基于马尔科夫随机场框架的单幅图像去雾*

雾或霾等天气会降低场景的能见度,给机器视觉的后续处理造成影响。针对图像雾霾退化的恢复、及现有基于马尔科夫随机场图像去雾算法的缺陷,提出了一种新的基于马尔科夫随机场和暗通道先验的图像去雾算法。该算法以雾天条件下退化模型为基础,通过介质传输图和原始无雾图像的约束条件,利用暗通道先验获取介质传输图的粗估计,构造MRF框架下的代价函数。为使去雾图像保持更多的纹理细节,引入纹理检测函数改进代价函数,最终求得去雾图像和介质传输图。实验结果表明,本文方法可以得到较好的去雾效果,同时保持较多的纹理细节和更快的运算时间。

Single image defogging based on Markov Random Field

The weather, such as fog or haze can significantly degrade the visibility of a scene, which is a major problem for many application of computer vision. To resolve the problem of image defogging and the defects of existing image enhancement algorithm based on markov random field, this paper proposed a novel method, which based on the markov random field and the dark channel prior. This algorithm is based on the degradation model under the condition of fog, using dark channel prior to achieve the coarse estimation of the optical transmission, through the constraints of the optical transmission and the original image, developing a cost function in the framework of markov random field. To keep more texture details, introducing the texture detection function to make better the cost function, and finally obtaining defogging image and the medium transmission. The experimental results show that this method can get better defogging image, while maintaining good texture details, and faster operation time.