An Improved Unsupervised Color Correction Algorithm for Underwater Image

Aiming at the problems of low contrast, dim, noise interference and color distortion caused by the attenuation and scattering of light in underwater propagation, an improved unsupervised color correction algorithm is proposed. Firstly, according to the attenuation characteristics of light R, G and B in underwater environment, color components of R, G and B are balanced and contrast enhanced; secondly, the image is mapped into HSV color model space, and the contrast of S and V components are stretched; and finally, the details and contours of the underwater images are enhanced by a method of limiting contrast adaptive histogram equalization. Experiments are carried out on the underwater image enhancement reference data set (UIEBD), the results show that the proposed algorithm can correct the underwater image color in different complex environments well. And contrast, saturation and image details of the enhanced underwater image are all improved, and the algorithm is evaluated by MSE, PSNR, SSIM UQI and VIF. Compared with deep learning based algorithm (DUIENet), the average value of indexes MSE with the proposed algorithm is decreased by 4.32%, SSIM and VIF are increased by 1.12% and 26.79%, respectively; compared with traditional algorithms, the total average value of indexes MSE with the proposed algorithm is decreased by 62.9%, PSNR, SSIM, VIF and UQI with the proposed algorithm are increased by 16.6%, 8.43%, 14.52% and 16.88%, respectively.