基于加权L1正则化的水下图像清晰化算法

摘要

水体对光能量有较强的吸收和散射作用,造成水下图像颜色失真,对比度下降.传统的图像增强方法和复原方法处理水下图像时各有不足,该文结合水下成像物理模型和基于Retinex理论的图像增强算法,提出水下图像清晰化方案.首先,基于图像统计特性给出一种简单的颜色校正方法,以去除颜色失真;在水下图像成像理论框架下,利用边界约束求得初始透射率,再使用自适应维纳滤波进行优化;在此基础上,提出加权L1正则化模型对亮度层进行增强,最后再进行自适应Gamma校正.实验结果表明,算法可以有效去除颜色失真,而且能够大幅提升图像的对比度和清晰度.%Due to the absorption and scattering when light is traveling in water,there are two major problems of underwater imaging:color distortion and low contrast.Traditional enhancement and restoration methods can not handle these problems very well,so,this paper proposes a new approach based on the underwater optical imaging model and a Retinex-based enhancing approach.Firstly,a simple color correction method based on statistical method is adopted to address the color distortion.Then the adaptive Wiener filter is used to optimize the initial transmission map with the boundary constraints.In order to make the result more naturalness,a weighted L1 regularization model is proposed to enhance the luminance layer.Finally,an adaptive Gamma correction operation is adopted for post-processing.Experimental results demonstrate the effectiveness of the proposed method in restoring the original color of the scene and enhancing image contrast and the visibility.