Wavelet based image compression
Talking nowadays about video data compression, MPEG is one of the most importand and well known algorithms for data reduction. After MPEG-1 and 2, the H.264 (MPEG4 Part 10) standard is the new method to achieve higher compression rates at the same image quality. Within all these standards time predictors are the main aspects to get high data reduction as done for one direction (IP coded) or bidirectional (IPB coded). The coding mechanisms of all the described methods have one common module: the block based cosine transformation.
A new algorithm based on the wavelet transformation was develeoped and is called JPEG2000. Wavelets are used as filters on the whole image data and doesn't need to be split into code blocks why blocking artefacts are eliminated. At high compression rates other artefact occure like blurring. The JPEG2000 standard is applicable for still images as well as for HDTV, where its skalable image size may be useful. For still images the target application can be the internet and for video application the professional film studios. In both cases the skalability may be used to generate preview images without reencoding. Another skalability with the help of wavelets can be the skalability in time to generate a video with a high quality and a hight frame rate, where the video may be displayed in different solutions or frame rates when the original video is encoded only once.
The goal of our research is to detect and compensate the side effects and artefacts at high compression rates for JPEG2000 with adequate methods. Within the development of our own standard conform software and hardware models, new methods should be found to make the motion JPEG2000 system comparable to the MPEG systems. Additional aspect are the implementation of the JPEG2000 encoding algorithms in embedded systems (e.g. DSPs or FPGAs) where new and less cost intensive algorithms are needed. Possible rearangements of moduls as well as developing new algorithms are also part of this research work to optimize the whole system.