The Research And Hardware Design Of Image Compression Coding Based On Lifting Wavelet Transform

Wavelet transform is a new theory, which is an important result on the history of the mathematics development. It has a profound impact on works and the application of mathematics. It is a core algorithm of the latest JPEG2000 image compression standard.The discrete wavelet transform (DWT) has been applied extensively to digital image coding Because of a good many choiceness characteristic. A lot of scholars have been done a great deal of researches in image coding based on the DWT. But the research of the hardware implementation relatively less, mobility difference, and the technology of special-purpose chip is something to keep secret. Therefore, it is very meaningful to design the special-purpose DWT image coding chip.In this paper, a more detailed analysis of the theoretical basis of wavelet transform, multiresolution analysis, Mallat algorithm and lift algorithm are introduced. Then Analysis of a wavelet filter used in Jpeg2000, and introduced a new LS97 wavelet filter. Which has simple wavelet coefficients, easy to implement, and has good compatibility with CDF97 wavelet. CDF97 wavelet can be used as a substitute. After determining the wavelet used, a VLSI of two-dimensional discrete wavelet transform was designed. Data for the border, using a mirror symmetric extension technology, no additional cache, the hardware resource is conservated. After image wavelet transform, need to quantify each sub-band, meaning that the number of spaces to express wavelet coefficients. Which often quantified as damage data, But in the circumstances to maintain a certain PSNR it can greatly reduce the resource for storage and data compression. Less byte can be used to represent the raw data. EZW algorithm is a good ompression algorithm which is based on scalar quantization. SPIHT algorithm is an extension of EZW. Set on the broad introduction of zero-tree split effective method and get a high coding efficiency. Even without arithmeticcoding it can get the same or even better effect than EZW coding algorithm. SPIHT wavelet encoding is a good choice for the design of hardware.But SPIHT includes three linked list to store the chain and set of pixel coordinates.for the reality of hardware, such as insertion and deletion of a chain operation expenses is high. On the other hand, the storage of image coordinates information required too much memory space, which is not conducive to the realization of real-time encoding. A Frederick W and W Pearlman proposed a no list SPIHT (No List SPIHT, NLS) algorithm, which use non-recursive structure, properties and SPIHT, However, the required storage space than SPHIT is greatly reduced, which needs about 248KB to store the coefficients for the 12 levels quantified. Lin K W and Burgess proposed a listless zero coding in 1998 (Listless Zero tree Coding, LZC) LZC tree algorithm use only two symbols C(i,j) and D(i,j),it occupied a smaller memory, But the scanning of LZC limit coding speed . NLS and SPIHT algorithm, using the same three processes, but the abolition of the chain structure, reduced memory requirements, made a fast encoding speed. NLS algorithm uses the same size as the original image with the state table Mark[i] Band replaced 3 chains of SPIHT, Mark state signs can be further simplified. Both MD and MG signs can be used for one-sixth and one-fourth the size of the original image size of the matrix to show. Significantly higher coefficient signs MIP, MSP MNP could be merged into a symbol FC according to the LZC algorithm district thinking. Optimization of SPIHT above zero SPIHT can simplify the coding process. Significantly make memory requirements lower and speed up the encoding. SPIHT algorithm can be simplified. If the signs were used to separate the coding can reduce memory requirements; Even RP IPP process could be combined; MNP MSP signs and the NLS algorithm signs can be combined into a significant marker.The main work and content are sum up as follows:Chapter 1, It is briefly introduced the basic knowledge of image coding. Such as the possibility and necessary of image compression coding, the type of data redundancy, the classic image coding and the modern image coding, andintroduced history and present research situation of image coding. At last, the evaluation criteria used in image compression is introduced.Chapter 2, the basic knowledge of wavelet encoding is introduced, such as the mathematical definition of wavelet transform, multiresolution analysis and so on. Finally, the author studied the embedded coding classical algorithm: EZW and SPIHT, and analyzed the progress of the algorithm.Chapters 3, the two wavelet filters of JPEG2000 are studied: Le Gall 53 reversible and irreversible CDF97 filter filter. Then, a new simple LS97 is deeply studied. This has simple coefficients, easy for the reality of VLSI, and its performance of image compression with CDF97 are almost the same. Both have good compatibility. The filter used in this study t is determined. Chapter 4, the lifting coefficient of LS97 wavelet fiter is analyzed. In light of the difficulties in achieving Hardware float, the computation method was adapted to lifting wavelet coefficients.amore simple and flexible implementation of the extension is used. Pipeline ranks of the parallel processing improve the speed of the processing. The structure of the line and row transform is desiged, and analyzed its function.at last, for the need of following coding, the author anylyzed the manner of the wavelet coefficients storage.Chapter 5, The SPTHT algorithm based on the wavelet transform has the high performance on the still image compression. An improved listless SPTHT algorithm suitable for the hardware is presented. The algorithm uses the state mark instead of the list to record the information of the partition of sets, and then simplifies the process of the coding as the Boolean calculation. results show that this improved listless SPTHT algorithm is simple, thus saving the resource and increasing efficiency. The method is efficient for realizing the image compression with a high speed by the hardware. Chapter 6, a brief summation is set out.Summation and outlook:1,In this paper, image coding was deeply studied, a hardware architecture based on two-dimensional discrete wavelet (LS97) transformation wasproposed, and simulation was done .Analysis of the performance of the VLSI provided a basis subsequent image coding.2,The author analyzed the optimal Listless SPIHT, reduced the need for coding of memory and demand for hardware logic.3,Because limited time and ability ,optimized no list SPIHT Algorithm was not realitied, This is the direction for the future.