Continuous Potts Model Based SAR Image Segmentation by Using Dictionary-Based Mixture Model
In this paper, Potts model based on the dictionary-based mixture model (DMM) is proposed to make image classification. Potts model is used for SAR image segmentation by minimizing energy functional, which is a weighted sum of data fidelity and the length of the boundaries of the regions. However, it needs prior information such as the number of regions and the probability density function of image. In this paper, we overcome this problem by using the dictionary-based mixture model, which can compute the optimal number of segments automatically and the probability density function of complex SAR image. Experiments on several real SAR images show that Potts model based on DMM has better performance in SAR image segmentation than that with sole distribution.
This work is supported by the National Natural Science Foundation of China under Projects 61271287.