Research On Convolutional Neural Network Based Image Segmentation Quality Evaluation And Segmentation Repairing Method

With the development of the information age,the amount of multimedia data has grown rapidly and images have become the main carriers of information.How to extract key information from massive image data quickly and effectively has become a basic requirement of many emerging industries.Among them,the first step in information acquisition is image segmentation,which provides analyzable areas for subsequent tasks such as high-level identification and understanding.However,the image data is massive and diverse,and extracting semantic regions from the image is generally recognized as a morbid problem.In actual segmentation applications,there are often incorrect segmentation results,which seriously interfere with the subsequent high-level analysis of the image.Achieving the evaluation and repair of image segmentation quality has positive significance for improving the performance of high-level image recognition and understanding.Segmentation quality evaluation and repair is dedicated to the characteristics of the segmentation results,evaluate the quality of segmentation results,and perform segmentation and restoration.At the same time,segmentation quality evaluation and repair have many challenges,including how to accurately describe the segmentation quality,how to effectively combine segmentation results with the original image information,and how to use evaluation methods to solve practical problems.Focusing on the above challenges,this paper has carried out a study on segmentation quality evaluation method based on convolutional neural network and segmentation result restoration method based on segmentation quality evaluation.The specific content is as follows:1.For the segmentation quality evaluation modelling problem,this paper proposes a segmentation quality evaluation method based on convolutional neural network and constructs a new evaluation model.The model analyzes and fuses the correlation between the original image and the segmentation result,and multi-scale analysis is used to describe the multi-level segmentation information,which effectively improves the segmentation quality evaluation performance.2.Based on the segmentation quality evaluation,it is difficult to determine the optimal parameters in the interactive segmentation process.This paper effectively combines segmentation evaluation results,and constructs an interactive segmentation optimization model of adaptive parameter selection.The segmentation parameters are automatically selected for different images and the solution is effectively improved.The performance of the interactive segmentation method.3.Further for collaborative segmentation results,this paper proposes a new multiimage collaborative segmentation and repair algorithm,constructs a collaborative segmentation quality evaluation model based on contour and region analysis,and designs a collaborative segmentation and restoration solution based on segmentation quality evaluation,aiming at existing collaborative segmentation results.An iterative repair algorithm was proposed to effectively improve the performance of collaborative segmentation.