Deep-Sea Image Enhancement Using Multi-Scale Retinex with Reverse Color Loss For Autonomous Underwater Vehicles

Advances on underwater optical imaging has been rampant recently because of numerous applications it offers and the increasing use of underwater vehicles. Unfortunately, due to the inherent property of light in water, images tend to be hazy, have non-uniform illumination and colors get distorted. Retinex algorithms were formulated for dynamic range compression and tonal rendition to fix the illumination and color distortion problem. Multi-scale Retinex, an extension of the original Retinex algorithm, has been proven to enhance images. However, Retinex Theory is based on human visual system in air medium. Due to this, light attenuation is not considered. Through Beer-Lambert Law, the light attenuation percentage can be estimated and added back to the original image in a process the researchers call Reverse Color Loss. Thus, combining Multi-scale Retinex and Reverse Color Loss, an enhancement algorithm for deep-sea application for the use of AUVs is proposed.