The Research On Feature Detection And Restoration Algorithm For Transform Invariant Low-rank Textures

Image feature extraction is currently a research focus in the area of image processing and computer vision,studies in recent decades have been a great development.Due to the application of the actual images are obtained by affine transformations and projection transformations,scale transform,perspective transform,lighting,and other transformations,using invariant feature of image for image content analysis and compared to the direct use of the pixels of the image to show a better performance,especially local invariant features of images became the focus of research and application.This dissertation will be based on texture invariance analysis to recover and extract the low rank textures from image regions,although research on texture invariance has a greater breakthroughs in recent years,in content-based image retrieval and object recognition,medical image analysis,biometrics and other fields also has an extensive research and application.But in the practical application,their accuracy and efficiency still cannot meet the demand.In this thesis,we will provide a way based on transform invariant low-rank textures to for image recovery and features extraction.First,we introduce sparse representation and matrix recovery of image information describing and processing applications.Then,depending on the problems faced by this dissertation,mathematical modelling of real images,and then,by the iterative convex optimization algorithm for obtaining optimal solution.This dissertation introduces the algorithm's solving process and the optimization scheme in the implementation of the algorithm.In this thesis,we show how to efficiently and effectively extract a class of"low-rank textures" in a 3D scene from 2D images despite significant corruptions and warping.The low-rank textures capture geometrically meaningful structures in an image,which encompass conventional local features such as edges and corners as well as all kinds of regular,symmetric patterns ubiquitous in urban environments and man-made objects.Our approach to finding these low-rank textures leverages the recent breakthroughs in convex optimization that enable robust recovery of a high-dimensional low-rank matrix despite gross sparse errors.In the case of planar regions with significant affine or projective deformation,our method can accurately recover both the intrinsic low-rank texture and the precise domain transformation,and hence the 3D geometry and appearance of the planar regions.Extensive experimental results demonstrate that this new technique works effectively for many regular and near-regular patterns or objects that are approximately low-rank,such as symmetrical patterns?building facades?printed texts?and human faces.