Research On Methods For Image Rectification Based On Transform Invariant Low-rank Textures

In the field of image processing,the extraction and representation of texture features directly affect the quality of subsequent work and the quality of experimental results,so it is of great significance.In recent years,based on the research and development of sparse representation and low rank matrix,more and more attention has been paid to visual invariant low rank textures.The most notable method is image rectification based on transform invariant low-rank textures.It does not need to locate and extract local features such as traditional edges or corners,but instead model the distorted or damaged low rank textures,and use convex optimization algorithm to solve it effectively.The experimental results show that the correction method has excellent correction effect for most of the synthetic and natural images containing abundant low rank components,and has strong robustness for noise,occlusion,illumination changes and scale scaling.However,existing image rectification methods based on i transform invariant low-rank textures still have some limitations in algorithm performance and complex scenes that can be processed.Therefore,the optimization algorithm is very important for its better processing performance and a wider range of applications.Firstly,this paper introduces the concept of low-rank texture,and systematically expounds the theory,model and derivation process of existing invariant low-rank texture rectification methods.Secondly,in view of the images containing multiple adjacent low-rank texture regions,the Hough transformation line detection method is applied to locate the intersecting lines of adjacent textures,then the correct segmentation of multiple textures is achieved by optimal segmentation of different texture.Finally,this paper introduces the multi-resolution and branch and bound strategy in detail,and applies it to the optimization process of the new correction method,which makes the optimized algorithm has been significantly improved in the convergence range and the efficiency of iteration.Through in-depth research and optimization of existing rectification methods,this paper extends the original algorithm to multiple low rank region rectification and effectively improves the processing performance of the original algorithm.The experimental results show that the algorithm studied in this paper has a wide range of application prospects for complex surface which include multiple external facades,and the correction effect and robustness of it is significantly improved.