基于非局部先验的雾线优化图像去雾算法

针对非局部先验去雾算法中雾线端点像素位置精确度不足的问题，提出了雾线优化的非局部先验图像去雾算法。首先分析雾线理论，结合暗通道理论确定最大聚类雾线真实端点，以其为已知条件补偿小聚类雾线端点与大气光之间的距离，根据类内不同像素与雾线对应夹角预估单个像素雾线端点进而求得像素级优化后的透射率，最后根据图像局部灰度值差异融合暗通道先验(dark channel prior, DCP)和非局部先验透射率得最终透射率图。将本文算法与其余3种去雾算法在多幅户外雾图下通过主观及客观两方面分析比较，实验结果表明该算法能取得更好的去雾效果，尤其在天空区域图像复原效果较为突出。

Haze-line optimization image dehazing algorithm based on non-local prior

The pixel position of the haze-line endpoint is not accurate enough in the non-local prior dehazing algorithm.To resolve this problem,an image dehazing algorithm based on non-local prior with an optimized haze-line was proposed in this study.We analyzed haze-line theory and combined the dark channel theory to find the real haze-line endpoint of the largest cluster.Then,we took it as the known conditions to compensate the maximum distance between other haze-line endpoints of small cluster and the atmospheric light. According to the different pixels in the class and the corresponding angles,the haze-line endpoint of individual pixel was estimated,and then,the transmission of every pixel after optimization was refined.Finally,local grey value difference fusion dark channel prior (DCP) and non-local prior transmission was used to produce our transmission map.We compared our algorithm with three existing algorithms by applying them to multiple outdoor hazy images through subjective and objective analyses.The experimental results demonstrate that proposed algorithm has a better dehazing effect, especially in the sky region,the image restoration effect is more prominent.