基于颜色失真去除与暗通道先验的水下图像复原

Underwater Image Restoration Based on Color Cast Removal and Dark Channel Prior

水下图像成像过程与雾天图像虽然类似,但因水对光的选择性吸收和光的散射作用,水下图像存在颜色衰减并呈现蓝(绿)色基调,传统的去雾方法用于水下图像复原时效果欠佳。针对这类方法出现的缺点,该文根据先去除颜色失真后去除背景散射的思路,提出一种新的水下图像复原方法。结合光在水中的衰减特性,提出适用于水下图像的颜色失真去除方法,并利用散射系数与波长的关系修正各通道透射率;另外,该文改进的背景光估计方法可有效避免人工光源、白色物体、噪声等影响。实验结果证明,该文方法在恢复场景物体原本颜色和去除背景散射方面效果良好。

The imaging process of the underwater image is similar to the haze image. However, the dehazing methods fail when used in the underwater image restoration because of the color attenuation and blue （green） color tone, caused by the selective absorption of water and light scattering. Thus, this paper proposes a new approach for underwater images restoration based on the idea of removing backscattering after the color cast removal. Due to the attenuation of light in water, a color cast removal approach is proposed. The relationship between scattering coefficient and wavelength is used to obtain a more accurate transmission estimation for each color channel. In addition, an improved algorithm for background light estimation is presented, which can effectively avoid the influence of artificial light, white object and noise. Experimental results demonstrate the effectiveness of the proposed method in restoring the original color of the scene and removing the backscattering.