Color Image Segmentation Based On Disjunctive Normal Level Set

Image segmentation is a technique or process that a given image is divided into several assumed homogeneous and inter-independent sub-regions to get ROI or target objects according to a certain attribute.Compared with monochrome image,color image contains more effective information,for example,hue,saturation and brightness.Although there are up to hundred models for expressing color space,none of them is suitable for all color image processing.Therefore,it is a difficult to take optimal color space according to the real requirements in color image segmentation.In this paper,a color image segmentation algorithm based on disjunctive normal level set is proposed.The main work is as follows:Firstly,the basic theory,level set function and the relative traditional CV model of the level set method widely used has been studied based on the analysis of the current research progress in color image segmentation at home and abroad.By combining the k-means algorithm with the HSV color space,a new color image segmentation algorithm is produced after improving the level set evolution function of the traditional CV model in image segmentation.Through the segmentation experiments on the images in the standard Berkeley library,and the comparison with the traditional CV algorithm and GMRF-CV algorithm,the improved algorithm in this paper has significantly improved the segmentation accuracy with the great advantage over the GMRF-CV algorithm in the terms of time performance.Secondly,a novel disjunctive normal level set model with HSV color space is created and implemented on the research basis of existing DNLS model to enlarge the application range from the original simple composite image and gray image to the color image with complex background and bright color.Through the segmentation experiments comparison on the images in the standard Berkeley library,its segmentation accuracy is better,and its time performance has great advantages over the traditional CV algorithm and GMRF-CV algorithm.It is not necessary for the improved disjunctive normal level set algorithm in this paper to be re-initialized and add penalty terms.And the algorithm can rapidly converge with excellent time performance and better segmentation result.Moreover,it not only can be applied to segment flower image to supply necessary parameters for automatic production of facility flowers,but also can provide pre-processing support for image recognition and machine vision.