Underwater Image Enhancement by Joint Lab and Opponent Color Spaces

Underwater images suffer significantly from severe color distortion and blurring due to the different attenuation of light as well as the scattering of light in underwater environments. To obtain clear underwater images, we propose a novel underwater image enhancement model based on joint Lab color and opponent color spaces. Lab color space can separate luminance and chrominance, and express rich color information based on human visual perception. Opponent color space is based on the biological vision mechanism, which divides colors into three opponent channels: red-green, blue-yellow and light-dark, and can resist the influence of light source variation. Our model utilizes the advantages of these two color spaces to restore the underwater images.We design a dual color space encoder-decoder module that fuses Lab and RGB color spaces to enhance degenerated images. Moreover, we also use an adaptive light source estimation module to learn a light source map from the opponent color space to correct the color deviation. Extensive experiments demonstrate that our proposed method achieves better enhancement results.