Adaptive weighted multiscale retinex for underwater image enhancement
Dayi Li,Jingchun Zhou,Shiyin Wang,Dehuan Zhang,Weishi Zhang,Raghad Alwadai,Fayadh Alenezi,Prayag Tiwari,Taian Shi
DOI: https://doi.org/10.1016/j.engappai.2023.106457
IF: 8
2023-05-24
Engineering Applications of Artificial Intelligence
Abstract:Vision-dependent underwater vehicles are widely used in seabed resource exploration. The visual perception system of underwater vehicles relies heavily on high-quality images for its regular operation. However, underwater images taken underwater often have color distortion, blurriness, and poor contrast. To address these degradation issues, we develop an adaptive weighted multiscale retinex (AWMR) method for enhancing underwater images. To utilize the local detail features, we first divide the image into multiple sub-blocks and calculate the detail sparsity index for each one. Then, we combine the global detail sparsity index with the local detail sparsity indices to determine the optimal scale parameter and corresponding weights for each sub-block. We apply retinex processing to each sub-block using these parameters and then subject the processed sub-blocks to detail enhancement, color correction, and saturation correction. Finally, we use a gradient domain fusion method based on structure tensors to fuse the corrected and enhanced sub-blocks and obtain the final output image. Our approach improves underwater images through comparisons with current state-of-the-art (SOTA) techniques on several open-source datasets, both quality, and performance.
automation & control systems,computer science, artificial intelligence,engineering, electrical & electronic, multidisciplinary