Enhanced Color Sensing and Recognition of Underwater Color Using Robust Adaptive Tone Mapping

Abstract

Sensing images in the underwater environment is a significant issue in ocean engineering. Acquiring clear underwater images involves many challenges, such as climate, environment, and human factors. The most important problems are the fogging effect caused by the dispersion of light and the energy of each light wavelength when it propagates in water. Then, a color cast is caused by inconsistent attenuation. A common issue is the dispersion of light that occurs in underwater photography, which can impact the overall color balance of an imager. While current research can make use of good approaches for obtaining good visual quality and quantitative indicators, having a wider color gamut space and a dynamic image range can improve visible details. Therefore, we propose a module for enhancing underwater color image sensing with robust adaptive tone mapping for inferring degradation models using deep learning models and with adaptive tone mapping for further improving the image dynamic range. We address issues with limited dynamic range and brightness in underwater image sensing and recognition using a robust adaptive tone mapping method. Quantitative and qualitative results show that our method performs relatively well in the Underwater Image Enhancement Benchmark dataset compared with other recent methods that apply appropriate tone mapping to the large-scale layers of the image to preserve details and avoid over-enhancement. Therefore, the color gamut of our augmented image has a large scale and is evenly distributed when visualized in the Y'CbCr color space. In the future, our research method is expected to be applied to different types of underwater work and environment, and to reduce the severe degradation problems that usually occur in underwater images.