Study And Implementation Of Image Coding Based On Integer Wavelet Transform

Image compression based on classical DWT(Discrete Wavelet Transform) is one of the most popular research fields due to the fact that it has the virtues of high compression ratio, good quality image reconstruction, progressive image transmission, etc. On the other hand, it has the drawback that it is hard for hardware implementation due to computational complexity. It is just the hardware implementation to push the development of image compression technology, because it can be manufactured to meet the demands of people, and can bring great business chances. IWT(Integer Wavelet Transform) greatly reduces the number of multiplication and memory usage compared to classical DWT, so it is convenient for hardware implementation. However, it has one main defect that using the IWT instead of the DWT degrades the performances of the lossy codecs. To make the tradeoff between coding efficiency and fidelity of reconstruction image, a novel image compression algorithm is proposed and implemented by software and hardware. The main contributions are:(1) A new image quality metric is presented, which is good consistent with human visual perception. The simulation results show that the correlation coefficient between the proposed algorithm and subjective MOS(Mean Opinion Score) is 0.95, but the corresponding correlation coefficient obtained by the PSNR measure is 0.85.(2) One kind of 2-D IWT based on lifting scheme is put forward. Wavelet transform is done in image space directly. The multiplication number of the method is rough 25% less than the conventional implementation based on 1-D tensor product, so the memory usage is also reduced. Experiments indicate that the 2-D IWT is about 10% faster than the classical one.(3) Conventional arithmetic coding is modified. There are no float-point operation and no integer multiplication and division, while the coding performance is maintained. So the computational complexity is greatly reduced. The simulation results show that the algorithm is nearly 30% faster than CACM87.(4) A suit of hardware platform based on TMS320DSC25 is developed successfully. The platform has the properties as follows: low power consumption, compact size, low cost, easy to be upgraded, etc. The algorithm and software of image compression can be expediently developed on it.(5) A set of image compression software is designed. The algorithm is based on 2-Dlifting scheme IWT, combined with modified EZW based on morphological dilation and modified adaptive arithmetic coding, and several questions of assembly optimized by hand are discussed according to author's experiences. Experiments indicate that the coding time of the algorithm is approximately 1 times faster than that of EZW, and the reconstruction image quality outperforms EZW 0.2-0.9dB in PSNR, its performance compares favorably with SPIHT. It can compress 5 frames per second for the 512x512 gray images at 0.25bpp after the software of the algorithm is debugged and optimized on the hardware platform designed in this thesis.