Underwater Image Restoration Based on Local Depth Information Prior

Abstract

Underwater images often suffer from color distortion and loss of contrast. This is due to the absorption and scattering of light as it travels through water. Although the physical process of underwater imaging is similar to that of haze images in the air. However, traditional dehazing methods cannot produce good results due to the different attenuation of light under different wavelengths in underwater conditions. To overcome this problem, we propose a novel underwater image restoration method based on local depth information priors. First, we use a computer vision-based multi-view geometry method to estimate the local depth information of the image for parameter estimation of the depth compensation model. According to the characteristics of underwater optical imaging, we introduce an underwater color correction method using depth compensation. Second, we propose a method for estimating the global depth image with local depth information priors. Finally, we adopt the global depth image to recover the underwater image. Experimental results demonstrate that the recovered images can achieve better visual quality of underwater images compared to several state-of-the-art methods.