基于改进暗原色先验模型的快速图像去雾方法

为了解决雾天图像质量退化问题，结合改进的暗原色原理与容差机制提出一种快速图像去雾算法。该算法首先基于暗原色先验估计大气参数，然后利用插值算法和最大最小估计法改进暗原色先验模型进而准确计算出不同场景深度的透射率，最后结合容差机制基于大气散射模型恢复无雾图像。实验结果表明，相比于原有的暗原色先验算法，该算法的计算速度可提高至少30倍，并且能够同时实现明亮与暗色区域的有效去雾，去雾图像清晰自然。基于插值算法与最大最小估计法改进的暗原色先验去雾模型可同时保证去雾处理的鲁棒性和实时性。

Fast image de-hazing method based on improved dark channel prior model

To deal with the image degradation problem induced by the haze or fog, a fast image de-hazing algorithm based on an improved dark channel prior model and tolerance mechanism is proposed. The atmosphere light is firstly estimated using dark channels, and then for calculating accurately the transmission map of the image with different scene depths, the dark channel prior model is improved by combining the interpolation and the minimum-maximum estimation methods, finally the de-hazing image is restored using the tolerance mechanism and atmospheric scatting model. The experimental results illustrate that, comparing with the original de-hazing methods based on dark channel prior, the proposed algorithm is very fast with a computation time reduced to 1/30 and effective for fog removal in both dark and light image areas. The improved dark channel prior model by combining interpolation and minimum-maximum estimation methods has a potential for real-time and robust image de-hazing.