The Study Of Scalable Video Coding Based On Network Transmission

With the rapid development of internet and global popularization of mobilecommunication, the transmission of image and video via network has become anew focus. Consequently the algorithm of video compression has attracted a lotof attention. Because of the large amount of data and limited network bandwidth,the development of multimedia communication is very difficult. How tocompress the digital video image effectively and make it scalable according to thestatus of the network automatically has become a problem which must be solved.With the development of the research of wavelet, many outstanding algorithmssuch as EZW, SPIHT etc have been brought forward. They have made waveletbased video and image compression be the most potential one.This paper is a summarization of research, which is concerning with themultimedia compression applied to network. It analyzes systemically thestatistical characteristics of the coefficient of DWT, the technique of fractalcoding as well as the region coding and explores more suitable coding scheme formultimedia network. The primary innovations include:1. Proposes the SPITH algorithm based on region of interest::We analyze the statistical properties of wavelet transform coefficients, anddo research into some typical zero-tree wavelet coding such as EZW and SPIHT.Offer a SPIHT algorithm based on region of interest, which split the region ofinterest using quarter tree, fill the coefficients based on line to suit the averageSPIHT coding, and then code the different objects using SPIHT, and allocate bitrate in the process of transmission according to the different bit budgets of regionof interest and background region.It proves in experiment that this algorithm can not only get highercompression ratio, but also get better reconstructed quality. The SPIHT algorithmbased on region of interest, not only has good compression characteristics, butalso can get completely embedded bitstream. Meanwhile, since different objectsuse different bit allocation, the acquired coding bitstream is more suitable fortransmission on network with complex environment. If this method was used inthe field of video coding, we would get higher compression ratio in theapplication of scalable coding based on object.2. Offers a new remote sensing image coding method with the combination of 3-D wavelet transform and vector quantificationConsider the characteristics of remote sensing image such as large size, fewwave band ,small relativity and complex veins and so on, we use 3D transformcoding for multi-spectrum image: use wavelet transform in spatial, while use KLtransform in spectrum; converge the coefficients in the relative same position intoa vector, make full use of the statistical properties of wavelet transform codingand the good adaptability of vector quantification to do efficient and frequentcoding for multi-spectrum image. Combining the vector construction andtree-shape vector quantification together, we can make full use of thecorresponding relation of high frequency hefts after wavelet transform: not onlyuse the linear relativity of image, but also use vector to realize non-linear reliantcoding. So we can get simple but efficient remote sensing image compression. Itshows in experiment that the reconstructed image has good video effect, canresist mistakes well, and the algorithm is simple to realize with hardware. It willhave wide application in the compression field of remote sensing image in thefuture.Then we analyze the properties of bitsteam got by wavelet transform, andoffer an algorithm which can extract region of interest from bitstream directlythrough computing ROI mask of each approach level; This can not only save a lotof recoding time, but also improve coding efficiency; It can suit many imagetransmission application with appointed region, especially for the application ofremote transmission of large image ROI.3. Does video coding of wavelet transform using motion information of objects After explaining object coding and frequent transmission of MPEG-4, weoffer a new 3D wavelet transform video coding method based on object: usingwavelet analysis aim at objects with any shape for input video sequence can tracemotion information of object efficiently, consequently do 3D wavelet codingbased on object. The method of video coding based on object we discussed , isthrough 3D temporal and spatial analysis to code the objects. It is very natural toconnect objects in temporal domain. For time analysis and compound motioncompensation filter, we solve the problem of continually tracing the motioninformation of objects efficiently through SA-DWT(shape arbitrary-discretewavelet transform), for we can get more reliable information of filter executingtime. Use motion compensation to insure most energy of signal to concentrate onthe lowest temporal and spatial subband. In addition, this method allows scalableframe rate, combines scalability and flexibility of objects together, and allowsselected objects of the same coding bitstream.to be decoded at different framerates It proves in experiment that it can not only get high compression ratio, butalso sustain scalable transmission based on object coding.4. Analyzes the properties of fractal techniques in wavelet domain and offers wavelet fractal compound video coding with zero tree We do research into the base of fractal theory, normal IFS model, the levelanalysis of IFS and IFS description of wavelet transform. After analyzing andsummarizing the advantages and disadvantages of current wavelet fractalcompound coding, we propose that fractal coding based on wavelet domainshould consider not only the comparability of wavelet coefficients of subbands,but also statistical properties of zero tree. We do research into new method ofwavelet fractal video coding using motion compensation combined with zero tree,and try to find the best combination of wavelet transform and fractal video imagecoding. We use entropy coding with no distortion for the lowest frequency...