A Study Of The Lossless Wavelet Color Image Coding Based On Set Partition

Today, multimedia technologies have been widely used and have become the characteristics of the times of the computer system. Image is the most important part of multimedia information, but it requires large space to store and spends large costs to transmit, it is necessary for image data to compress. In real life, color images contain more information and meet more demand of people, so the application of color images will become more and more popular. However, color images need a much more storage space. So color image compression is very important. In addition, some color images need for lossless compression, such as fingerprint images, medical images, aerial images and satellite remote sensing images. Therefore, lossless compression of color images has a certain value of research.This paper firstly describes the development of digital image compression technology. Then in the second chapter, some commonly image coding technologies, especially transformed coding technologies are described. Chapter III studies the image compression based on set partitioning theory and expounded the principle of set partitioning coding, and gives some instances which describe the set partitioning process of the transformed image coefficients to explain the theory in detail. Set partitioning coding must be combined with a spatial scanning order to demonstrate its efficiency, so a spatial orientation tree structure and a quadtree block structure are usually used. SPIHT and SPECK are two kinds of examples of algorithms which efficiently use the above two kinds of structures.Experimentally, it has been noticed that SPIHT consumes an important amount of bits to encode direct descendants although they are not significant during coding of different test images. Therefore, it wastes bits in the initial iterations as it encodes the four direct descendants of each set that has been found significant with respect to the current threshold. In fact, if new sets are defined to perform a separate sorting of the sets that have insignificant children, we only need one bit. Obviously, it will be more efficient if there are more nodes whose descendants are significant and direct descendants are insignificant, it is one side. On the other side, in one bit plane, the number of significant sets is much higher than that of insignificant sets. Thus, all significance tests on the entries can be removed so that every set is directly partitioned. As a result, one bit is saved for each set if it is significant. If not, three extra bits are transmitted since the resulting four sub-sets will be evaluated against the current threshold. Based on the above considerations, an improved SPIHT algorithm is proposed in the fourth chapter of this paper, for the first point to add a new type to mark the nodes whose descendants are significant and direct descendants are insignificant, so that it can be processed separately. As regards to the second point , the number of all types of significant descendants and insignificant descendants will be counted separately at each bit plane, according to the number an optimal switching bit plane is selected. After this bit plane, all significant tests will be removed and directly partitioned, then we will save some bit.In order to achieve lossless compression of color images, this paper conducted a reversible color image YUV space conversion, and each component of the YUV space is independently applied with reversible S + P wavelet transform and is applied with improved SPIHT algorithm. Compared with the original SPIHT algorithm, the improvements obtained for grey level images vary between 0.03 and 0.16bpp while the improvements obtained for color images are in the range 0.25 and 0.50bpp.