Based On The Mumford-shah Model Vector Image Segmentation Method

Image segmentation is the fundamental issue in the field of image processing and computer vision, and is a classical difficult problem in image processing. Vector-valued image (including color image and multi-channel image) provides more vision information than gray one, and it can describe the real world more vividly. As the rapid development of multi-media technology, vector-valued image segmentation has a larger academic and practical significance, and is becoming increasingly concerned in image processing research. In this paper, we extend the Mumford-Shah model to the vector-valued case. The obtained method based on partial differential equations (PDE) can be used to segment and denoise vector-valued images.Firstly, our paper introduce some image segmentation methods in common use and some methods for vector-valued image segmentation . Then we summarize the merits and drawbacks of these methods.Secondly, we depict the basic theory of level set method and its application in image segmentation field. Then we introduce a few image segmentation models based on PDE and mainly focus on the analysis of Mumford-Shah model for gray image and its simplified model (C-V model) .Thirdly, we propose a method for vector-valued image segmentation by extending the Mumford-Shah model to the vector-valued case. We apply the method in our paper to the segmentation of both color and texture images. The experimental results show that our method has a good performance for vector-valued image segmentation. It can be used more widely than C-V model. Then multiphase level set method based on Mumford-Shah model is implemented and the programs are successfully used in complex image segmentation.Finally, the future research direction is put forward.