Image Segmentation Based On Improved GAC Model

Image segmentation is one important aspect of the image processing. Compared with the traditional segmentation method, the method, which combines the theory of partial differential equations and the theory of level set, obtains some satisfying results and becomes the focus of the image segmentation methods research.On the basis of the traditional level set method in the image processing, we design a new signed distance function method with the help of the extension of concentric circles. The new method avoids the massive computation in the construction of the distance function .In this dissertation, we studied the inadequacy of the traditional GAC models using the level set method, and proposed a improved GAC model. From the basis of geometry curve evolution, we find an embedding function on closed geometric curves, get a functional of embedding function, then calculated the gradient decent flow of the functional by using variation method, thus the evolution equation of embedding function is obtained. It is also one kind of parabolic type diffusion equation. The method avoids the velocity extension and the re-initialization of the level set function. Moreover, there is no spontaneous singularity in the evolution, so we need not apply upwind schemes to do the numerical computation. This method is better than traditional methods in computational complexity and stability, so it serves as a powerful tool of the application of GAC models in image segmentation.