Improved ROI Coding Method Based On SPIHT

ROI(Region Of Interest) coding can ensure that the ROI is encoded and transmitted firstly,and the quality of reconstructed image of ROI is better than BG (Back Ground). When ROIcoding is combined with SPIHT algorithm, arbitrary shape of ROI image can be encoded withthis method and without embedding the ROI mask. At the decoder side, ROI area can bereceived and decoded in the first place, and even without receiving complete image code, ROIimages can be reconstructed well.In the analysis of the ROI coding methods, some defaults can be found. The fusion betweenROI image and BG image isn't good enough. It's difficult to adjust the relative importance ofROI and BG image quality. Or the coding system is very complicate. An improved method ofROI coding is presented in this thesis with the combination with SPIHT. In this new ROI codingsystem, pre-process is improved in order that ROI and BG image can be fused naturally. Thewavelet transform strategy is different of ROI image from BG image. CDF 5/3 lifting wavelettransform is used in ROI area, CDF 9/7 in BG area. CDF 5/3 is lossless and fast transformmethod, CDF 9/7 is loss and compression method. The two-dimensional wavelet transform isreplaced by lifting wavelet transform such as CDF 5/3 and CDF 9/7. Two-dimensional wavelettransform is very complex because of its convolution computation, while lifting wavelettransform is very simple because of its only four fundamental operations of arithmetic tocomplete the transform. The replacement not only reduces the computation, but also simplifiesthe system structure. Any shape of ROI image can be coded without embedded the shapeinformation of ROI due to SPIHT embedded coding algorithm in the improved system.The improved method has been proved feasible and useful through theoretical analysis andsimulation. The fusion of ROI and BG is better ;image quality is higher; the coding system issimpler.