Research On Image Segmentation Method Based On Graph Cut

Image is a vivid description of objective objects,which is the most commonly information carrier in activities.Image segmentation is the process of segmenting an image into a number of meaningful areas according to certain features(such as color,texture,etc.)and extracting the interested objects.In recent years,the graph cut method has become a research hotspot in the field of image segmentation,compared with the traditional image segmentation method,graph cut method has significant advantages,First,the graph cut technique is based on graph theory,and graph theory is a mature discipline,with a better mathematical foundation;Secondly,graph cut formulated the image segmentation problem as an energy minimization problem,and energy minimization can be applied to soft constraints,to avoid there only hard constraints in image segmentation.Therefore,it is necessary to the research.At the same time,the graph cut method is still in the developing stage,and its research in image segmentation needs to be further deepened and improved.For example,the longer processing time is needed for the 2-D image with higher resolution;the error rate is high during the segmentation of noise polluted and partially shielded image.Therefore,improving the speed and quality of image segmentation is of great significance in the process of image processing.In this thesis,we focus on the problem which is caused by the segmentation of noise polluted and partially shielded image according to graph cut technology.Study on the new method of image segmentation based on the technology of image segmentation to improve the image segmentation quality and protect the detail information of the image.The main contents and innovations are as follows:(1)To address the problem which is caused by the segmentation of image that is noise polluted and partially shielded,a self-adaption shape priors image segmentation method is proposed.In this method,the shape priors is represented by a zero level set distance functions(the advantage of this route is that the overall feature and change of the target shape can be captured),the energy function of shape priors for the single fixed shape template is further defined,this energy function is added into the boundary item of the graph cuts energy function,the information of image and shape priors is transferred by the weight edges of the graph;by adaptive adjustment of parameters to adjust the shape prior in the different images;Speeded up robust features and Random Sample Consensus algorithm is used to implement shape template and object registration,the shape of affine transformation invariance.This method is used to cut natural image that corrupted by noises and barrier,compared to the case without shape prior,the object by constraining the adaptive shape prior information of the edge,can effectively respond to shadows,occlusion and noise problems,and obtained the ideal segmentation results.(2)To overcome the unsatisfactory result of the graph cuts by single fixed shape priors template when a great difference exists between undivided target and shape template in the image,the method based on the graph cuts and nonlinear statistical shape prior is proposed.Firstly,the shape priors are represented by the nonlinear feature of shape priors sample which extracted according to KPCA,this means that the input shape templates are registered in the input space,which obtain the training sets.Secondly,the target shape prior is mapped to afeature space with principal component analysis by using a nonlinear kernel function,and the projected shape is obtained,which is mapped back to the original input space to obtain the average shape of the target,and thus forms a new energy function.Thirdly,through the weight coefficient self-adaptive adjustment of the shape prior term,the shape prior term of the energy function becomes adaptive to the image to be segmented.Finally,the image segmentation is accomplished by graph cuts technology so as to minimize the energy function,Experimental results show that the proposed method can not only correctly segment the images which are different than the shape prior templates,but also has a segment results on the images which have the object with occlusion and pollution.Moreover,the proposed method can improve the efficient on image segmentation.(3)To address the problem that the combination of shape cut and single shape prior template cannot segment multiple objects in a graph,a new method based on graph cut segmentation for multiple prior shape is proposed.In this method,the shape is represented by signed distance function,and then the energy function of multiple shape prior is defined by the similarity between the shapes,this model is merged into the regional item of the graph cut framework.The priori energy function is expanded by adding multiple shape priors.The weight coefficient of shape prior item is adaptively adjusted to realize the adaptive control of shape items accounted for the proportion of the energy function.And thus,the problem of artificial selection of parameters is so-lved and the efficiency of segmentation is enhanced.To obtain the invariance of the method proposed in this research for shape affine transformation,the techniques comb-ining the scale invariant feature transform and the random sample consensus are emplo-yed to align.The experimental results indicate that multiple targets in the image can be segmented by the proposed method.Moreover,the image noise pollution as well as occlusion is inhibited.(4)To address the problem of graph cut algorithm cannot get the ideal segment results.a new method based on combines fuzzy C-means algorithm with graph cut image segmentation algorithm is proposed,which first,the image is segmented intoseveral small areas(super pixels)by using the mean shift algorithm,the pixels of the image is replaced by the super pixels obtained as thevertex of the image,and the relationship between of the adjacent pixels is used as the edge to construct the graph model.Secondly,we adoptthe improved FCMA to carry out clustering analysis on the Gauss model mixed with foreground and background.Finally,we use max-flow min-cut algorithm to find the global optimal solution of energy function,and the solution is just the image segmentation result.Experimental results indicate that the proposed method has stronger regional consistency and clearer,smoother edge of the image in segmentation result,moreover,the proposed method has a better segmenting outcome to the noised image.This thesis based on the technique of graph cuts,we have designed different image segmentation models and algorithms to improved the performance of these models in terms of the image segmentation,in the view of the problems existing in image segment-ation.