Research On Wavelet Transform-based Head-shoulder Sequence Coding Techniques

With the age of information coming, the research of image coding method has become one ofthe most active fields in information technology by way of computer communication,information store, Internet communication technology, etc. Especially in 21st century, with thedevelopment of electronic and communicate technology, it is possible to realize the Videotelephone, meeting TV, signal TV, information high way etc. On this occasion, it inevitablybecomes one of the main tasks to seek after the effective image coding method, and it willbecome one hot point of research at home and abroad. Thet analysis of wavelet is developing in the 1980's to deal with unbalanced signals. On thebasis of multi-resolution, it overcome the default of Fourier analysis. The study of wavelet basedimage and video compression has drawn much attention in this field. The image and videocoding based on wavelet transform is studied in this thesis, made improvements to its classicalgorithms, and conducted experiments certification.Firstly, Introducing the theory of wavelet transform, the concept of multi-resolution analysis,and the Mallat algorithm. The signal of Nossin is composed and decomposed by Mallatalgorithm. The advance of wavelet transform is further analyzed. Wavelet transform can changethe distribution of image energy and make the energy concentrating. So it can avoid the diamondeffect and utilize the feature of progressive transmission. EZW algorithms for the followingaspects of the proposed improvement : low frequency sub-belt failed to take full advantage ofcontains images of the original energy;it can not fully utilize wavelet coefficient characteristics;coding time long, resulting in coding efficiency decline, Dispart Frequency Embedded ZerotreeWavelet was proposed (DFEZW). Recommend Multi-Resolution motion Estimation andCompensate(MRME) in wavelet field, it have many disadvantage including the sport vectordiscontinous, true body in motion border with block border inconsistent, make thehigh-frequency weight of transform coefficient increase, reduce the code efficiency of DFD, sothis text propose Multi-Resolution motion Estimation and Compensate based on differenceassign neighbor lump predict (FFMRME ). Finally, according to the image coefficient of waveletdistribution characteristic, and the different sensitivity of human visual system on edge, smoothregion and texture of the image, wavelet images will be presented that high frequency partitioninto edge, smooth , texture zone fractal coding, through experiments indicate that the method canbe further improved image quality restoration.