Research On Video Coding Techniques Based On Multi-wavelet Transformation

It was just because multi-scaling function and multi-wavelet function provide more flexibility and scattered than single wavelet in quantified coefficient of high ratio video compression, they are widely used in the 2-D image compression, and play it to the 3-D video image compression. Nowadays,3-D video image compression based multi-wavelet is used primarily in the field of low bit ratio compression. But the time-domain of video image is dynamical, and it should be distinguished from the static image.And, despite the 3-D transforming technology has been reduced the influence from scene change on video compression, but the effects are still existence, and the coding efficiency has the superior rise space.The purpose of this paper is put forward more scattered representing than 3-D multi-wavelet video compression, improve coding efficiency, and resolve the influence of scene change to video compression.This thesis consists of three parts.The first part, practical experiments are made with sparse expression of video frame, the result of the experiment manifests:multi-wavelet has sparser expression than wavelet to each video frequency frame, but wavelet and DCT have sparser expression compared to the multi-wavelet to the temporal dimension of video.The second part proposed the combination plan based on multi-wavelet video image coding according to sparse performance in the first part. It combines fine properties of multi-wavelet and wavelet or DCT. Experiments proved that it has great progress in both PSNR and visual quality. To solve disadvantage influence of scene change to 3-D video coding, in the third part is to study the scene change, presents a detecting method of scene change based on luminance component. GOP is redistributed through the testing result. Analysis and simulation show that the method has improved the quality of video. Accordingly, according to assaying the scene change and the quality of video coding, added a further refinement to the GOP packet which further reduce the influence of scene change. The experiment results show that the improved method has better coding efficiency, higher PSNR, and milder change to inter PSNR than the original.