A New Color Image Segmentation Algorithm Based On Region Growing

In the applied mathematics and computer science, image processing and analysis has a solid theoretical foundation and extensive application .The primary research of this paper is color image segmentation based on region growing which is a focal point of research in image processing, including image feature selection, algorithm analysis and implementation,comparison and evaluation of segmentation results and other content.This paper proposes a new color image segmentation algorithm based on region growing, realizes the algorithm and analyses the experimental results.First ,this article describes color space in detail, studiess the main color space used in recent years, discussed conversion relationship between the different color space.In image feature extraction, the paper studies the image color feature, the textural feature and so on. Relative to other features, color feature is insensitive for the rotation, translation, scale changes, or even not sensitive to a variety of deformation, showing very strong robustness, and is the main feature of this algorithm used.Second, by extracting the image color, texture, shape and other feature, considering the image edge and internal information, this paper proposed a new seed point generation algorithm. The pixel is defined as a seed point, by calculating it neighbord similarity factor ,which is more than 0.5 points. Thus, according to certain rules and order of the growth ,the image is divided into different regions. At the same time, a new algorithm is proposed for region merging. For the isolated point and the regions which's pixel points is less than the threshold, calculating the feature similarity between each pixel of them and the pixel at outer boundary,the pixel of them is divided into the same class with the pixel at outer boundary ,which they have the bigest feature similarityThe regions which is greater than the threshold value come to be split into new classes. Specific image segmentation results show that the algorithm has good results.Then, this paper decoleps an evaluation algorithm based on artificial segmentation evaluation index ,compares with this algorithm and JSEG algorithm and analyses the algorithm complexity and parameter selection through specific experiment.Finally, the algorithm software is built based on C #,.You can directly get the best segmentation results in Window environment.