| With the popularization and application of electronic equipment,a large number of pictures are produced every day,so images have become one of the important ways to transmit information.Processing image information through computer technology has become a recognized method,and image segmentation is one of them.Image segmentation is the basis of image processing and computer vision,so its quality will directly affect the process of subsequent image processing.Therefore,the technology of image segmentation is more and more important.In the medical field,various medical image technologies are becoming more and more mature,but medical images often have noise and artifacts.In view of this,this paper proposes two segmentation and fusion frameworks,which can remove noise while preserving relatively complete image information.First,an image segmentation and fusion algorithm based on clustering by fast search and find of density peaks(DPC)and Markov Random Field(MRF)is proposed.First of all,this method applies the traditional DPC to the grayscale image,and combines it with the MRF model to build a new image segmentation framework to avoid the shortcomings of Markov random field that depends on the initial marker field and needs to manually select the classification number.In order to reduce the impact of image noise,this algorithm adopts local Laplacian filtering technology to obtain a smooth image of the source image.However,simple denoising preprocessing usually affects some normal areas with high-frequency information.Therefore,we introduce a fusion method that uses majority voting rules to re partition ambiguous pixels,thereby obtaining a unified segmentation label and improving the accuracy of image segmentation.The source image and smooth image are passed into the proposed segmentation method to obtain two different results,and the fusion method is used to merge the two results to obtain the final segmentation result.Secondly,a multi-threshold(IIMT)image segmentation and fusion algorithm based on interval iteration is proposed.Compared to most other multi threshold methods,IIMT iteratively searches for sub regions of the image to achieve segmentation,rather than treating the original image as a whole.The algorithm iteratively applies Otsu single threshold algorithm in the gray histogram of the original image,and completes the classification through the combination of class mean and threshold.In addition,in order to reduce the impact of noise on the image,this algorithm uses the hybrid L1-L0 layer decomposition method to obtain the base layer of the original image,and then inputs the source image and the base layer image into the interval iterative threshold for segmentation,and finally fuses the two segmentation results.In this paper,the two algorithms are tested on some public medical images.The experimental results show that the segmentation results of the two algorithms in this paper have good visual effect,clear gray level,and high rating index score.In addition,this algorithm is compared with some other algorithms,such as particle swarm optimization(PSO)and bacterial foraging algorithm(BF).The experimental results show that the algorithm in this paper has higher evaluation index results and stronger anti-noise ability. |
