Research on adaptive enhancement of robot vision image based on multi-scale filter
Qin Dong · International Journal of Image and Data Fusion · 2022
Contrast enhancement and histogram equalisation are two image enhancement methods, which can lead to changes in the edge position of the resulting image, blurring or even loss of details. Therefore, this paper introduces a multi-scale filter to adaptively enhance the robot visual image, improve the brightness of the robot visual image, enrich the image details and reduce the image enhancement time. According to Retinex theory, the characteristic information of robot visual image is obtained, the logarithmic domain operation form of Retinex algorithm is obtained, the robot visual reflection image of high-frequency part is determined, the robot illumination visual image is estimated by multiscale filter, and the scale constant of Gaussian filter is obtained; According to the Retinex algorithm of weighted guided filtering, the robot visual image enhancement process is designed. The experimental results show that the average value of the robot visual image enhanced by this method is 88.63, the standard deviation is 62.78, the information entropy is 8.18, the robot visual image enhancement time is only 5.9s, and the PSNR of the robot visual image is up to 39.92, which proves that the robot visual image enhancement effect of this method is good.