Mahmoud R. El-Sakka, "Adaptive Digital Image Compression Based on Segmentation and Block Classification", Ph.D. Dissertation, Systems Design Department, Faculty of Engineering, University of Waterloo, Canada,1997.
Over the last few decades, many good image compression schemes have been developed. The performance of these schemes varies from low to high compression ratios with low to high levels of degradation of the decompressed images. Since the end users of decompressed images are usually human beings, consequently, it is natural that attempts should be made to incorporate some of the human visual system properties into the encoding schemes to achieve even further compression with less noticeable degradations.
This thesis presents a new digital image compression scheme which exploits one of the human visual system properties---namely that of, recognizing images by their regions---to achieve high compression ratios. It also assigns a variable bit count to each image region that is proportional to the amount of information it conveys to the viewer. The new scheme copes with image non-stationarity by adaptively segmenting the image---using quad-trees segmentation approach---into variable-block sized regions, and classifying them into statistically and perceptually different classes.
These classes include, a smooth class, a textural class, and an edge class. Blocks in each class are separately encoded. For smooth blocks, a new adaptive prediction technique is used to encode block averages. Meanwhile, an optimized DCT-based technique is used to encode both edge and textural blocks.
Based on extensive testing and comparisons with other existing compression techniques, the performance of the new scheme surpasses the performance of the JPEG standard and goes beyond its compression limits. In most test cases, the new compression scheme results in a maximum compression ratio that is at least twice of JPEG, while exhibiting lower objective and subjective images degradations. Moreover, the performance of the new block-based compression is comparable to the performance of the state-of-the-art wavelet-based compression technique and provides a good alternative when adaptability to image content is of interest.