Underwater Image Enhancement Based on Red Channel Correction and Improved Multiscale Fusion.

The underwater environment is complex and everchanging, and color distortion and reduced contrast are the two primary issues encountered in original underwater images. This paper applies image enhancement to address these concerns by proposing a novel image color correction (RAG) algorithm and an improved multiscale image fusion (IMF) algorithm for each problem. In terms of color correction, traditional algorithms typically implement a single color correction step, leading to issues such as locally unrealistic colors and limited applicability in certain scenes. The proposed RAG algorithm advances upon these methods by further dividing the input image into multiple local blocks after performing global color correction using the gray world algorithm. Secondary color corrections are then applied within each local block, overcoming the limitations of the traditional single-step color correction process. The corrected images achieve a UICM mean value of 5.4671, representing an enhancement exceeding 20%. In contrast, the existing fusion algorithms often neglect the distinct characteristics of different channels and fail to maximize the advantages of fusion. The IMF algorithm proposed in this paper calculates weights in the RGB and LAB color spaces of images to effectively incorporate channel disparities, refining the weights to provide a more thorough fusion process. The corrected images achieve a mean UCIQE value of 0.630 and a mean CCF value of 37.495. Significant improvements are observed across various other metrics.