Study And Application Of Image Segmentation Based On Ambrosio-Tortorelli Model

Image segmentation is to divide the image with the same or similar characteristics of the sub region, and the different sub regions will show significant differences. After so many years of research and development, many image segmentation methods have become more and more mature, that various theories and methods have been applied in real life and the development of industrial design, which obtains good results. Among them, segmentation methods based on the variational energy model of image segmentation in recent years has become one of the more popular segmentation technology. Mumford-Shah model is one of the famous variational segmentation model. The image segmentation problem can be transformed into a functional extremum problem by this model. Mumford-Shah model is also a multi-scale segmentation model, which has a good noise immunity. However, the solution of the model is very difficult. In this solution, there are two well-known methods: one is that the original model is simplified and combined with the level set method, such as Chan-Vese model; another is using the model of the Ambrosio-tortorelli to approximate the original model weak functional form, and then by using the variable method of Euler equation and the gradient descent method,we can get the final iterative evolution equation.We did further studys on the Ambrosio-tortorelli model, you can get multi-scale image segmentation by adjusting the parameters. In addition to the boundary image, you can also get a smooth image. But the model also has its own shortcomings. This paper presents a new algorithm and model for this model. At the same time, the model has a good noise immunity, which is good for the segmentation of noise image. The main work of this paper is as follows:(1) Due to the phenomenon of the rounding effect of Ambrosio-tortorelli model,a method of weak iteration of image corner position was designed, and the moel of an elliptical approximation function has been improved to achieve the best improvement effect. If the parameter is too large, the model will generate rounding effect, in addition, it will also bring the phenomenon that the boundary of the final segmentation image is not obvious. The method of extract the effective image segmentation pixels, and of the threshold segmentation was used to enhance the edge of segmented image. The experimental results show that this model has a certain effect on the improvement of model.(2) The model was applied to the noise image segmentation, where the parameters have a great impact on the final segmentation. In order to improve efficiency and to avoid manual adjustment of parameters, parameter setting method proposed in this paper, which related to image information. Meanwhile, in order to achieve Ambrosio-tortorelli model automatic segmentation capabilities, in addition to setting the parameters, this paper used a method of feature extraction to get the piexl number of the noise and the image boundaries. When the noise and boundary number tends to be stable,it is the time to end the segmentation. In order to facilitate the experiment, a fuzzy satisfaction measurement model was proposed. Finally, the experimental results show that the proposed method is effective for achieving the automatic segmentaion of the model on multiplicative noise and additive noise image, which can achieve good segmentation results.