DESIGN OF FRAMEWORK FOR ENHANCEMENT OF UNDERWATER IMAGES

Abstract

Underwater image processing has received tremendous attention in few past years. In the last few years underwater image processing has attracted much attention because of its importance in marine engineering and aquatic robotics. The reason for increased research in this area is due to the process of image taken in under water. When we capture the image in under water then the quality of image is degraded. To address this problem, we need some other methods to increase the quality of image while capturing it under water. But capturing the image in normal circumstances as well as in under water are same, thus once we get an image, some mechanism to increase the quality of captured image will be required. Though some methods are already present in image enhancement and restoration, still some comparative and deep survey is required to improve the image quality. Various algorithms have been proposed for underwater image enhancement, but for their assessment either synthetic datasets or few selected real-world images are used. There are few latest underwater image enhancement methods based on deep learning and machine learning. These methods not only enhance the images but also provide better results as compared to enhancement and restoration method. These deep learning-based methods increasing the variety in underwater imaging, enhancing underwater image quality and providing wider scope in terms of improvisation. This research work presented the four significant contributions in the underwater image enhancement system. v First, we have conducted a literature review of underwater image enhancement systems to highlight the challenges of the existing work and identify some good quality works in the domain of underwater image enhancement. A complete and in-depth study of relevant accomplishments and developments, particularly the survey of underwater image methods and datasets, which are a critical issue in underwater image processing and intelligent application, has been done. In this, we first provide a review of more than 100 articles on the recent advancements in underwater image restoration methods, underwater image enhancement methods, and underwater image enhancement using deep learning algorithms, along with the techniques, data sets, and evaluation criteria. To provide a thorough grasp of underwater image restoration, enhancement, and enhancement using deep learning, we explore the strengths and limits of existing techniques. Second, we developed a robust model for improving the quality of underwater images using enhancement techniques. This technique is split into two sections. The first section focuses on boosting contrast, while the second section focuses on improving color. Our enhanced results stand out for their brilliant color, greater contrast, and enhanced features. When compared to other approaches this technique improves image quality by increasing entropy, peak signal to noise ratio (PSNR), and underwater color image quality evaluation (UCIQE) values while lowering mean square error (MSE). It is an entirely algorithm-based technique that is independent by image datasets. The images used to evaluate the results come from a variety of datasets, and their enhanced performance confirms their robustness. Because of its single image-based approach, our method is very vi compelling in terms of processing speed. Comprehensive findings on a variety of underwater image datasets demonstrate that our approach outperforms the vast majority of them. Third, we designed an underwater image enhancement framework to recover deep sea images. In deep sea underwater images, the uneven attenuation of sunlight, when it spreads underwater, they have high color distortion and very low intensity. Furthermore, the amount of attenuation changes with wavelength, yielding in asymmetric color traversing. As the research, this framework demonstrates that assigning the appropriate context based on the color channel traversal range may result in a significant performance speedup for the objective of underwater image enhancement. Furthermore, it is critical to reduce irrelevant multi-contextual characteristics and improve the model's representational strength. Therefore, we included an important reduce method to dynamically modify the learnt multi contextual characteristics. DeepSeaNet, the suggested framework, is enhanced via conventional pixel-wise and feature-based estimation methods. Comprehensive tests were conducted to demonstrate the efficiency of the proposed technique with the best published paper on standard datasets. Fourth, we do a comparative result analysis of the developed models with the other existing techniques. The comparative analysis shows that the proposed system is better than the existing techniques. The experimental results, analysis, and performance evaluation demonstrate that the proposed work provides feasible and efficient techniques. Thus, this research work successfully provides an effective and optimal underwater image enhancement system.