Enhancing Underwater Imagery Using Multicriteria Decision-Making with Machine Learning Techniques
Ganesh S. Khekare, Shivani Kerai, Anil Vithalrao Turukmane, Urvashi Khekare, Rahul P. Sharma, Rahul Agrawal · 2024
This chapter focuses on the pivotal task of image enhancement as a foundational step in underwater image captioning followed by underwater object detection for advanced marine exploration; given the indispensable role of image quality in downstream tasks, the necessity of underwater image enhancement becomes even more pronounced in the context of image captioning. Employing sophisticated machine learning techniques, particularly histogram equalization and manual white balancing, the authors of this chapter delves into the intricacies of refining underwater images, harnessing histogram equalization to optimize the distribution of pixel intensities, thereby enhancing the overall contrast and revealing subtle details submerged in the aquatic environment. Complementing this, incorporating manual white balancing fine-tuned the color representation, mitigating the distortions induced by the complex underwater light conditions. Combining these image enhancement methodologies set the stage for robust underwater object detection and ultimately contributing to the overarching goal of augmenting marine exploration capabilities. The improved image quality by using multicriteria decision-making not only enhances the interpretability of detected underwater objects but also empowers the subsequent captioning model to generate more accurate and descriptive textual narratives, thus enriching the overall marine exploration experience.