Efficient underwater image restoration utilizing modified dark channel prior
Sheezan Fayaz,Shabir A. Parah,G. J. Qureshi
DOI: https://doi.org/10.1007/s11042-022-13828-6
IF: 2.577
2022-10-06
Multimedia Tools and Applications
Abstract:An effective algorithm called Modified Underwater Dark Channel Prior (MUWDCP) based on the image prior for haze removal in underwater images is proposed. The proposed algorithm computes underwater dark channel information using only blue-green channels exploiting the fact that red color due to low frequency undergoes high absorption and drops at about a depth of 3 m. To address the issue of atmospheric light estimation and enhance its robustness, we propose a novel global atmospheric light estimation method based on arithmetic Mode operation. MUWDCP does not rely only upon the bright pixels of the underwater dark channel for atmospheric light estimation instead, it computes global atmospheric light from the degraded image itself. Unlike most underwater image restoration methods, which estimate transmission maps of only one channel, MUWDCP estimates transmission of all three-color channels while maintaining less computational complexity. MUWDCP has been found to provide significantly improved visual quality of the restored underwater image. The proposed method has been evaluated using subjective and objective quality metrics like Entropy, Natural Image Quality Evaluator (NIQE), Underwater Color Image Quality Evaluation (UCIQE), and Underwater Image Quality Metrics (UIQM). The experimentation shows that MUWDCP performs better as compared to various state-of-the-art algorithms. The proposed algorithm proffers Entropy, NIQE, and UIQM scores of about 7.18, 2.88, and 4.57 respectively, which are better than the state-of-the-art. Thus, the results demonstrate the effectiveness of the proposed scheme.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering