基于改进暗通道先验的图像去雾算法

Image Defogging Algorithm Based on Improved Dark Channel Prior

针对暗通道先验去雾算法在含有浓雾、亮白、非均匀光照区域造成的图像失真的问题,提出了一种改进的自适应局部阈值分割和自适应参数优化相结合的去雾算法。首先根据暗通道先验理论运用局部阈值分割出亮白区域和非亮白区域,然后采用引导滤波将求取的原始透射率进行细化,并通过亮白区域与非亮白区域加权求取更加精准的大气光强,提高了大气光强的鲁棒性,使得该算法适用于暗通道去雾效果不好的浓雾高亮区域和非均匀光照区域。最后,通过雾天图像降质模型恢复出无雾图片,将该算法与几种常用的去雾算法进行比较。结果表明,该算法在绝大多数情况下恢复的图片清晰自然,解决了图像去雾后视觉效果不好的问题,同时也有效改善了亮白区域色彩失真的现象。

To solve the problem of image distortion caused by dark channel prior defogging algorithm in areas with dense fog,white light and non-uniform illumination,an improved defogging algorithm which combined adaptive local threshold segmentation and adaptive parameter optimization was proposed.Firstly,according to the dark channel priori theory,the local threshold was used to divide the bright white region and the non-bright white region.Then the original transmittance was refined by guiding filter.After that,the more accurate atmospheric light intensity was obtained by weighting the bright white region and the non-bright white region,which improved the robustness of atmospheric light intensity.Therefore,this algorithm was suitable for the dense fog highlight area and non-uniform light area with poor fog removal effect in dark channel.Finally,the fog-free image was restored by the image degradation model.Comparing the algorithm with several common defogging algorithms,the experimental results showed that the image restored by the algorithm was clear and natural in most cases,which solved the problem of poor visual effect after defogging and effectively improved the color distortion in the bright white area.