The 'TeleVision' project (IWT project IT GBO 202) targets generic research on using images in high-end applications (e.g. multi-media and telematics applications). The project is a close collaboration between the major Flemish university departments that are active in the domain of image processing (KUL, VUB, UA, IMEC, UG, LUC). The goal of the project is to acquire the basic knowledge and competence necessary to combine multimedia and communication in a number of high-end applications. Six key technologies (generic research fields) are being investigated : Hypermedia Authoring and Annotation Management, Image Analysis, Interaction and Presentation, Transmission and Cooperative Work, Compression, System Exploration and Prototyping. The activities of the VUB ETRO/IRIS department in this domain are mainly situated in the two last key technologies the goals and the research topics of these key technologies are listed below (the respective contributions of the partners VUB, KUL, RUG and IMEC are indicated) : COMPRESSION: Goals : In this workpackage, we conceive and explore new compression methods for images and image sequences in order to obtain the high compression ratios needed in interactive tele-multimedia applications. Depending on the bandwidth, the quality of the connection and the application, different approaches are needed. Currently a number of different application areas have their own standards. Research will be done to identify the image type and characteristics in order to identify the optimal standardized compression scheme and corresponding parameters. For other applications, such as high-quality color images, image sequences for surveillance and mixed image data like video images with graphical annotations, there exists no standard yet. Traditional methods for low-bitrate video coding are even insufficient and need further research. In this project we focus on new region-based (RB) algorithms for low bitrate color images and image sequences. Aspects such as efficient color image segmentation and motion estimation for RB compression are investigated. Progressive coded image transmission will also be studied via the approaches of region-based (RB) compression, but will be compared to a new multiresolution compression scheme based on Wavelet subband coding. Active participation in the MPEG-4 standardization committee will ensure that the research proceeds along the lines of ongoing standardization activities. SYSTEM EXPLORATION AND PROTOTYPING: Goals : The objective of this workpackage is to develop and apply a system design methodology which allows to specify video and image processing applications (especially video compression and image analysis) at the system level and to evaluate them effectively in terms of desired functionality and optimized realization cost. A fast prototyping environment is developed for testing the required functionality of the most promising algorithms when applied on real-life data without having to rely on slow simulation based on workstations. For this purpose, a hardware environment will be used which is based on programmable DSP's. In this task, the necessary tools will be developed to efficiently map a given application to this hardware set-up. All this work will be performed with a tight link to application drivers which are used to demonstrate and validate the project results. These include: (1) 2D wavelet transform coding kernel for image compression and transmission, (2) an image analysis module to extract motion. RESEARCH TOPICS : 1. Region based color video codecs (VUB - ETRO/IRIS and RUG - ELIS/Medisip): Research on new region-based codecs for color images and color video. Very high compression ratios are aimed at (a codec is a complete compression/decompression scheme). We have consolidated the results on compression of still color images we now proceed and start to tackle the compression of color video (motion modelling & estimation) 2. Image compression characteristic for adaptive region based coding / motion characteristic for compression (KUL, ESAT-MI2) : Research is ongoing on identifying the image type and characteristics of (parts of) images, in order to identify the most appropriate standard compression scheme which should be applied on these (parts of) images: e.g. JBIG for line-art graphics, JPEG for photographic images. Later, the research will focus on the motion characteristic for compression 3. Codecs based on the wavelet transformation (VUB - ETRO/IRIS and IMEC - VSDM) : Conception and elaboration of a codec based on Wavelet subband coding. The codec will support progressive transmission (still images) and adaptivity to varying channel bandwidth (video). The final goal is to deliver algorithmic specifications for a complete codec (real-time), comprising Wavelet transformation, quantization and motion vector coding. Whereas the algorithmic explorations are part of the work package "COMPRESSION", the optimization aspects for the mapping of the Wavelet transformation on VLSI architectures, and the methodology for doing this, belong to the work package "SYSTEM EXPLORATION AND PROTOTYPING" (collaboration VUB - IMEC) 4. Efficient system exploration of data storage and transfer cost (IMEC - VSDM and VUB - ETRO/IRIS) : Development of a system design methodology, which allows to specify video and image processing applications at the system level, and to evaluate them effectively in terms of desired functionality and optimized realization cost. Concretely, it is aimed at optimizing real-time implementations of these applications in terms of area and power. With the eye on validation, this work will be performed with a tight link to application drivers, e.g. with the mapping of the Wavelet transformation on VLSI architectures (see research topic 3) 5. Fast prototyping of video processing algorithms on MVPs (KUL - ESAT/ACCA) : A fast prototyping environment will be developed where the required functionality of video and image processing algorithms can be tested out on real-life data without having to rely on slow simulation based on workstations. For this purpose, a hardware environment will be used which is based on a programmable video processor, the MVP/C80 of Texas Instruments. In this task, the necessary tools will be developed to efficiently map a given application to this hardware set-up. The driver application will be an adaptive image compression scheme which uses the algorithms of research topic 2 for identification of image type and characteristics, and consequently applies the selected standard compression schemes (JPEG, JBIG) on (parts of) the images.

Runtime: 1995 - 1998