Research on Interest Region Detection of Images Based on Enhanced Visual Quality

With the development of hardware, image pixels are constantly improving. At the same time, the pixel size is constantly increasing, resulting in a large image dataset to be processed, and traditional image segmentation algorithms can no longer meet the requirements. In order to detect regions of interest quickly and accurately in stereo images, a stereo image region of interest detection method combining visual perception characteristics is proposed. Analyze multiple important visual characteristics such as human visual sensitivity and visual persistence, and use spatiotemporal joint analysis of damaged stereo images to improve viewpoint synthesis, effectively suppress synthesis distortion, and enhance the visual quality of stereo images. Train a prior model that is more in line with human visual characteristics by using eye tracking data and stereo image data obtained from experimental markers as input samples for the artificial neural network, and fuse all features to form a salient image. Based on salient images, locate regions of interest, and achieve detection of regions of interest in stereo images. The experimental test results have shown that the proposed method can effectively improve the accuracy of detection results and reduce detection time.