Alexei Yavlinsky's research while affiliated with Imperial College London and other places

Publications (26)

Article
This paper presents a framework for improving the image index obtained by automated image annotation. Within this framework, the technique of keyword combination is used for fast image re-indexing based on initial automated annotations. It aims to tackle the challenges of limited vocabulary size and low annotation accuracies resulting from differen...
Conference Paper
We describe an online application that allows users to pro- vide similarity judgements whilst browsing a collection of 60,000 photographs. One immediate goal is to modify the initial browsing structure in response to the feedback. We thus suggest a long-term relevance feedback technique that integrates user information over multiple sessions. The p...
Article
This short report outlines the indexing techniques and the user interface of a prototype image search engine called Behold. It com-plements traditional image metadata with automatically generated im-age annotations and an image similarity network to make images more accessible. The user supplies the initial query in the form of one or more keywords...
Article
This paper outlines a system for searching and browsing 1.14 million images from the World Wide Web (WWW) based on their visual content. At the heart of the system lies an automatically constructed network of images that can be navigated quickly by following its edges. The browsing experience is enhanced in a number of ways including multidimension...
Conference Paper
This paper outlines the technical details of a prototype system for searching and browsing over a million images from the World Wide Web using their visual contents. The system relies on two modalities for accessing images — automated image annotation and NNk image network browsing. The user supplies the initial query in the form of one or more key...
Conference Paper
This paper summarises the scientific work presented at the 28th European Conference on Information Retrieval and demonstrates that the field has not only significantly progressed over the last year but has also continued to make inroads into areas such as Genomics, Multimedia, Peer-to-Peer and XML retrieval.
Conference Paper
Keyword-based queries are an important means to retrieve information from XML collections with unknown or complex schemas. Relevance Feedback integrates relevance information provided by a user to enhance retrieval quality. For keyword-based XML queries, feedback engines usually generate an expanded keyword query from the content of elements marked...
Conference Paper
This paper describes an efficient method to construct reliable machine learning applications in peer-to-peer (P2P) networks by building ensemble based meta methods. We consider this problem in the context of distributed Web exploration applications like focused crawling. Typical applications are user-specific classification of retrieved Web content...
Book
th These proceedings contain the refereed papers and posters presented at the 28 Annual European Conference on Information Retrieval (ECIR 2006), which was held at Imperial College London in South Kensington between April 10 and 12, 2006. ECIR is the annual conference of the British Computer Society’s Inf- mation Retrieval Specialist Group. The eve...
Article
Full-text available
We describe our experiments for the high-level feature extraction and search tasks. For the search task, we tested the system we have used in previous years, which encapsulates content based image search, image browsing, automated image annotation and named entity extraction. For the feature task we apply the nonparametric density estimation model...
Conference Paper
Full-text available
We describe our experiments for the shot boundary detection, high-level feature extraction and search. In the shot boundary detection task, we employ our proven method based on the calculation of distances between colour histograms of frames over a range of timescales. For the search task, we tested a different system. This year, content based sear...
Conference Paper
This paper describes a simple framework for automatically annotating images using non-parametric models of distributions of image features. We show that under this framework quite simple image properties such as global colour and texture distributions provide a strong basis for reliably annotating images. We report results on subsets of two photogr...
Article
We describe our experiments for the Image CLEF medical retrieval task. Our efforts were focused on the initial visual search. A content based approach was followed using global features that would perform well with the medical image collection. We used structure, texture, localisation and colour features that have been proven by previous experiment...
Conference Paper
We describe our experiments for the Image CLEF medical retrieval task. Our efforts were focused on the initial visual search. A content-based approach was followed. We used texture, localisation and colour features that have been proven by previous experiments. The images in the collection had specific characteristics. Medical images have a formula...
Conference Paper
Summary from only given. We present an image library with similarity-based searching and novel browsing structures and a news video library with content-based search and summarisation features. Our approach puts the traditional library search via textual meta-data upside down by allowing access to collections through browsing and search of images t...
Conference Paper
The paper reports on experimental results obtained from a performance comparison of feature combinations strategies in content based image retrieval. The use of support vector machines is compared to CombMIN, CombMAX, CombSUM and BordaFuse combination strategies, all of which are evaluated on a carefully compiled set of Corel images and the TRECVID...
Article
Full-text available
We describe our experiments for the shot boundary detection, high-level feature extrac-tion, search and story boundary detection tasks. In the shot boundary detection task, we employ our proven method based on the calculation of distances between colour histograms of frames over a range of timescales. In the feature detection task we confined our e...
Conference Paper
We present an image library with similarity--based searching and novelbrowsing structures and a news video library with content--based search andsummarisation features.
Conference Paper
This paper reports on experimental results obtained from a comparative study of retrieval performance in content-based image retrieval. Two different learning techniques, k-Nearest Neighbours and support vector machines, both of which can be used to define the similarity between two images, are compared against the vector space model. For each tech...
Article
This book constitutes the refereed proceedings of the 28th European Conference on Information Retrieval Research, ECIR 2006, held in London, UK in April 2006. The 37 revised full papers and 28 revised poster papers presented were carefully reviewed and selected from 250 submissions. The papers are organized in topical sections on formal models, doc...
Article
This work investigates the application of Support Vector Machines (SVMs) and a number of feature selection methods to Content Based Image Retrieval (CBIR) based on feature generation schema proposed by Tieu and Viola in 2000. This study largely builds on the work published by Pickering M. and Rüger S., which investigates a number of approaches to C...

Citations

... Another concept is the dual modality query scheme, where a combination of text and image example is submitted as query. In (Yavlinsky, 2007), the author proposed to solve the semantic gap problem based on Automatic Image Annotation (AIA). A learning model that depends on the correlation between the low-level features and the training annotation words was developed and then machine learning techniques were employed to automatically apply annotations to new images using the learned model. ...
... In SL0, sparse Homotopy, sparse PALM, and HER (Li et al., 2015), Heesch et al. (2004) and SCFV (Araujo et al., 2015) methods in the CC_WEB_VIDEO dataset have been shown. In this experiment, the best performance 84.82%. ...
... The last step in our multimedia processing pipeline was to automatically extract keyframes from the MP4 video. Rather than naively extracting keyframes at some fixed period, we elected to use the content analysis based Hive2 automatic keyframe extractor [6]. Hive2 is able to detect scene changes ranging from sudden 'cuts' through to more gradual 'fades', and extracts the keyframes as JPEG images with the same dimensions as the video. ...
... Comprehensive surveys of the automatic image annotation methods are available at [11] and [12]. On-line image annotation retrieval systems are available at [5] and [13]. The proposed method employs the Joint Composite Descriptor (JCD) [14] and utilizes two sets of keywords in order to map the low-level features of the descriptor to the high-level features constituted by these keywords. ...
... Deve-se notar que o radiologista continua interpretando a imagem e sendo responsável pelo laudo, com o CAD apenas auxiliando-o [9]. Um sistema CAD processa imagens para: 1) melhorar a qualidade [10] (aumentar o contraste, por exemplo); 2) segmentar [11] (delimitar os achados na imagem); 3) parametrizar [12] (calcular atributos para cada um dos achados – área, perímetro e forma, por exemplo); 4) reconhecer [3] (classificar os achados); e 5) analisar [13] (associar os parâmetros calculados ao problema dado). ...
... In a user study they found out that the most popular view was the storyboard view, which is a similar result as already found by Komlodi and Marchionini [1998]. Heesch et al. [2004] presented a tool for video retrieval and video browsing (Figure 2.3). The tool allows searching and browsing a video in different dimensions in a storyboard manner. ...
... To provide temporal browsing they also use a fisheye visualisation at the bottom of the window (5) in which the image-of-interest (selected on any view) is always shown in the center. Heesch et al. [2004] An extension of this approach is introduced by Ghoshal et al. [2006]. Their interface, shown in Figure 2.4, is split into two main panels with the browsing panel taking up to 80% of the screen. ...
... Our Multimedia Information Systems (MMIS) group at the Knowledge Media Institute (KMi) conducts research in the area of multimedia information retrieval in-cluding content-based search, automatic image annotation and video shot-boundary detection. MMIS and previously the Multimedia Group at Imperial College London, also led by Stefan Rüger, have participated in both TRECVid and ImageCLEF tasks since 2002 (Pickering and Rüger, 2002;Heesch et al, 2003Heesch et al, , 2004Howarth et al, 2005;Jesus et al, 2005;Magalhães et al, 2006;Overell et al, 2006Overell et al, , 2008Llorente et al, 2008Llorente et al, , 2009Zagorac et al, 2009). ...
... On the other hand, the late fusion strategy is based on combining multiple retrieval results, each of which is based on one specific feature [19]. Conventional approaches include such as the Borda count, CombMIN, CombMAX, CombSUM, and CombMNZ algorithms [1,17,22,24] (c.f. Sect. ...
... Furthermore, annotation of personal photos requires a higher level of descriptive annotation of people's activities. This is beyond the scope and capability of classic image retrieval systems [107], [108], [109] [110]: most of these systems were designed for simpler data such as the popular Corel image database and typically only provide simple image semantics (such as sunset, mountain, and lake), while the photos taken by personal cameras are much more complex and involve different people and various activities (such as beach time, birthday parties, wedding, and graduation). ...