Research On Color Image Segmentation Based On Intelligent Optimization Algorithm And Normalized Cut

Image segmentation problem is an important step of image analysis, segmentation quality affect the subsequent processing. In recent years, color image segmentation based on graph theory become a research hotspot, color image color is rich, contain more details, and with the development of science and technology, the channels to obtain more color images and cost reduction, color image segmentation is become more and more important.This article is based on intelligent optimization algorithms and normalized division criterion, for color image segmentation, main work of this paper is described as follows:(1)A fish swarm optimization algorithm was proposed combined with normalized divided guidelines to solve the color image segmentation. First, fuzzy c-means clustering is used to color image preprocessing, then employs fish swarm optimization algorithm to minimize Normalized Cut, finally through the optimal state of individual fish guidance image segmentation(2)A method based on bacterial foraging optimization and normalized cut algorithm was proposed. In order to solve the problem of premature convergence, introduced fuzzy C-means (FCM) dealing with color image for reducing the algorithm dimension; Meanwhile the bacterial foraging optimization algorithm was introduced to minimize Normalized Cut, so the stability and convergence speed of algorithm was improved; Then got segmentation result by the optimal individual bacterium.(3)Proposed a speed limited-disperse artificial bee colony optimization algorithm to solve the normalized cut color image segmentation problem in image field. According to the problem model, the position of the artificial bee colony algorithm was redefined discrete position, and increased the speed definition of individual bees. In order to solve the problem of premature convergence, introduced a speed limit process, and designed a speed limit function to increase the diversity of the population. Meanwhile the adaptive weighting adjustment strategy was introduced to update the position of individual bee. So the stability and convergence speed of algorithm was improved. Experiments show that the algorithm is superior to other similar algorithm in convergence rate and efficiency. So the algorithm in the normalized cut color image segmentation problem was verified to be efficiency and superiority in simulation experiments.(4)An image segmentation method based on shuffled frog-leaping spectral clustering was proposed. Classic spectral clustering algorithm can be mapped high-dimensional non-linear space to lower dimensional linear space, but often can not find the best value when deal with the optimal threshold by heuristic to search for the best value, This paper presented a shuffled frog-leaping spectral clustering method basing on the character. This method had the ability of depth of local search and global information exchange, and could get better clustering results. The FCM was being used to reduce the computing complexity when handled the problem of color image segmentation. Experiment showed that this method could full accuracy extract the target in the color image.