The Study Of Multiple Description Lattice Quantization Image Coding Based On Coefficients Direction Scanning In Multi-scale Geometric Transform

In recent years,with promoting of Internet media terminal such as mobile phone,ipad,mini computer.People often pay money for low speed,high distortion in the process of information transmission to browse the information in unstable network environment.Multiple description coding guarantee the real time and robustness of information transmission,reduce data packet loss,bit error rate and as much as possible to reduce the image distortion in unreliable and heterogeneous network.Therefore,multiple description coding is becoming a hot spot of research in recent years.Lattice vector quantization coding is an extension of the multiple description coding,which the optimization of the performance also has been widely concerned.Classic Lattice quantization multiple description coding adopts wavelet transform method to extract the information of direction characteristic in subband.Wavelet Transform has a lack of directional and not well capture two-dimensional image of the line or plane singularity,which is not best to retain the geometric information such as the edge texture.The geometric information such as the edge texture is basic features of images,the information acquisition is particularly important in the process of image coding.Multi-scale family of the Non-Subsampled Transform,Contourlet Transform and Contourlet mix Wavelet reflect the anisotropy of curve better than Wavelet Transform,at the same time,as the transformation scale and fine,which is more effectively capture the details and texture information of the image.Therefore,the Non-Subsampled Shearlet,Contourlet and Contourlet mix wavelet coefficient direction scanning in Lattice quantization multiple description image coding is proposed,which is to avoid image lost texture,edge and contour geometry information in transmission.This article elaborated the image coding technology research status domestic and abroad firstly.Secondly,this paper summarizes the Multiple description coding knowledge,and mainly introduces the Multiple description coding scalar quantization,vector quantization of multiple description coding method and the multiple description coding method of the vector quantization.On the basis of theoretical analysis of the Multiple description type lattice vector quantization,This paper puts forward three algorithms based on the direction coefficient scanning lattice quantization and Multiple description image coding:(1)The algorithm which is based on the sampling under the Shearlet(None Subsampled Shearlet,NSST)direction scanning type quantization of Multiple description coding.Use the image after the next sampling Shearlet transform which has the translation invariant feature extraction subband,and reorganize the coefficient.Under the lattice vector quantizer form two description,ensure the description is lost,the other a description can restore image contour edge and texture information,improve the quality of single channel decoding.(2)The algorithm which is based on multiple description coding Contourlet direction scanning lattice quantization,According to the image after Contourlet transform subband coefficients have more directional,better capture texture,edge and contour geometry information,"optimal" to express two-dimensional image singular line or plane.(3)The algorithm which is based on wavelet and Contourlet joint in the direction of the scanning lattice quantization multiple description coding.Make full use of Contourlet transform the direction of the opposite sex with the advantages of stable structure after wavelet decomposition,not only can reduce the Contourlet transform redundant data,but also can very good capture a bit bizarre.Make the use of the members of the family of multi-scale sampling of shear wave Shearlet transform,Contourlet transform and Contourlet transform and wavelet transform of the joint scanning strategy which is based on the method of lattice vector quantization,the application of multiple description coding,the Three kinds of scheme can effectively improve the single channel of the peak signal-to-noise ratio,and retain the original image texture and edge geometry information,the experimental results show the effectiveness of the above three kinds of solutions.