Research On Image Retrieval Algorithm Based On Region Of Interest

With the development of big data technology,images have become an indispensable way to obtain information in our daily life.Although the research on image processing and image retrieval has been very in-depth and has made great achievements,there are still many shortcomings that need to be optimized.In view of this,based on the content-based image retrieval technology,this paper analyzes the main part of the image,and proposes an image retrieval method based on the region of interest(ROI).The main contents of this paper are as follows:1.Method of extracting regions of interest in single-subject images based on color features is proposed.Interested area is the main part of the image,image can be used to replace the entire image retrieval task,interested in using the edge of the area unique shape features and evident in the contour,the advantages of accurate interested region is proposed.The influence of irrelevant information such as background is eliminated to the maximum extent,and only the image main body is retained to match the image in the image library,so as to enhance the accuracy of retrieval.2.An image enhancement method combining spatial domain and frequency domain is proposed to process the image to highlight the region of interest.In order to make the proposed ROI more complete and accurate,the ROI of the image is enhanced before extraction to make the characteristics of the image body more significant.The first to use bilateral filtering in image noise reduction processing,on the one hand can realize image denoising,on the other hand also can keep the edge details of image body.Then the wavelet transform is used to process the image and highlight the main body as the low-frequency component,so as to enhance the characteristics of the region of interest,so as to better propose the region of interest,and further enhance the retrieval efficiency.3.Comparative analysis of experiments.The ROI enhancement method and ROI extraction method are tested by using the standard image library,and the ROI mean square error and peak signal-to-noise ratio are proposed by comparing with other algorithms to prove the advantages of the proposed algorithm.By comparing the average precision of various methods,the retrieval accuracy of the proposed algorithm is proved.