Study On Hybrid Intelligent Algorithms For Color Image Segmentation

Image segmentation is critical to image processing and pattern recognition, and it is also one of the classical dilemma of the image processing .It is an important component of image analysis and pattern recognition system, and which decided to pattern recognition and image quality of the discriminate analysis of the final results. Origin image was converted to a more abstract and compact style by the combination of image segmentation, object separation, feature abstraction and parameter determination, which make it possible more advanced analysis and understanding. As a higher level of image processing and image analysis and pattern recognition itself, Image segmentation has more important significance of the research.Image segmentation has been widely applied in practice prospects, such as industrial automation, online product testing, production control, document image processing, remote sensing and biomedical image analysis, security surveillance and military, sports, agriculture and other aspects of the project. Generally speaking, in various image applications, as long as the need for image extraction, measurement and tracking, we all rely image segmentation.The technology and process which is divided the image into each characteristic region and extracted interested goals is called image segmentation. Here characteristics can be gray, color, texture, it could be a single regional goal, or could be more regional. Images can categorize color images and gray images. Color image external objective world is the most urgent, because the eyes of adaptive brightness, is there any point in a complex image can identify dozens of gray level, but thousands of color can be identified. So in many cases, simply using gray background information can not be extracted from the goal, we must also help color information. Color images to gray-scale image are much greater than the message. For gray-scale image segmentation, the only information available was a bright. However, the bright vision of the general feeling only reach 20% level, brightness and color images not only provide information, but also include with hue and saturation. Hence, the use of gray can not be divided; we can try to split the use of color information. Now, following the rapid increase in the ability of computers, color image segmentation technology is increasingly concerned about and has a broad space for development.Color image segmentation has two main aspects: Firstly, choose a suitable color space model; Secondly, choose segmentation strategies and approaches which are suitable for this space and model. It can be expressed as following:Color image segmentation method = Color Space + gray color image segmentationThe aim in this thesis is to color images. Firstly, we introduce thoroughly of all color space and color vision. Secondly, have a systemic discussion on the current traditional color image segmentation techniques. Including the method based on histogram thresholding, method based on color clustering, method based on neighboring district, method based on vertical-square map, method based on the model and so on. Compare with their respective advantages and disadvantages, it is found that the traditional single arithmetic of color image segmentation inevitably has some deficiencies and defects. Image segmentation combined with artificial neural network or genetic algorithm is a new research field with rapid development in recent years. The methods have better effect, in that they have human-like intelligence and concurrency. Although many works have been done in this area, we still have a long way to get a satisfactory result. Based on this situation, we can combine different intelligent algorithms according to the actual situation or have a hierarchical division of Image segmentation. By means of these, it wants to seek more effective segmentation algorithms in accord with human visual and perceptive traits. This can make up for existing deficiencies and shortcomings. This color image segmentation technology will be a future trend.This thesis is based on the intelligent application. In particular, we use the most representative algorithms such as artificial neural networks, genetic algorithms, ant colony algorithms and fuzzy algorithm, and mix them together for color image segmentation. Therefore, this paper introduces three orientations in image segmentation, one of which is combined clustering segmentation with neural network, one of which is combined entropy-threshold segmentation with genetic algorithms, and the other of which is combined genetic algorithms with ant colony algorithms. This will enrich the color image segmentation algorithm and has a certain extent practical value of the image processing system enhancing.