| How to segment images accurately and quickly is a problem that needs to be solved in the field of image segmentation.It was difficult to meet our requirements of segmentation accuracy that the early days of digital image processing depended on lowlevel visual features as processing elements.Then we try to understand the image content,and transform the image segmentation problem into a mathematical theory problem,which makes a big progress in the field of image segmentation.With the development of artificial intelligence,the artificial neural network algorithm is applied to the field of image segmentation,which makes the accuracy of image segmentation improved.However,the field of image segmentation is still a state of "hundred schools of thought."In order to solve the local optimization problem in the minimum spanning tree structure,and to reduce the degree of over-segmentation and under-segmentation in image segmentation,the maximum inter-class variance(Ostu)was incorporated into the minimum spanning tree method.Firstly,the edge weights between the vertices of the graph are critical to construct the minimum spanning tree.Therefore,the Euclidean distance is calculated in the HSV space and weighted in each channel component to obtain the final edge weight.Secondly,the merging strategy of segmented regions,which incorporates the variance between classes,can effectively reduce the looseness caused by the minimum spanning tree method.It has certain effect on color image segmentation.We use neutrosophic set(NS)theory to remove the indeterminacy factors in the saliency map in order to neutralize the shortcomings of the saliency map and NS theory.Firstly,to overcome the problem of weak edges in the image,we highlight the details and use the guided filter to filter the various channels of the natural image.Then,the initial saliency map is generated.After the weighted superposition of the initial saliency map,the local entropy map and the gray scale map,the final saliency map can be generated using the nonlinear function,and it can effectively highlight the foreground information of the image.Secondly,the saliency map is transformed to the NS domain.According to NS theory,the indeterminacy is reduced,and the segmentation results are finally obtained by using thresholding.A new segmentation algorithm based on NS theory and saliency map,which could reduce the detail in the saliency map and prevent NS theory from filtering out image details excessively.In this way,the effect of undersegmentation and over-segmentation can be neutralized to achieve the purpose of improving the segmentation accuracy. |
