009] with necessary simplification.
is also popular due to its small file size (exactly for this reason it is often called a
microvideo), interactive capabilities, and progressive downloading (i.e., video begins
playing without having fully downloaded the clip) [Owens and Andjelic 2007]. To sup-
port different connection speeds, FLV codecs provide a wide range of encoding profiles,
from a total bitrate of 44 Kbps to 800 Kbps. Recent versions of Flash have added acces-
sibility functions [Reinhardt 2008]. For example, additional content, such as subtitles
or transcripts, can be associated with a Flash movie, consequently making FLV con-
tent visible to search engines; Flash Lite is available for some cell phones and other
devices.
Discussion. Due to high streaming efficiency, FLV codec is quickly becoming the
mainstream codec of online video. However, the quality of FLV videos suffers greatly.
By comparison, typical DVD bitrates are around 8 Mbps, and DV video has a bitrate of
25 Mbps. With more high-bandwidth networks being rolled out, higher-quality video
delivery techniques will be adopted to provide vloggers with better experiences.
3.2.2. Media Syndication.
For a number of blogs, tracking updates everyday is a cum-
bersome and tedious task. Thus, we need an XML-based format for publishing head-
lines of the latest updates posted to a blog or Web site for use by other sites and direct
retrieval by end users. The format, known as a feed, includes a headline, a short de-
scription, and a link to the new article. This syndication process enables Web sites or
blogs that share a common interest to expand their content. By subscribing to feeds,
bloggers can quickly review the latest updates on those blogs from a consolidated index
rather than browsing from blog to blog.
Currently, the major blogging syndication formats are RSS and Atom. RSS stands
for rich site summary, RDF (resource description framework) site summary, or re-
ally simple syndication, and is actually an XML-based metadata description and
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:14 W. Gao et a l .
Fig. 5. An element tree for RSS 2.* and Media RSS.
syndication standard for publishing regular updates to Web-based content. RDF is
a family of W3C specifications originally designed as a metadata model, but which has
come to be used as a general method of modeling information through a variety of syn-
tax formats. Currently, RSS 1.* and RSS 2.* are two major lineages of RSS. RSS 1.*
uses namespaces and RDF to provide extensibility, and thus is suitable for the applica-
tions that need advanced RDF-specific modules; while RSS 2.* uses modules to provide
extensibility, and thereby is suitable for general-purpose, metadata-rich syndication.
Figure 5 shows a simple element tree for RSS 2.*. At the top level, an RSS document
is a <rss>element, with a mandatory attribute-called version that specifies the ver-
sion of RSS that the document conforms to. Subordinate to the <rss>element is a
single <channel>element, which contains information about the channel (metadata)
and its contents. (For more details about RSS, see http://www.rss-specifications.com/
rss-specifications.htm).
To further enhance the enclosure capabilities of RSS 2.*, a group from Yahoo! and
the Media RSS community worked on a new RSS module in 2004, called Media RSS.
It extends enclosures to handle other media types, such as short films or TV, as well
as provides additional metadata with the media, thereby enabling content publishers
and bloggers to syndicate multimedia content. (For more details about Media RSS, see
http://search.yahoo.com/mrss).
By incorporating lessons learned from RSS, the Internet Engineering Task Force
(IETF) is working on another syndication format—Atom. Atom defines a feed format
for representing and a protocol for editing the Web resources such as blogs and sim-
ilar content. The feed format is analogous to RSS but more comprehensive, and the
editing protocol is novel [Parker and Pfeiffer 2005]. (For more details about Atom, see
http://www.ietf.org/html.charters/atom-pubcharter.html).
It should be noted that some multimedia syntax standards such as MPEG-21 can
also be used to syndicate video in vlogs. MPEG-21 is the newest of a series of MPEG
standards. With the aim to enable the use of multimedia resources across a wide range
of networks and devices, MPEG-21 provides an open standards-based framework for
multimedia delivery and consumption [Burnett et al. 2003].
Discussion. Since overlaps exist between a blogging language such as RSS or Atom
and a multimedia metadata markup language such as MPEG-21, a combination might
be the best solution [Parker and Pfeiffer 2005]. Therefore, media syndication of vlogs
is more complex than that of textblogs.
3.2.3. Vlog Aggregation.
With media syndication, we can use a program or a Web site,
known as an aggregator, to regularly check the updates of various vlogs, and then
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:15
present results in summary forms for vloggers. A more exciting use is to automatically
aggregate vlogs into topic-based repositories (i.e., vlog directories). Thus, the aggrega-
tion process lets updates from widely distributed Web sites be viewed as an integrated
whole, consequently allowing the automated creation of continuous feeds for a partic-
ular topic from distributed sources.
Media aggregators are sometimes referred to as “podcatchers” due to the popularity
of the term “podcast” used to refer to a feed containing audio or video. They can be used
to automatically download media, playback the media within the application interface,
or synchronize media content with a portable media player. Current media aggregators
essentially employ the textblog aggregation technique. Once feeds of various vlogs are
available, a vlog aggregator just retrieves the headlines, textual descriptions, thumb-
nails, and links of new video items, without performing more complex, but impor-
tant, video content analysis and summarization. Despite this, there are still some ad-
vanced features that differentiate vlog aggregators from textblog aggregators, like the
following:
Bad internet archive warning: Even if it only contains a video file of 1 to 3 min-
utes, a vlog entry is much larger in size than a textblog entry. To reduce the storage
cost, many vloggers choose to periodically remove their out-of-date video items from
their vlog archives but retain the corresponding feeds in the archives. Thus, detecting
whether the archived videos are valid is indeed an important feature of vlog aggrega-
tors such as FireANT.
Filtering and alerting adult feeds: With more image/video content being placed
online, the chance that individuals will encounter inappropriate or adult-oriented
content increases. Hence detecting adult feeds is also a necessary function of vlog
aggregators. Towards this end, a simple method is to exploit the <media:adult>or
<media:rating>element in Media RSS, or classify feeds based on their text content.
Some more advanced methods such as pornographic content-detection based on im-
ages [Rowley et al. 2006] or classification of texts and images [Hu et al. 2007] can also
be used.
Support for a streaming media preview mode: Due to the large size of video files
in vlogs, some existing vlog aggregators (e.g., FireANT) and video surrogates [Song and
Marchionini 2007] provide support for a streaming media preview mode. This means
that the user can check video content before deciding to download it. This is one of the
special features of vlog aggregators compared with textblog aggregators.
Full BitTorrent support and scheduling for downloading media files: Again due
to the large size of video files in vlogs, some existing vlog aggregators (e.g., FireANT)
also support downloading media files in BitTorrent to improve the transmission effi-
ciency. Users can even schedule the downloading time for further reducing network
congestion. This is featured by desktop vlog aggregators.
Discussion. Video aggregation is one of the key innovations in the online video
ecosystem. As users are increasingly content-source agnostic and enjoy both profes-
sional and user-generated content equally, the role of aggregators is growing [Owens
and Andjelic 2007]. One possible research direction for vlog aggregation is to integrate
the existing video content analysis and summarization technologies into video aggre-
gators. We will further explore these topics later.
3.2.4. Commenting and Hyperlinking in Vlogs.
Typically, bloggers may post a response
from others, allowing a critique of ideas, suggestions, and the like, which some ar-
gue is a defining characteristic of a blog. Meanwhile, each blog entry contains one,
and often several hyperlinks to other Web sites, blogs, and stories. Usually, there is
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:16 W. Gao et a l .
a standing list of links, known as blogrolls, to the author’s favorite bookmarks. Some
current blogging systems even support a linkback functionality. A linkback is a method
for Web authors to obtain notifications when other authors link to one of their docu-
ments, enabling authors to keep track of who is linking to or referring to their articles.
There are three linkback methods, that is, Refback, Trackback, and Pingback, but they
differ in how they accomplish this task [Wikipedia 2008]. In some sense, linkbacks are
similar to incoming hyperlinks in Web pages.
Similarly, vloggers can still post their text comments on other vlogs, and put hyper-
links or blogrolls directly in the supporting text of videos. A more powerful technique
is to embed outgoing and incoming hyperlinks in the video clips themselves. Embed-
ded outgoing hyperlinks enable a video to link “out” to a Web page or another video,
so as to offer some supplementary materials when the viewer watches the commen-
tary under discussion; while incoming hyperlinks let the portion of the video under
discussion be directly referenced by a specifically constructed URL pointed from an-
other video, Web page, or email document [Parker and Pfeiffer 2005]. Currently, most
popular media players such as QuickTime, RealPlayer, or Media Player can crudely
support embedded hyperlinks.
Discussion. In current vlogs, video is mainly used as a powerful narrative media,
and readers can only add comments to vlog entries in text format. However, if the
discussions among vloggers were in video format, they would be much more lively.
Researchers in the multimedia community have conducted some initial explorations
on this topic. For example, Yan et al. [2004] proposed a tool called “conversant media,”
which is a video-based learning application that supports asynchronous, text-based
discussion over a network. Comments and replies are attached to the specific locations
on the time-based media that inspired them. Via this process, participants can learn
from both the video content and the resulting peer-to-peer commentaries.
3.2.5. Social Tagging.
Social tagging (a.k.a. folksonomy, collaborative tagging, social
classification, social indexing) is the practice and method of collaboratively annotat-
ing and categorizing digital content by using freely chosen keywords (i.e., tags) [Jakob
2007]. Hence social tagging can be viewed as a user-driven and user-generated clas-
sification of content—both created by and used by the community of users. Users are
able to trace who has created a given tag, and thereby can also discover the tag sets
of another user who tends to interpret and tag content in a way that makes sense to
them. The result is often an immediate and rewarding gain in the user’s capacity to
find related content (a practice known as “pivot browsing”) [Wikipedia 2008]. More-
over, social tagging also makes it possible to establish a form of multiple, overlapping
associations that resemble the brain’s way of functioning instead of the rigid categories
of a given taxonomy [Jensen 2007].
Social tagging is very useful for the indexing and search of online videos in vlogs.
When there is not enough text description provided by a vlogger, tags created by the
community of users seem to be the only searchable metadata that is simple to im-
plement and currently available. So in the vlog hosting sites or aggregators such as
Youtube, most of the search operations are based on the textual tagging of the visual
content.
Discussion. Tags are subjective labels that might be misleading in semantics, and
limited in representing the content of a video. Some recent studies are attempt-
ing to integrate automatic annotation and social tagging to better understand me-
dia content. ALIPR, an automatic image annotation system at http://www.alipr.com,
has recently been made public for people to try and have their pictures annotated
[Datta et al. 2008]. As part of the ALIPR search engine, an effort to automatically
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:17
validate computer-generated annotation with human labeled tags is being made to
build a large collection of searchable images. In this way, the manifold results of au-
tomatic and semiautomatic annotation and social tagging might make the image and
video data truly “the next Intel inside” (as O’Reilly [2005] calls it).
3.2.6. Vlog Search.
As vlogs increase in population, how to effectively index these
vlogs and make them more easily searchable is becoming a new challenge. The aim of
vlog search is to make vlogs as easily searched by Web search engines as normal Web
pages. Searching video clips or streams in vlogs is, however, much harder than search-
ing textblogs. There are two main approaches. The first, and most widely used, is to
employ the traditional text-based retrieval techniques, which rely on the description
text or user-generated tags in RSS feeds of vlogs. Tags may describe the genre of a clip,
the actors that appear in it, and so on. The tagged video can then be easily searched.
The second approach uses software to “listen” to the video’s soundtrack, as used by
Blinkx, Podzinger, and Truveo video search engines [Wales et al. 2005]. Turning spo-
ken dialogue into text requires fancy algorithms, and is not always reliable, as anyone
who has ever used a speech-recognition system can testify. But the resulting text is
then simple to search.
In addition, vlog search services may provide some advanced functions. For exam-
ple, VlogMap (http://community.vlogmap.org) combines the geographic data about the
location with vlog entries to provide a world map of searched vlogs.
Discussion. Ideally, search engines should index the video content automatically by
scanning for embedded transcripts and timed metadata or even by performing auto-
mated analysis of the video content [Parker and Pfeiffer 2005]. However, currently
few vlog search engines can support content-based video search from vlogs. Given the
(still) poor capabilities of video analysis technology, it is not surprising that a video
search on the Web remains rather hit-or-miss. This is indeed ongoing work for future
vlog search engines, which will be discussed in detail in Section 4.1.3.
3.3. Current Vlogging Tools
Nowadays, most vlogs are powered by vlogging hosting services or stand-alone tools.
These systems make it easier to edit a self-made video, to set up a vlog, to update, dis-
tribute, and archive its content, and to search vlogs and videos from widely distributed
sites. Some systems are free and open sources, others are commercial products. In this
installment, current vlogging tools and services are surveyed from several different,
while related, perspectives.
3.3.1. Online Video Editing and Publishing Tools.
Video editing is the process of rearrang-
ing or modifying segments of video to form another piece of video. The goals of video
editing are the same as for film editing—the removal of unwanted footage, the isola-
tion of desired footage, and the arrangement of footage in time to synthesize a new
piece of footage [Wikipedia 2008]. With video editing softwares, vloggers are able to
cut and paste video sequences and integrate audio (background music, special effects,
and so forth) so as to create more attractive vlogs.
In the home video domain, there are a number of inexpensive yet powerful video
editing tools such as iMovie, Abode Premiere. Ordinary users often want to create
more professional videos using a set of videos that they captured. Towards this end,
there are many recent studies that aim to automate home-video editing, ensuring that
the edited video is of satisfactory quality. For example, Hua et al. [2004] proposed
an optimization-based automated home-video editing system that could automatically
select suitable or desirable highlight segments from a set of raw home videos and then
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:18 W. Gao et a l .
align them with a given piece of incidental music, to create an edited video segment of
the desired length. Similar MTV-style video generation system can also been found in
Lee et al. [2005].
The increased popularity of vlogging has resulted in a large increase in online video
editing activity. Currently, the Web is populated by tools and sites that allow users to
edit videos online within their browsers, without the need to install a specific software
on their local computers. These video editing tools and sites, such as International
Remix, Jumpcut, Videoegg, Eyespot, Motionbox, Photobucket, or One True Media, and
so on [Bordwell and Thompson 2006], are generally characterized by their great ease of
use, which makes it a breeze to do both basic and advanced editing even for nonexperts.
Collaboratively developed by Yahoo! Research Berkeley and the San Francisco Film
Society, Remix is a platform for Web-based video editing that provides a simple video
authoring experience for novice users [Schmitz et al. 2006]. This system was featured
at the 2006 San Francisco International Film Festival Web site for one month. Another
well-known online video editing tool is the Jumpcut, a 2007 Webby Award nominee
and Yahoo acquisition. Jumpcut offers very strong editing and enhancment tools, with
a clean, clear interface and community features.
Online video editing and vlog publishing are often integrated into one system. Given
a user-generated video, the next thing to do to set up a vlog is to publish the video over
the Internet. Vlog It, developed by Serious Magic, Inc., and Broadcast Machine, devel-
oped by the Participatory Culture Foundation (PCF), are two typical examples of such
tools. Vlog It makes it easy to share a video online, since all technical details like video
compression, file transfer, and thumbnail creation are handled automatically. Simi-
larly, Broadcast Machine can also be used to publish internet TV channels, including
vblogs and podcasts, by using BitTorrent. For a large video, network bandwidth is
always a key concern for the vloggers who do not have sufficient high bandwidth avail-
able. Keeping this in mind, Brightcove has developed a platform to allow vloggers to
distribute video content of all sizes over the Internet.
Discussion. A pioneering digital filmmaker, Kent Bye, presented an excellent idea for
creating a distributed, collaborative video-editing system, in which multiple individu-
als could actively participate in the putting together of a full-feature movie. We believe
that with the advances in interactive multimedia research, more powerful techniques
or tools might be developed to allow vloggers to collaboratively build more expressive
vlogs in the near future.
3.3.2. Vlog Aggregators.
Currently, vlog aggregators can choose from two conven-
tional means of displaying content: Web aggregators that make the view available in a
Web page, and desktop aggregators that deliver feeds through stand-alone, dedicated
software on a user’s desktop.
FireANT is a desktop vlog aggregator. By harnessing the power of RSS, FireANT
can automate the download and display the freshest video content in vlogs through
easy point-and-click interfaces, streaming media preview mode, and BitTorrent clients.
Moreover, FireANT integrates Yahoo! video search, and allows vloggers to tag their
audios or videos for easy categorization, search, and retrieval. Another example of
desktop vlog aggregators is Democracy Player (a.k.a. Miro) developed by PCF. Aim-
ing to make online video “as easy as watching TV, Democracy Player combines a
media player and library, content guide, video search engine, as well as podcast and
BitTorrent clients.
Founded in December 2004, Mefeedia is the first Web-based vlog aggregator, and is
also the first and most complete vlog directory on the Web. Through Mefeedia, vloggers
can easily find and watch videos in vlogs, and fill their video iPod or Sony PSP with
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:19
videos. In Mefeedia, a user can add tags to videos to help him or her as well as other
people to find videos more easily. Moreover, Mefeedia also provides some advanced
functions such as vertical video communities (i.e., categorizing and featuring popular
videos by topic, event or location); vlogger social networks (i.e., making vloggers aware
of which friends are watching); and personalized recommendations (i.e., recommending
new videos, shows, and channels based on vloggers’ personal interests and what they
have watched). By combining satellite imagery, e-maps, and Google Earth, VlogMap
is a fun and interesting vlog aggregator that shows where participating vloggers are
located around the world, along with links to key information about their vlogs. So far,
YouTube is the most popular online vlog aggregator. It fueled the growth of other video
aggregation destinations, like MySpace Videos, Revver, and DailyMotion. Today, these
sites represent general destinations that function primarily as diverse user-generated
and remixed content.
Discussion. In recent years, YouTube and similar vlog aggregators celebrated huge
commercial success. However, they are not necessarily navigation-friendly user en-
vironments. For this reason, a number of topic-specific and meta-aggregators have
emerged in the online video space. By adding a layer of filtering over the mass of
content, some niche or vertical aggregators (e.g., brand aggregators such as ON Net-
works and SuTree) can be developed to focus on a specific topic or an area of interest
[Kelley 2008]. On the other hand, with the sheer bulk of video that is becoming avail-
able, meta-aggregators can be developed by mixing “push” and “pull” models. Give the
preselected types of video content, the meta-aggregators can bring the most relevant
videos to vloggers by using content-based video filtering [Owens and Andjelic 2007].
This is also called prospective search [Irmak et al. 2006].
3.3.3. Vlog Search Tools.
Search engines are the most popular tools on the Web to
help users find the information they want quickly. Similarly, vloggers can utilize search
engines to easily find required vlogs and videos.
In general, most vlog directories or hosting sites are equipped with some kind of
search tools or services, which are either customized according to particular needs or
integrated with well-known search engines such as Google or Yahoo!. As mentioned
before, most vlog search techniques are rather simple, that is, they employ the tra-
ditional text-based retrieval techniques on RSS feeds of vlogs. Similarly, most of the
current video search engines (e.g., Yahoo! video search, Singingfish, Mefeedia, MSN
Video search) also search the text descriptions (i.e., metadata) of all videos in vlogs
for relevant results. Furthermore, some commercial search engines can extract and
index close-captions for video search. For example, Google Video searches the close-
captioning content in television programs and returns still photos and a text excerpt
at the point where the search phrase is spoken; Blinkx builds its video search engine
by combining speech recognition with context clustering to understand the actual con-
tent (spoken words) of video, thereby facilitating more semantic searches of video files
[Blinkx 2007]. Table III shows a simple comparison among five popular video search
engines from several aspects.
Discussion. Generally speaking, the video search engine is still in its infancy. Thus
we should investigate how to exploit the existing video content analysis and retrieval
technologies to facilitate video searching in the online vlogosphere. We address this
issue in the next section.
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:20 W. Gao et a l .
Table III. Features of Existing Main Video Search Engines on the Web
Feature Google Video Yahoo! Video MSN Video Blinkx Mefeedia
Navigation by Hot videos Hot videos Hot videos, 12 genres Topic, event,
Categories 20 genres 9 categories location
Metadata-
based Search
Close-caption × × ×
based Search
Content-based × × × Speech recognition, ×
Search phoneme mapping
Constraints of By for-sale/ By format, size By sources: By source channels ×
Search Results free or length length, domain MSN or Web
Result By relevance, By relevance,
By date By relevance By date
Ranking popularity, most viewed, or date
rating or date newest
Related Video × ×
of Results
Result Views List, grid, TV List, grid, TV List, grid, TV List, TV Wall List
Support of ×
Video Upload
RSS Support Media RSS, Media RSS RSS 2.0 RSS 2.0 RSS 2.0
openSearch RSS
Subscription ×××
Support
Multi-lingual ××
Support
Adult Video ×Do not show
Filtering thumbnails
Personalized × × ×
Settings
Search Assist ×× × ×
4. VLOGGING 2.0: ADDRESSING THE VALUE-ADDED AND INCUMBENT ISSUES
With better user experience than textblogs and more social networking features than
the traditional online video services, vlogs provide a new and rapidly evolving ap-
plication in the era of broadband and mobile multimedia. However, current vlogging
technology is essentially an understandable consequence of textblogging technology.
It is clear that with the evolution of the vlogosphere, vlogging technology will grow
synchronously.
To present a possible futurescape of vlogging technology, we would like to look at
the swarms of emerging user behaviors in online video and vlogging. According to
the study by Owens and Andjelic [2007], there are seven key behaviors relating to
online video and vlogging: creating personal taste, seeking peer pressure, gravitating
toward quality, searching for Web videos, plucking and aggregating, exploring social
scenes, and DIY (directing it yourself). It was also found that for shaping online video
behavior, immediate access, convenient navigation, and intelligent content discovery
appear to be central; vlogging, social interaction, and creative involvement make up
the other budding consumer trends. Keeping this analysis in mind, we argue that the
next-generation vlogging technology (called vlogging 2.0) should be characterized by
at least the following salient features:
scalability: to support scalable online video and vlogging services so as to meet the
different demands of vloggers (e.g., various video quality requirements and access
bandwidths);
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:21
interactivity: to support feasible user interaction in the form of interactive
video, such as deep linking [Lynch 2007], interactive navigation of video content
[Hammoud 2006];
searchability: to enable more sophisticated indexing and search of user-generated
video based on metadata, video content, and relevant deep links;
accessibility: to support any vlogger to access any format of online video in vlogs
using any device anywhere.
Moreover, vlogging 2.0 technology should also provide support for vloggers and vlog-
hosting providers to reduce potential copyright, moral, and legal risks.
Once realized, vlogging 2.0 technology will be immensely smarter and more relevant
to users, regardless of their devices or personal interests. Due to the efforts made
by researchers in multimedia, we have accumulated many valuable results, ranging
from content creation and understanding to user-centered delivery of and interaction
with videos. Some of these results can be introduced to help make this vision possible.
Similar discussions can also be found in Boll [2007], which covers what multimedia
research and Web 2.0 have in common, where the two meet, and how they can benefit
each other. Instead of a cursory overview of the convergence of the two broad fields,
we focus mainly on two key challenges that vlogging 2.0 should tackle with the help of
multimedia technology, that is, the “value-added issue” and the “incumbent issue” (see
Section 2.3).
4.1. Value-Added Techniques
We will describe several techniques that can potentially be used to support the scal-
ability,interactivity,searchability,andaccessibility of vlogging 2.0, including scalable
video delivery, interactive video, content-based retrieval, and user-centered content
adaptation. Finally, we will discuss how to flexibly combine these techniques and sys-
tems at all stages of vlogging via the mashing-up approach.
4.1.1. Scalable Video Delivery.
As mentioned in Section 3.2.1, FLV video has reached
near ubiquity in the current crop of online video and vlog sites such as MySpace,
YouTube, and Google Video. However, the great streaming efficiency of FLV videos is
obtained at the cost of greatly degraded quality. In other words, the service providers
can effectively reduce the running cost of network bandwidth by using low-quality
and low bitrate FLV videos; on the other hand, users must passively accept watch-
ing low-quality videos, even some user-generated videos are of higher-quality when
they were initially created and uploaded. To attract more users to vlogging, we must
nevertheless improve the quality of online video while retaining reasonable streaming
efficiency. This trend can also be evidenced by more and more high-definition videos
recently placed online.
With the rapid growth of network broadband and that the increasingly fierce compe-
tition for broadband market share, prices continue to decline while speed is on an ever
upward trajectory. On the other hand, online video delivery technologies and solutions
are continuously improve via optimizing scalable video codecs and developing central-
ized and/or peer-assisted delivery solutions, enabling more efficient video transmission
over the Internet. In short, it is now possible to distribute high-quality video efficiently
and economically.
Scalable video coding. Bandwidth and QoS characteristics of the varying broadband
networks suggest that online videos in vlogs will require variable, or scalable bitrate
encoding schemes that leverage network intelligence to dynamically tune bitrates
up or down. To effectively support scalable delivery of online video, there are three
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:22 W. Gao et a l .
Fig. 6. An illustration of various scalable coding schemes: (a) spatial scalability; (b) temporal scalability;
(c) quality scalability; and (d) attention scalability.
requirements the video codecs must meet: (1) Be scalable: online video should be able
to scale from the smallest encoding all the way to full high-definition quality, so as
to provide more flexibility in meeting different streaming demands (e.g., different ac-
cess bandwidths and latency requirements). (2) Real-time decoding: since streaming
video applications running on mobile devices or embedded in Web browsers must be
simple, real-time decoding with low decoding complexity is desirable. (3) Embedded
playable: video can be placed within HTML in today’s browsers by using some simple
JavaScript to load the video and control playback and interactivity using any HTML
user interface [Lynch 2007].
To address the scalability issue, various video coding techniques have been proposed,
including scalable video coding and multiple descriptive coding. In general, a scalable
coder generates a high-quality video bitstream consisting of multiple layers, achieved
through scaling frame rate, size, or quality (see Figure 6(a)–(c)); while a receiver, de-
pending on its capability, can subscribe to the base layer only with the basic playback
quality, or subscribe to additional layers that progressively refine the reconstruction
quality [Liu et al. 2003; Ohm 2005; Schwarz et al. 2007]. The early scheme was called
layered coding, used in MPEG-2 [Ohm 2005], in which the stand-alone availability
of enhancement information (without the base layer) is useless, because differential
encoding is performed with reference to the base layer. By data partitioning, the bit-
stream is separated into different layers, according to the importance of the underly-
ing elements for the quality of the reconstructed signal. To support even more flexible
scalability, a new form of scalability, known as fine granular scalability (FGS), was
developed and adopted by MPEG-4 [Li 2001]. In contrast to the conventional scalable
coding schemes, FGS allows for a much finer scaling of bits in the enhancement layer.
Slightly different from scalable video coding, a multiple descriptive coder generates
multiple streams (referred to as descriptions) for the source video, and any subset
of the descriptions, including each single one, can be used to reconstruct the video
[Castro et al. 2003; Padmanabhan et al. 2002]. The scalable and multiple descrip-
tive coding schemes provide two flexible solutions for transmission over heterogeneous
networks, additionally providing adaptability for bandwidth variations and error con-
ditions. However, scalable coding has efficiency concerns due to the iterative motion
estimation and transformation of all the layers, and often requires great computa-
tional power on the receiver’s side in order to assemble and decode multiple layers
[Ohm 2005].
To address the real-time decoding issue, low decoding complexity is desirable. For
example, Lin et al. [2001] employed a least-cost scheme to reduce decoding complexity.
However, the solution to this issue is more relevant to the implementation of decoders
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:23
by optimizing the decoding process or using decoding pipelining [Chen et al. 2005]. Be-
sides the lightweight implementation of real-time decoding, the embedded playability
can often be realized using Web-related programming languages such as JAVA.
Standardization is the key to interoperability and, by defining open interfaces, we
open the doors to seamless convergence of online video. At present there are four main-
stream video coding standards: MPEG-2, MPEG-4, H.264 (a.k.a. JVT, MPEG-4 AVC or
AVC), and AVS. The first three standards are specified by MPEG, while the fourth is
a Chinese independent formulation. Technologically, MPEG-2 is the first generation of
information source standards, and other three are the second generation standards.
MPEG-4 offers some FGS solutions [Diepold and M˝
oritz 2004], where the changing
network capabilities for video delivery can be taken into account in real-time by ap-
pending enhancement layers to the video stream in case of more bandwidth becoming
available. H.264 is another up-to-date video coding standard, which employs variable
block-size motion compensation for intercoding and directional spatial prediction for
intracoding to achieve high coding efficiency [Wiegand et al. 2003]. AVS is the abbre-
viation for “Advanced audio and Video coding Standard” (see http://www.avs.org.cn).
Specified by a standards workgroup led by our lab, AVS is a set of integrated stan-
dards that contains system, video, audio, digital rights management, and so forth. In
terms of coding efficiency, MPEG-4 is 1.4 times of MPEG-2, and AVS and H.264 are
more than twice of MPEG-2. Currently, MPEG-4 and H.264 have been widely adopted
in mobile phones, set-top boxes, game consoles, camcorders, and in Web video play-
back via clients such as QuickTime. As a newcomer, AVS has also been applied in
many commercial applications such as IPTV, internet live video, and mobile TV. These
coding standards are likely to become the optional video encoding formats for video on
the Web.
P2P video streaming. There are more and more vlogs that employ the Web-TV-like
presentation. Video embedded in Web browsers enables almost all PCs to view video
reliably and consistently in a completely transparent way [Lynch 2007]. Technologi-
cally, this will pose a significant challenge to supporting platforms when large-scale
high-quality videos are allowed to broadcast simultaneously. Due to its scalability,
peer-to-peer (P2P) technology is an appealing paradigm, providing video streaming
over the Internet for vlog service providers.
As mentioned above, P2P file-sharing technology such as BitTorrent has been ap-
plied in some vlogging tools such as FireANT to effectively enhance the file transmis-
sion performance. However, live video streaming faces several challenges that are not
encountered in other P2P applications [Liu et al. 2008; Locher et al. 2007]:
large scale: corresponding to tens of thousands of users simultaneously participating
in the broadcast;
performance-demanding: involving bandwidth requirements of hundreds of kilo-
bytes per second;
real-time constraints: requiring timely and continuously streaming delivery (al-
though minor delays can be tolerated through buffering; nevertheless it is critical
to get uninterrupted video);
gracefully degradable quality: enabling adaptive and flexible delivery that accom-
modates bandwidth heterogeneity and dynamics.
The early P2P streaming proposal was built upon a similar notion of IP multicast, re-
ferred to as application-level multicast or end-system multicast [Chu et al. 2000]. In
recent years, a large number of proposals have emerged for P2P video streaming, which
can be broadly divided into two categories [Liu et al. 2008]: tree-based and data-driven
randomized approaches. In tree-based approaches, peers are organized into structures
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:24 W. Gao et a l .
Fig. 7. An example of the P2P video streaming system, PPLive: (a) the system diagram; and (b) the peer
node architecture.
(typically trees) for delivering data, with each data packet being disseminated using
the same structure. When a node receives a data packet, it also forwards copies of
the packet to each of its children. Data-driven approaches contrast sharply, with tree-
based designs in that they use the availability of data to guide the data (e.g., gossip
algorithms) instead of constructing and maintaining an explicit structure for deliver-
ing data. In a typical gossip algorithm [Eugster et al. 2004], a node sends a newly
generated message to a set of randomly selected nodes; these nodes do similarly in the
next round, and so do other nodes until the message is spread to all. More insights
on P2P streaming technology can be found in two comprehensive surveys [Li and Yin
2007; Liu et al. 2008].
P2P video streaming is not only an active field of research, but there
are already practical systems and commercial products emerging, for exam-
ple, Coolstreaming [Zhang et al. 2005]; JumpTV (http://www.jumptv.com); PPLive
(http://www.pplive.com); SopCast (http://www.sopcast.org); and AVStreamer [Huo
2006]. Built in March 2004, Coolstreaming has been recognized as one of the earli-
est large-scale P2P streaming systems. It is based on a data-centric design, in which
every peer node periodically exchanges its data availability information with a set of
partners and retrieves unavailable data from one or more partners, while also sup-
plying available data to other partners [Zhang et al. 2005]. At its peak, it supported
80,000 concurrent viewers with an average bitrate at 400 Kbps with two PC servers.
Since then, PPLive became one of the largest P2P streaming systems, with more than
100 channels and an average of 400,000 daily viewers [Hei et al. 2006]. Figure 7 shows
the system diagram and peer node architecture of PPLive [Huang 2007]. JumpTV and
SopCast are similar commercial products that have attracted tens of millions of down-
loads and supported several hundreds of thousands of daily viewers. Developed by our
lab, AVStreamer is the first AVS-based P2P streaming media broadcast system, and
currently has been applied in China Netcom and Unioncast.tv.
Discussion. Video coding and streaming are possibly the two most active fields of
research and development currently. The success of Youtube confirms the mass mar-
ket interest in Internet video sharing, where scalable video coding and P2P stream-
ing serve as two underlying vehicles. For video coding, a promising research topic
is to perform scalable coding by assigning more bits for regions of interest (ROIs)
which might attract more user attention and less bits for other regions (as shown by
Figure 6(d)) [Lu et al. 2005]; while for P2P video streaming, probably the biggest prob-
lem is that it is still unclear what the appropriate revenue model for P2P broadcast
should be, with users scattered over the global Internet [Liu et al. 2008]. Nevertheless,
several popular P2P streaming services, such as PPlive and SopCast, are gradually
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:25
Fig. 8. Two interactive video diagrams: (a) interactive video via hyperlinks; and (b) interactive video via
key frames and summary.
moving into mainstream business, suggesting that the online video business will fi-
nally live up to its hype.
4.1.2. Interactive Video.
Interactive video is becoming the future standard for the most
attractive video formats, which will offer users nonconventional interactive features,
powerful knowledge-acquisition tools, as well as nonlinear ways of navigating and
searching [Hammoud 2006]. At this point, vlogging seems to be one of best applica-
tions for emerging interactive video technology. Here we adopt from Hammoud [2006]
the definition of interactive video, defined as a digitally enriched form of the original
raw video sequence, providing viewers with attractive and powerful forms of interac-
tivity and navigational possibilities. In this context, a vlog can be viewed as an in-
teractive video document that contains not only the raw video but also several kinds
of data in a time-synchronized fashion for enhancing the video content and making
the video presentation self-contained. This section will address the following two ma-
jor issues that concern different types of interactivity in vlogging (as illustrated in
Figure 8): (1) interactive video that enriches videos by hyperlinking; and (2) video
structuring and summarization for interactively navigating video content.
Video hyperlinking.Hypervideo (a.k.a., hyperlinked video) is a displayed video
stream that contains embedded, user-clickable anchors, allowing navigation between
video and other hypermedia elements [Smith and Stotts 2002]. For example, hyper-
video might involve creating a link from an object in a video that is visible for only
a certain duration, and making a deep linking that points not only to a video but to
particular time positions or markers in video streams (as shown in Figure 8(a)). The
initial concept regarding hypervideo can be traced back to HyperCafe [Sawhney et al.
1996], a popular experimental prototype of hypervideo that places users in a virtual
cafe where the user dynamically interacts with the video to follow different conversa-
tions. Since then, many interesting experiments have followed, and some hypervideo
products were developed to support interactivity in video content, such as HyperSoap
[Agamanolis and Bove 1997]; Advene [Aubert and Pri´
e 2005]; HotStream [Hjelsvold
et al. 2001]; HyperFilm [Pollone 2001]; VideoClix [VideoClix 2008]; MediaLoom [Tolva
2006]; and HotVideo [HotVideo 2008]. Nevertheless, the extensive application of hyper-
video has not happened yet, and the authoring tools are, at the moment, available from
only a small number of providers. This situation is rapidly changing, perhaps with the
unprecedented prevalence of online video sharing and vlogging. Hypervideo will in-
troduce a new set of interactivity options that give vloggers advanced interaction and
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
15:26 W. Gao et a l .
navigational possibilities with the video content, for example, directly commenting or
providing supplementary videos. Direct searching of videos could also be greatly facil-
itated by hypervideo models. Perhaps the most significant consequence of hypervideo
will result from commercial advertising [Owens and Andjelic 2007]. Hypervideo offers
the possibility of creating video clips where objects link to advertising or e-commerce
sites, or providing more information about particular products. This type of adver-
tising is better targeted and likely to be more effective (for further discussion, see
Section 6.1).
Compared to hypertext, hypervideo is challenging, due to the difficulty of video
segmentation, object detection, and tracking [Gelgon and Hammoud 2006]. Basically,
viewers can click on some user-clickable items of interest (called hot-spots) in hyper-
video to watch a related clip, which is often logically attached to objects or regions
within the video [Smith and Stotts 2002]. In order to automatically build hyperlinks
in a video, it is necessary to segment the video into meaningful pieces and to provide
a context, both in space and time, to extract meaningful elements from the video se-
quence. Unlike the nonsemantic segmentation that aims at extracting some uniform
and homogeneous regions with respect to visual properties, semantic video segmen-
tation is defined as a process that typically partitions the video images into mean-
ingful objects according to some specified semantics. There are many video segmenta-
tion methods in the current literature that exploit different kinds of information (i.e.,
spatial, temporal, spatio-temporal, and attention). These methods can be summarized
simply as the implementation of four phases: object-mask generation, postprocessing,
object tracking, and object-mask update [Li and Ngan 2007]. For more details, see Li
and Ngan [2007] and Zhang [2006].
Once the required nodes have been segmented and combined with the associated
linking information, the metadata must be incorporated into the original video for
playback. The metadata is placed conceptually in layers or tracks in a video stream.
Several existing standardization efforts in multimedia already allow annotations and
hyperlinks to be attached to media resources. The first is the W3C’s Synchronized Mul-
timedia Interaction Language (SMIL) 2.0 [SMIL 2001], an XML-based language that
allows authors to write interactive multimedia presentations. SMIL has outgoing hy-
perlinks and elements that can be addressed inside by using XPath and XPointer. In
the strict sense, however, a SMIL document is not a single, temporally addressable
time-continuous data stream [Pfeiffer et al. 2005]. The second effort is the Continuous
Media Markup Language (CMML), which is an XML-based markup language for au-
thoring annotation tracks for time-continuous data. With timed metadata and outgo-
ing hyperlinks, it can be used to create both free-text annotations and structured meta-
data to describe a video’s content, and can be easily serialized into time-continuous
frames. For synchronized delivery of the markup and the time-continuous data over
the Web, a streamable container format, called Annodex, was developed to encapsu-
late CMML together with the time-continuous resource [Pfeiffer et al. 2003]. Recently,
MPEG has specified a new standard, Lightweight Application Scene Representation
(LASeR) [LASeR 2007], as a scene description format for lightweight embedded de-
vices such as mobile phones. The LASeR specification defines a LASeR engine which
has rich-media composition capabilities based on a scalable vector graph (SVG) Tiny
1.1 and enhanced with key features for mobile services (e.g., a binary encoding, dy-
namic updates, fonts, and so on). To multiplex the scene stream and the synchronized
media, a simple aggregation format (SAF) is provided as a streaming-ready format
for packaging scenes and media together and streaming them onto such protocols as
HTTP, TCP, MPEG-2, and so on [LASeR 2007].
With the hypervideo representation, there are some hypervideo authoring works
(e.g., Gelgon and Hammoud [2006] and Smith and Stotts [2002]) that focus on the
ACM Computing Surveys, Vol. 42, No. 4, Article 15, Publication date: June 2010.
P1:IAZ
CSUR4204-15 ACM-CSUR May 9, 2010 18:19
Vlogging: A Survey of Videoblogging Technology on the Web 15:27
automated markup of hyperlinks in a real-time video stream as well as stored video.
By utilizing the hypervideo model, the MUMMY project developed a mobile interac-
tive video tool that could be used to exchange and retrieve knowledge among mem-
bers of a community in a mobile scenario [Finke 2004]. Some hypervideo authoring
tools and systems were also developed to semi- or automatically create hypervideo
content, including OvalTine [Smith and Stotts 2002]; Hyper-Hitchcock [Girgensohn
et al. 2003]; VideoClix [VideoClix 2008]; Asterpix (http://www.asterpix.com); Adivi
(http://adivi.meticube.com); AdPoint (http://www.adpointonline.com); and so on.
Video structuring and summarization. Another type of interactive video technology
is to use video structure analysis and summarization to provide a representative short
summary of the video prior to downloading or watching it, or to present a list of vi-
sual entries (e.g., key frames) that serve as meaningful access points to desired video
content, as opposed to accessing the video from the beginning to the end [Hammoud
2006]. The two interactive video forms are useful for vloggers to preview and navigate
video content in a very efficient nonlinear fashion.
For efficient access of video data, parsing video structure at different granularities is
the first important step. Generally speaking, video can be hierarchically represented
by five levels: key frames, shots, scenes, story units, and video. Hence video struc-
ture analysis typically involves four steps: shot boundary detection, key frame extrac-
tion, scene analysis, and story unit segmentation. As the fundamental element above
the frame level, the video shot is often defined as a series of interrelated consecutive
frames taken contiguously by a single camera and representing a continuous action in
time and space [Rui et al. 1999]. Therefore, to build up the semantic table-of-contents
structure for the video, the first step is to locate every shot by finding its beginning
and end, which is often called the shot boundary detection task. In general, the video
creators may