Underwater Fish Image Enhancement Algorithm Based on Dark Channel Prior and Retinex

Due to the intricate underwater conditions and lighting challenges prevalent in marine farms, underwater fish images often suffer from issues like blurriness, low contrast, color distortion, and uneven illumination. In response to these hurdles, we present an underwater fish image enhancement algorithm that leverages the dark channel prior and Retinex. We commence by applying the dark channel prior to dehaze underwater fish images. To discriminate between foreground and background regions within the image, we employ a superpixels segmentation technique based on variations in image brightness. Consequently, we employ a multi-scale Retinex enhancement on each region and perform a secondary enhancement exclusively on the background region. Finally, a Poisson fusion method is utilized to amalgamate the enhanced foreground and background segments, resulting in a unified composite image. We conducted comprehensive experiments employing the Underwater Image Enhancement Benchmark Dataset (UIEBD). The experimental findings substantiate the efficacy of the proposed algorithm, with notable improvements observed across three key evaluation metrics: information entropy, average gradient, and standard deviation. These enhancements translate to an average percentage increase of 18.22%, 133.67%, and 59.29%, respectively. Additionally, the algorithm notably enhances underwater fish image clarity and ameliorates issues related to uneven illumination. It demonstrates superior aptitude in preserving image intricacies and elevating contrast when juxtaposed with alternative algorithms, substantiating its superior performance.