基于熵加权的四叉树分解单帧图像去雾

Single image dehazing algorithm based on quadtree decomposition of entropy weighting regularization

为解决现有图像去雾算法运行效率低下且对白亮目标区域色彩失真的问题,提出一个基于四叉树分解与熵加权上下文正则的单帧图像去雾算法。原始雾图像估计出边界的约束量并进行四叉树分解,利用输入图像的统计特性,自适应地获得逐块的传输函数。引入每一个颜色通道的熵,改善权值函数的精度,通过最小化优化函数计算最终传输映射。该算法可获得具有最逼真的颜色与最少的光晕效应的高质量无雾图像。仿真的定性定量结果表明,该算法相比于现有算法能够获得最佳去雾效果,极大降低计算复杂度。

A single image dehazing algorithm based on quadtree decomposition was proposed and it was combined with the entropy weighting contextual regularization to improve the dehazing performance. To achieve the effects , the boundary constraint was es-timated ,and the transmission function of sub-patches was obtained adaptively using statistical properties of input image. The en-tropy weight of each color channel was introduced to improve the accuracy of weight function to minimize the optimization func-tion to compute final transmission mapping. The proposed algorithm can obtain high-quality dehazed image with better color effects and minimal airlight artifacts. Comparing with the existing algorithm , simulation results show the proposed algorithm is superior.