Abstract: For all researchers, recovering lost colours in underwater photographs with complete flavour is still a difficult undertaking. A physically realistic model was provided, and it was demonstrated that the atmospheric image creation model's broad application to underwater photographs was partially to blame for this problem. The updated model demonstrated that: 1) the attenuation coefficient of the signal is not constant throughout the scene but rather depends on reflectance and object range; and 2) the back scatter coefficient, which controls the increase in backscatter with distance, is different from the attenuation coefficient of the signal. Here is the method that uses RGBD photos to recover colour using an updated model. Using the darkest pixels and their known range information, the Sea-thru approach estimates backscatter. The range-dependent attenuation coefficient is then obtained using the spatially variable illuminate. It is demonstrated that the approach utilizing the updated model performs better than the models utilizing the atmospheric model. With the aid of potent computer vision and deep learning algorithms, the regular removal of water will lead to the opening up of enormous underwater datasets for technical development, intriguing potential for underwater research, and conservation.

Keywords: Underwater, image, colour, Sea-thru

Cite:
Dhulipalla Tejaswi, Yerru Charitha, Menikonda Harika, Tummalacharla Sreshta,"Underwater Image Enhancement using Deep Learning", IARJSET International Advanced Research Journal in Science, Engineering and Technology, vol. 11, no. 3, 2024, Crossref https://doi.org/10.17148/IARJSET.2024.11342.

| DOI: 10.17148/IARJSET.2024.11342