Image tile pyramid with 5 levels. In this example, a 16 MP image is tiled into 341 sub-images of 256 × 256 pixel size. 

Contexts in source publication

Context 1

... images. Its goal is to help scientists discover new phenomena and link related data sets from various instruments that are currently often analyzed in isolation. In addition, it will make a huge amount of information available to the general public by visualizing it in intuitive and appealing ways. The current number of data browsing tools is limited and each offers very specific functionality. For example, SOHO’s web-accessible Near Real-time Image Browser [1] is serving over twelve years of heliophysics data as JPEG images in two sizes: 1024 × 1024 and 512 × 512 pixels. The SOHO Movie Theater [2] uses the same data to create on-screen animations with basic movie control functionality. The Solar Weather Browser [3] allows quick image browsing of highly compressed data and can handle up to two overlays. All of these widely used tools work well with current data volumes and met many of the current browsing needs but will be severely challenged in the immediate future. To address the new challenges, Jhelioviewer has been built to provide an integrated solution for image encoding, storage and dissemination. Its advanced browsing capabilities (e.g. movies with arbitrary time cadence, unlimited overlays, image processing, event data integration etc.) underscore the superiority and novelty of this approach. The focus of this paper is the application of a relatively new image compression technique that is both established and well known in the digital image processing community to the challenging task of the solar physics community to cope with the ever-growing amount of data available. While the impetus for this article is handling solar physics data, similar requirements of accessing, browsing and searching image data exist for applications in other areas such as the earth sciences and medical research [4, 5, 6]. In the following sections we will: compare the hierarchical structure of JPEG 2000 files to the popular image tiling approach for rendering and visualization of large image data sets over networks and give a brief overview of the JPEG 2000 compression standard. We will then outline the specific requirements that led to the development of Jhelioviewer and describe the architecture and implementation of the system. We will conclude by giving an outlook on possible future uses and extensions of the project. A popular technique to handle image rendering and visualization of large data sets over networks is image tiling. For this method, each original image is divided into sub-images, or tiles, at various resolution (zoom) levels, thus creating a pyramid of image tiles for each image (Figure 1). Used by many providers of geospatial data, e.g. Google TM and Mapquest TM , this approach has the advantage of transferring only data for the chosen region of interest (ROI) and zoom level as image tiles from the server to the client. While this method works well for data repositories with only a few files, it has ...

Context 2

... disadvantages when the number of files increases. • The number of tiles increases as a power of the number of resolution levels: It is given by the finite geometric series k n =0 − 1 z k = (1 − z n ) / (1 − z ), where z is the number of tiles each sub-image is divided into, to create the next resolution level, and n is the number of resolution levels. Even for a modest image size of 16 MP divided into tiles with a size of 256 × 256 pixel at five resolution levels, the tiling approach increases the number of files to be stored by a factor of 341 and the total data volume by at least about a factor of two (the total overhead in file size depends on the compression rate and the image content). SDO’s Atmospheric Imaging Assembly (AIA) 3 instrument will take 16 MP images of the Sun in 8 spectral channels at least every 10 seconds, i.e. on the order of 70,000 images per day or 30 million files per year. For data sets of this magnitude the number of tiles is staggering. Even for a modest fraction of the data, generating tiles becomes prohibitive. • Typical use cases for image browsing involve repeated zooming in and out of different ROIs. For each zoom level, a new set of image tiles has to be transferred from the server to the client. This method uses significantly more network bandwidth than necessary as it fails to exploit the fact that part of the information contained in the image has already been transferred at a different zoom level. Clearly, image tiling is not a sufficent solution. A method that eliminates the need for tiling is the discrete wavelet transform (DWT), an image transform that is the basis of the JPEG 2000 image compression scheme. The JPEG 2000 standard [7] was created by the Joint Photographic Experts Group (JPEG) with the intention of improving upon, and superseding, the very successful JPEG standard [8] that has been in use for almost 20 years. It is a novel image compression ISO standard that offers both a lossless and a lossy compression mode and provides many new features, which make it a promising format for handling massive amounts of image data and associ- ated meta data. Images have to be encoded only once in the highest desired quality and can subsequently be decoded in many ways to extract subfield images with a chosen spatial resolution, level of detail and region of interest. This offers significant advantages compared to storing multiple versions of images or tiles for different resolution levels and drastically reduces the size and complexity of storage and network transmission requirements. Figure 2 shows the JPEG 2000 equivalent of the image tile pyramid of Figure 1: Level 1 (L1) represents the image encoded at the highest quality and ...