Compression algorithm based on segmentation is one of the most perspective directions in image compression, since it conforms to human visual system and information theory. In this thesis image segmentation and texture synthesis are adopted to improve and complement current compression standards. The research fields involved in are image segmentation, texture synthesis and wavelet coding based on adaptive shape. Image segmentation includes two aspects in this thesis, i.e. region segmentation and texture segmentation. There are many existing methods about the former, but no one can deal with comprehensive segmentation by itself. Consequently, one integrated algorithm is adopted, which makes use of strongpoint of different methods. Wavelet textons are described in texture segmentation. And then, statistics and vector field are applied respectively in order to segment texture images from coarse to fine. There are tow problems about texture synthesis. One is how to select samples from a large texture. A color structure is defined as features of each pixel at first. Subsequently, fuzzy clustering is applied to simplify types of color structures. Histogram matching algorithm is presented in order to choose suitable samples of textures. The other problem is texture synthesis. EMD algorithm is introduced for extracting texture structure of samples. And then, we synthesized texture structure with Patch-based sampling and MRF. Wavelet coding based on adaptive shape is presented in MPEG-4. We classify images to smoothing region and texture regions. The algorithm similar to standard method in MPEG-4, is applied for texture regions. To reduce Gibbs effect, we code smoothing regions with a new method based on diffusion equation. Compression algorithm based on segmentation is compatible to current compression standards. Compared to existing algorithm, the algorithm here is suitable to images containing large texture regions and has higher compression ratio. At the same time, code stream can be distributed optionally so that image quality can be guaranteed in important regions
修改评论