全局平滑和边缘保持的单幅图像去雾

Global smoothness and edge preservation for single image dehazing

针对现有基于暗原色先验理论的去雾方法在天空区域容易产生失真和边缘定位不准确的问题,提出了一种雾天图像直接去雾方法。根据雾天成像模型和空间变化图像复原思想,构建了数据项;通过深入分析天空区域产生失真的原因、透射率图像和复原图像的边缘特征,构建了约束项,并通过线性组合数据项和约束项,构建了一个能量泛函;利用分步梯度下降流法最优化该能量泛函,实现了复原图像的精确求解。实验结果表明,与传统方法相比,该方法不但能更好地抑制天空区域失真现象的产生,也能更精确地定位复原图像的边缘。

The methods based on dark channel prior have two disadvantages: the first is that the recovered image fails to achieve good effects in correspondence of object edges in image; the second is that images fail to contain a larger area of the sky. In order to solve the problem, a novel method is proposed. Firstly, based on haze imaging model and image restoration theory, a data term is proposed. Secondly, a constraint term is proposed based on edge feature of transmission and image, and then the function is proposed which is based on the linear combination of data term and constraint term.Finally, accurate restoration image is obtained, through stepwise gradient descent method to optimize the function. Experimental results clearly indicate that the proposed method performance is better in comparison with traditional methods.