Star Shape Constraint For Automatic Object Segmentation

Image segmentation has always been a hot topic in computer vision and image processing,and also,it is one of the most challenging research directions.It is the basis of image processing techniques.The results of segmentation will directly affect the relevant tasks of image processing and analysis.In recent years,the variation model of image segmentation algorithm has been the concern of many scholars and researchers,and also been proved that it had good results in practice.The multiple piecewise constant(MPC)model was rather a typical segmentation algorithm,which mainly introduced the shape information of target into the segmentation process and improved the segmentation performance.However,due to the complexity and diversity of targets,the acquisition of shape prior information will become a main problem in the process of segmentation.Although the prior information can be obtained by artificial interference which is time-consuming,it undoubtedly increases the difficulty of segmentation.Therefore,the automatic image segmentation of MPC model always arouses more attentions.This paper obtained prior information based on the star convexity constraint,and carried out the improvement research on the MPC model,and achieved the automatic image segmentation algorithm.The research has obtained corresponding results.Firstly,this paper proposes the multiple piecewise constant(MPC)model with star convexity constraint.This model uses multiple star convexity constraints to conduct prior constraint on the target shape based on the MPC variation model,which enhances the anti-jamming performance on the object surrounding environment compared with the original variation model,and improves the performance and robustness of segmentation algorithm.Secondly,this paper constructs the labels model of automatic acquisition of target and background label information.In order to satisfy the required label information,we analyze the performance of different saliency detection models,and give out the saliency detection method that can effectively label target and background information,and then gets the result of image segmentation.The method sets the foundation of automatic acquisition on star convexity centers.At last,this paper proposes the star shape constraint for automatic object segmentation.The algorithm automatically gets the star convexity centers by dilation and erosion operation based on saliency detection,and then the multiple piecewise constant model based on star convexity constraint is used to achieve the automatic effective segmentation.Our proposed automatic segmentation method is evaluated in different image databases,and large numbers of data indicate that the method of this paper improves the results of image segmentation and has a good universality,especially for the target details or multiple targets.