基于明暗像素先验的随机游走图像去雾

针对雾霾天气下大气粒子的散射作用导致户外图像质量严重退化和暗通道对天空区域估计失效的问题,提出了一种单幅图像去雾算法。该算法从雾天图像模型出发,首先利用估计天空区域更加准确的明暗像素先验获取介质传输图粗估计,然后在随机游走模型的框架下,将粗估计的介质传输图作为先验约束传统随机游走能量模型,进一步优化介质传输图,得到最终的无雾图像。实验结果表明,该算法有效地在随机游走模型下结合了明暗像素先验对雾天图像估计的优缺点,证实了所提方法的可行性和有效性。

Random walk image dehazing through a priori of light and dark pixel

To solve the problem of severe degradation of outdoor images caused by the scattering effect of atmospheric particles in foggy or hazy weather and failure of dark channel for sky areas estimation, this paper puts forth an algorithm of single image dehazing. Based on the image models in foggy weather, the algorithm works out a rough estimation of the medium transmission images through a priori of light and dark pixel which is more accurate for sky areas estimation. Then under the framework of random walk model, by using roughly estimated medium transmission diagram as a priori to constrain the traditional random walk energy model, and further optimizing the medium transmission, accurate images are eventually achieved. The experimental results show that the proposed algorithm effectively integrates the advantages and disadvantages of a priori of light and dark pixel for image estimation in fog and haze in random walk model, and verifies the feasibility and effectiveness of the proposed method.