Study On Image Coding Algorithm And Its Applications Based On Wavelet

The study of the image coding algorithms and its applications is one of the most active areas in the information technology. To image coding considerations, the selected transform method should being matched with the characteristics of the compressed image. Meanwhile, it must consider distortion ration, realizing complexity, coded bits rate and so on. The wavelet transform has advantages over other transform methods upon these aspects. The wavelet transform can satisfy these aspects by selecting proper basis function, setting reasonable threshold, applying fast algorithm and constructing embedded bits stream. But the theories of wavelet transform are not perfect and the algorithms of wavelet transform have many pitfalls in regard to its speed, complexity and compressing performance. The classical EZW algorithm has slow computing speed, poor reconstructed image quality, etc. The selection of de-noising threshold is not perfect. The performance of video compressing circuit based on wavelet cannot turn up trumps. So it is important and imperative that developing the study of wavelet image compression algorithm and its applications.This paper divides six chapters. The first chapter introduces overall knowledge about image coding, including two key stages of image coding technology developing, the conditions of image coding and most international standards of image coding. It also points out the advantages of wavelet methods over other methods and the importance of developing the study of wavelet image coding.Some key theories about the wavelet image compression are elaborated in chapter 2, especially on all kinds of wavelet transform, multi-resolution analysis and Mallat algorithm. This chapter also studies the criteria of selecting wavelet basis in engineering applications. The conclusions of this study are: the regularity of reconstructed wavelet function is most important to the quality of reconstructed image; the bi-orthogonal (9,7) wavelet basis has stable excellent performance in the lossy image compressing.Chapter 3 is about the study of embedded zerotree wavelet algorithm.First it illustrates the properties of original image wavelet decomposition coefficients and analyzes the standard EZW algorithm to point out some pitfalls. To solve those pitfalls, some studies have been done in regard to reasonable organization of wavelet coefficients, multi-threshold setting for quantization and scan times. Theoretical deducing gets the conclusion that the relation of quantization intervals between different decomposition scales should be the power of 2. This chapter brings forward RMEZW that is a new embedded zerotree wavelet algorithm. The characteristics of this algorithm include: (1) setting different quantization threshold in different direction according to HVS property; (2) adopting different quantization threshold in different decomposition scales according to MSE conception; (3) adopting T transform to reorganize high-frequency coefficients; (4) adopting DPCM to manage lowest band; (5) only one scan being done correspond to set threshold; (6) adopting multi-zerotree to produce more zerotree.The study of wavelet de-noising and compression for noised image is in the chapter 4. This chapter illustrates most different time-domain and frequency-domain de-nosing methods and point out that those methods cannot satisfy de-noising as well as compression. After the discussion of different wavelet de-noising methods and criteria, this chapter invents a new de-nosing threshold decision method that is independent in different sub-band and has visible formula. A new wavelet de-noising and compression scheme WT_DC based on minimum Bayes risk is also put forward. This scheme can create the extreme image compression as well as de-nosing. The estimation of noise level of noised image and the procedure of de-noising threshold estimation are put out in this chapter. For de-nosing image noised by coherent noise, a logarithm operation is done to convert the coherent noise to additive noise. The simulation results sh...