Snow Model Based Image Contour Extraction Method

As a very important process, contour extraction directly affects the results in the intelligent vision system, especially pattern recognition. As a classical image segmentation method, watershed has been widely used, but existing serious over-segmentation problem. The level set method, which is based on the edge, can naturally change the topology of contour curve, so it is widely used in image contour extraction,however, large number of iterative calculation causes great time consuming.This paper studied these focus problems above, and main research work as follows:1) Adaptive surface evolution snow modelBased on the new image segmentation idea of simulating snowfall, this paper focuses on exploring image contour extraction method, which fits on the people visual characteristics and has the image contour curved evolution guided by regional characteristics. According to the image regional characteristics information, data adaptive Gaussian kernel can realize adaptive image surface evolution, in view of this point, in this article data adaptive Gaussian kernel will be introduced into the snow model, simple snow model, i.e. adaptive evolutionary snow model surface, is proposed. And similar to the process of simulating snowfall, the ground surface evolution based on that new model, implies a simple surface contour evolution, and has the topological adaptability. These discussions and researches will provide some new ideas for image contour extraction.2) Watershed image contour extraction based on adaptive curved surface evolution of snow modelCombined with the adaptive surface of snowfall model, watershed image contour extraction method based on snow model adaptive surface evolution is proposed in this paper. Firstly, image curved surface evolution is realized under adaptive curved surface evolution mechanism of snow model, to smoothing the large scale details information in the images, meanwhile suppressing noise effect. Secondly, the contour information of original image is enhanced through the fusion with the gradient and detected edge information. Finally, the bottom filling is introduced to control the watershed segmenting region, and realizing contour extraction,which can fit better with the people visual property requirements.3) Level set image contour extraction method under adaptive curved surface evolution mechanism of snow model Combined with adaptive curved surface evolution of snow model, a level set image contour extraction method under adaptive curved surface evolution mechanism of snow model is proposed in this paper. Based on adaptive curved surface evolution mechanism of snow model, the important feature information underlying the image in the can be shown. In traditional level set method, Gaussian filter is used to calculate edge operator, in order to improve the method, in the new method Gaussian filter is instead with the data adaptive curved surface evolution kernel function. The original image is processed by adaptive curved surface evolution, and then the evaluated characteristics image is obtained, based on the characteristics image level set method can be realized, so as to realize the contour extraction.Combined with the snow model, aiming at the over-segmentation, the improved watershed contour extraction method is proposed. Under the adaptive evolution mechanism of the snow model, a new level set contour extraction method is discussed, and the improvement and innovation with classical method is realized. The experimental and theoretical analysis proves the superiority of the new method, all of these new method and new ideas will further perfect the contour extraction theory based on the watershed and the level set.