区域透射率融合的暗通道图像去雾方法

Dark Channel Image Defogging Method Based on Regional Transmittance Fusion

针对传统基于暗通道先验的去雾方法容易导致边缘区域和天空区域分别出现伪影和颜色失真的问题,提出了一种基于区域透射率融合的暗通道图像去雾方法。首先,将含雾图像分为非天空、天空及过渡边缘等3个区域。其次,在暗通道结合块透射率估计和点透射率估计分别对非天空区域和过渡边缘区域进行融合,并利用梯度域引导滤波进行平滑处理,得到融合后的暗通道透射率。然后,合成天空区域的亮度透射率和暗通道融合透射率,以获得最终的透射率。最后,利用得到的透射率和改进后的大气光值对图像进行复原,得到去雾结果图像。实验结果表明,与传统暗通道方法相比,所提方法明显更优,其在有效抑制边缘伪影的同时能够较好地保留含雾图像的颜色特征。所提方法得到的去雾图像在主观评价和客观评价方面均能取得更好的结果。

In order to solve the problem that the traditional defogging method based on the dark channel priori is easy to cause artifact and color distortion in the edge region and the sky region,a dark channel image defogging method based on regional transmittance fusion is proposed.First,the fog image is divided into three regions:non sky,sky,and transition edge.Then,the nonsky region and transitionedge region are fused in the dark channel combined with block transmittance estimation and point transmittance estimation,and the fused dark channel transmittance is obtained using gradient domain guided filtering.Next,the luminance transmittance of the sky region and the dark channel fusion transmittance are synthesized to obtain the final transmittance.Finally,the image is restored using the obtained transmittance and the improved atmospheric light value to obtain the defogging result image.Experimental results show that the proposed method is significantly better than the traditional dark channel method.It can effectively suppress edge artifacts and preserve the color features of foggy images.Compared with the traditional methods,the defogged images obtained by the proposed method can achieve better results in both subjective and objective evaluation.