Image coding method based on total variation

Abstract

The invention relates to an image coding method based on total variation, comprising the following steps of: (1) decomposing an input image (I) into a structure component (u) and a texture component (v); (2) analyzing the (u) at a coding end, selecting an analyzed area, mapping the (u) to the (v), and coding non-synthetic areas in the (u) and the (v) by using JPEG (Joint Photographic Experts Group) traditional coding mode; analyzing areas which need to be analyzed in the (v) by using a block connection method at a decoding end, v'' represents an analyzed result of a texture area, and u' and v' respectively represent compressed results of the non-synthetic areas in the (u) and the (v); (3) superimposing the u', the v' and the v'' at the coding end to obtain an image I', and I'' represents a result of the input image I which is compressed by the traditional FPEG; and under the same bit rate, judging whether the I' or the I''is nearer to the I by a JND (Just Noticeable Difference) model, then, if I'' is nearer to the I, reducing a decomposable model parameter lambda, returning to the step (1) to continue a whole process until the I' is nearer to the I as compared with the I'', finishing circulation, and selecting the I'. The invention can increase the coding quality of the image at the coding end.

Description

technical field [0001] The invention relates to an image coding method, which is especially suitable for image coding with a large number of texture regions. Background technique [0002] Efficient image / video compression coding is one of the basic scientific problems to be solved urgently in multimedia applications with digital storage and transmission as the core. Image / video compression methods can be roughly divided into two categories: signal processing-based coding methods and vision-based coding methods, which are widely used at present. The former mainly uses digital signal processing technology to remove spatio-temporal redundancy and statistical redundancy in images through prediction, transformation and entropy coding. The latter uses the results of human vision research and uses computer vision analysis to remove visual redundancy. The similarity of the content of the texture region leads to a large amount of visual redundancy in the texture region, so the text...

AI Technical Summary

Problems solved by technology

[0004] Currently commonly used image compression coding standards, such as JPEG and JPEG2000, do not reflect human visual characteristics well, and the proposed texture analysis and synthesizer do not specifically deal with texture features.

Images