The Research On Threshold Image Segmentation Based On Image Of Weak Edges

Image segmentation is the field of image processing and computer vision in low-level visual areas of the most basic and important one, it is pattern recognition and target detection premise, has important practical value. However, when weak edges in the image of the target case, the target and background is not much difference between the gray scale, increasing the difficulty of extracting the target in a image. The task of the follow-up treatment is very difficult in the practice. The main goal of this paper, the boundary fuzzy image segmentation algorithm, the algorithm through analysis of past deficiencies, propose a new algorithm to improve the quality of image segmentation. The main contents are as follows:1. This paper describes the research background and significance of the domestic and Image Segmentation Research. Threshold method is introduced by the basic principles of the advantages that a threshold segmentation algorithm in dealing with the limitations of the weak boundary image.2. Two-dimensional minimum error threshold method in dealing with weak boundary image with noise, especially with salt and pepper noise ,the use of oblique two-dimensional histogram method is unreasonable. Although two-dimensional histogram method consider the oblique projection in the plane rectangular region of all four. But in deal with noisy images, the noise points have taken into account threshold selection strategy, reducing the quality of the image segmentation. This paper introduces a new two-dimensional histogram construction methods to reduce the noise point of impact threshold selection strategy, effectively reducing the noise in the segmented image point, improving the quality of image segmentation.3. The basic idea of graph cut spectral theory is divided into an image as an undirected weighted graph, each node represents the pixel image or a region, the weights between nodes that the close links between the nodes level, then according to certain criteria determined by the energy function, from the energy function to determine the best division of the image. When some of the characteristics of the image with weak edges, the existing map based on the normalized threshold power division method in calculating the value of considering only the difference between nodes and the spatial location of gray, it is difficult to get a proper solution. Thus the weak boundary in the segmentation image, the image is not well preserved details. Proposed algorithm pay enough attention to the relationship between pixels when calculate weight by introducing a new constraint which is made by Gaussian Mixture Model to the algorithm. Experiments show that this segmentation is better to retain the image details.4. Graph cut measure calculated by the existing map of the gray image needs to grayscale range traverse, increase the calculation of graph cut measure.Proposed algorithm computes the distribution of threshold range adaptively by the median parameter of Gaussian Mixture Model, therefore proposed algorithm makes the computation of normalized graph cut measure much more efficient.