Research On Image Segmentation Methods Based On Superpixels

Image segmentation is the most basic research topic in the field of computer vision.And it is an important part of image analysis and processing.It means that the image is divided into several disjoint sub regions(superpixels)based on certain features,such as color,texture,transparency and similarity criteria.The purpose of image segmentation is to simplify or change the representation of an image,so that subsequent image analysis and processing can be carried out.After decades of research of image segmentation,a large number of segmentation algorithms have been proposed,yet there is not a general image segmentation algorithm.Because different image segmentation algorithms are applied to specific application scenarios.At present,image segmentation technology has been widely used.It plays an important role in target tracking,human posture recognition,biomedical image and machine vision.Superpixel segmentation is a hot topic in image segmentation.Superpixels can reduce the redundant local image information effectively,significantly reduce the complexity and computation of image processing,and retains the effective information for further processing of the image.At present,superpixels have been widely used in many fields of image processing.In this paper,a superpixel segmentation algorithm based on clustering is studied,and the main work and research results are as follows:Firstly,we introduced the typical superpixel segmentation algorithms based on graph theory and gradient information.And for the normalized cut method,FH method,entropy rate method,Turbopixels method and SLIC method,we tested and compared them with the experimental results of these algorithms.In addition,the subjective evaluation and objective evaluation criteria of superpixel segmentation are introduced,which is the criterion for evaluating the above algorithms,and also provides an objective basis for the improvement of the algorithm we proposed.Secondly,according to the requirements of the superpixel algorithms in segmentation accuracy guarantee under the premise of operational requirements faster,with fast searching and find of density peaks methods,we proposed a superpixel segmentation algorithm based on gradient information.The experimental results show that the algorithm can guarantee the accuracy of the segmentation of the premise,while can quickly generate superpixel blocks,and provides a good foundation for the subsequent image segmentation.The algorithm has achieved the ideal segmentation effect.Finally,aiming at the slow convergence of the algorithm has no segmentation with image pixel segmentation,character segmentation boundary is not smooth,the superpixel segmentation method based on general image segmentation method,segmentation and extraction of the target image,in the accurate segmentation effect at the same time,it can reduce the computation time.