Research On Texture Image Segmentation Based On Phase Field Model

As an important aspect of digital image process and pattern recognition, texture segmentation has always been one of the hottest and most difficult study topics. Texture segmentation involves accurately partitioning an image into sections according to the textured regions or by recognizing the borders between different textures in the image. The segmentation of textured images aims to partition an image into several disjointed regions that are homogeneous with regards to some texture measures, so that subsequent higher level computer vision processing can be performed. It has long been one of the most important branches of digital image processing and has drawn considerable attention of researchers from around the world. Therefore, texture segmentation is, up to now, still an open topic with great challenge in image processing field.This dissertation is devoted to the segmentation of textured images. After comprehensively reviewing the basic principles and existed methods, the author chooses the variational theory based approaches to solve this problem, and proposes a novel texture segmentation method which has a good performance and avoids the complexity and difficulty in texture modeling and analysis.This dissertation discusses the problem of jacquard image segmentation by approaching phase-field paradigm from a numerical approximation perspective. A modified phase model based on the classical phase-transition theory in the field of material science is introduced to separate and extract geometrical pattern features in jacquard images. The role of the proposed model is to direct a dynamic separating process of phase transition for image pixels with different gray levels. The images to be segmented are partitioned into connected regions by grouping neighboring pixels of similar intensity levels, and the edges of the images are represented by dynamic interfaces between different phase values in the model. In particular, a novel algorithm for numerical solving of the model is developed based on finite element approaching and adaptive remeshing technology. First, the above model is discretized using P1-lagrangian finite elements in space. Then, a mesh refinement strategy is applied to generate a mesh structure that well represents a jacquard image by adapting to its pattern content. The analysis and experimental results on synthetic and jacquard texture images demonstrate the effectiveness of the algorithm.