Natural Image Segmentation Method Based On The Theory Of Gestalt

Natural images have a wide range of application in the field of traffic control and military security and social networking by the development and popularity of the imaging device and the Internet technology. Natural image segmentation become difficult and hot in segmentation fields because of the uncertain light and the complex of background and texture and other characteristics of natural images. In this paper,natural images are taken as its subject and Gestalt theory of human visual perception are taken as the basic theory, do researches further and analysis on the natural image segmentation based on Gestalt rules, first the natural image segmentation of research status were summarized and the existing problems were analyzed; then the concept of Gestalt theory were introduced; then the scheme and algorithm of natural image segmentation based on Gestalt rules is proposed and experimental using Matlab software, experimental phenomena was showing; and finally the work was summarized.This paper the main works and achievements are as follows:1、Natural image segmentation methods was commonly used by based on the edge and the area and the particular combination of segmentation theory, and analyzes the problems of these methods, such as limitations and edge intermittent and poor scalability.2、The origin and development of Gestalt theory is simply explained, and the core theory of Gestalt rules are illustrated in detail, and the theory used in computer vision are highlighted.3、In this paper, a new closed contour extraction method based on Gestalt rules is proposed to reduce the impact on the background of the edge. First, the integration significant constraint Canny edge detection method, which integrates significant constraints in the framework of Canny edge detection, is given to reduce the background edge in edge extraction process. Then, edge information into the edge fitting algorithms about the fold approximation is proposed to reduce the effects of noise to the the edge by Canny edge detection method. Then, the edge of the fitting is measured by Closed Rules in Gestalt rules, that can obtain the relationship between the edges. Finally, the new objective function is proposed by analyzing the relationship between the edges and the significance of the target area, the objective function issolved by the minimum weight perfect matching algorithms. Experimental results show that our algorithm improves the effectiveness and accuracy of the contour extraction.4、In this paper, a new natural image segmentation method based on Gestalt rules is proposed to to reduce the impact of illumination and other factors and improve the accuracy of segmentation. The novelty lies in three aspects: firstly, the original image is segmented into several sub-region by using Ncut algorithm. Secondly, the Gestalt rules are introduced to measure the regions, and a quantitative calculation model based on the regions, which can get Ratio-Region, is proposed. Thirdly, a new merge algorithm on the basis of Ratio-Region is proposed. The final segmentation result is obtained by merging the regions with the merge algorithm which is simple and high efficiency. Both quantitative and visual inspections operated in 30 images show that Gestalt rules will be good applied to image segmentation, and compared to the comparative experiments, the effect of the algorithm is more in line with human visual perception.