Underwater Image Enhancement Method Based on Dark Channel Prior and Guided Filtering

Abstract

This paper presents a comprehensive enhancement method based upon Underwater Dark Channel Prior (UDCP) and Guided Filtering for standard RGB underwater images without depth information. Firstly, color compensation and Gray World Algorithm are used to correct the color of images obtained underwater. After that, the restored image is dehazed by using the optimized dehazing algorithm created on UDCP. The dehazing algorithm proposed in this study is obtained by reconstructing the ambient light transmittance expression in UDCP. It effectively avoids the “excessive dehazing” caused by traditional dehazing algorithms, and it can also optimize the depth of field of dehazed images. At the same time, due to the complexity of underwater dark channel image dehazing, the dehazed image will still produce fuzzy white areas in zones with large pixel color difference changes (such as the boundary of objects). Therefore, our method adds an image fusion approach built upon guided filtering to optimize the dehazed image to eliminate the white areas to enhance the image clarity further. At last, this paper compares the image enhancement effect of our method with that of other four methods such as Unified Generative Adversarial Networks (UGAN) by using five objective image evaluation indexes such as Underwater Color Image Quality Evaluation (UCIQE).