Research On Image Segmentation Based On Shape Prior

The segmentation of digital images is an important part of image processing.It is an important method for feature extraction,image measurement,image registration,and image display.It is also a prerequisite for subsequent recognition and analysis of image processing.The objective segmentation method based on shape priors has very good advantages when dealing with issues such as image blur,occlusion,and missing.Based on the in-depth discussion of segmentation methods based on shape priors,this dissertation focuses on how to accurately introduce prior shape information,mine local information in a priori dictionary,and efficiently learn dictionary information while reducing dictionary dimensions.Adapt the target image to achieve more accurate and efficient segmentation.The main content of this paper is divided into the following sections:Starting from the level set and shape representation,the effect and method of shape priors guiding image segmentation based on level set were studied.The method of integrating the prior shape information into the energy functional was discussed.Aiming at the pose matching problem of the shape prior,a prior shape matching method based on the internal alignment method was proposed.For the problem of segmentation guided by multiple samples,a classification-based prior shape target segmentation model was introduced to eliminate the reconstruction errors that may result from nonidentical priors.For the problem of weak edges and blurred regions of medical images,an image model based on local shape prior is proposed.Based on the sparse representation of existing shape dictionary,a local decomposition of dictionary shape is introduced to generate supplementary dictionary.A local sparse shape prior constraint implements a local description of the target shape.By replacing the traditional global sparse shape representation method with a partial decomposition of the similarity shape of the dictionary,the target which is only partially similar to the shape dictionary can be well segmented,and the application scope of the shapes in the dictionary is expanded.Aiming at the problem of large prior information and high dimensionality in the extraction of priori shape dictionary information,a priori information extraction method based on PCA is introduced.The principal component analysis method was used to learn the dictionary and most of the redundant information was excluded.The distance measurement of priori shape and the target image are proceed in the PCA space,and less critical prior information is used for shape reconstruction,a small-scale deformation adaptability is achieved,the accuracy of the shape representation is improved.The local prior model is mapped to the PCA space to guide the segmentation of the kidney.