基于改进暗通道和自适应容差的图像去雾算法

Image defogging algorithm based on improved dark channel and adaptive tolerance

传统暗通道去雾算法计算的透射率图存在块效应,易造成复原图像白边现象,同时图像中天空、白云等明亮区域不适用暗通道原理,易引起去雾图像失真。本文结合引导滤波和自适应容差机制提出了一种基于多尺度暗通道和自适应容差的去雾算法,可有效避免以上问题。首先,计算3种不同尺寸滤波窗口下的透射率初估计,并对估计结果进行有效融合;接着,通过引导滤波对透射率进行细化,以获得鲁棒性和准确性更好的多尺度透射率图;然后,引入自适应容差策略对图像中明亮区域的透射率进行修正;最后,由于暗通道去雾图像整体亮度偏暗,因此对去雾图像的亮度和对比度进行亮度补偿。实验结果表明,采用不同算法对不含和少量天空区域的图像去雾,信息熵约提高0.2 bit/symbol,平均梯度约提高0.5,PSNR约提高8 dB。对较多和大量天空区域图像去雾,PSNR约提高3 dB,SSIM约提高0.1。较好地实现了去雾图像细节清晰、颜色可靠且明亮区域去雾效果良好等要求。

In the traditional dark channel defogging algorithm,there is blocking effect in the transmittance estimation,which is easy to cause the white edge phenomenon. At the same time,the dark channel principle is not applicable to the bright areas such as sky and white clouds in the image,and results in defogging image distortion. In this paper,a defogging algorithm based on multi-scale dark channel and adaptive tolerance is proposed,which can effectively avoid the above problems. Firstly,the rough transmittance estimates under three windows with different sizes are calculated,and the results are fused effectively. Then,the transmittance is refined by guided filtering to obtain a multi-scale transmittance map with better robustness and accuracy,and the adaptive tolerance strategy is introduced to correct the transmittance of the bright areas. Finally,the overall brightness of the defogged results is dark,so the brightness and contrast of the defogged image are compensated. Experimental results indicate that the information entropy increases by 0.2 bit/symbol,the average gradient increases by 0.5,and the PSNR increases by about 8 dB when different algorithms are used to the image without or with a small amount of sky. For processing more and a large number of sky region images,PSNR and SSIM are improved by about 3 dB and 0.1,respectively. The requirements of clear image details,reliable color and good defogging effect in bright areas are better realized.