一种联合多正则化项的双域去雾模型

在有雾天气下,图像传感器获得的图像可能会出现能见度低、对比度差和其他退化现象。针对暗通道先验算法在景深较大区域可能导致失真的局限性,提出一种联合小波,TV以及轮廓波正则化的模型,该模型旨在提高场景传输率和图像质量。首先,该文通过结合小波,TV和轮廓波惩罚来增强暗通道得到的初始场景传输率,并对此应用高提升滤波来进一步增强边缘与细节,从而得到精细的场景传输率;接下来,使用这种精细后的传输率来获得初步的去雾图像;最后,对得到的这幅图像进行后处理,再次作用小波,TV以及轮廓波惩罚与高提升滤波,在保证平滑的同时保持局部细节。实验结果表明,该算法的去雾结果是有效的,并且主观和客观的评价都表明,该方法表现良好。

In conditions of hazy weather,images captured by image sensors may suffer from reduced visibility,low contrast,and other degradation issues.To address the shortcomings of the dark channel prior algorithm,which can lead to distortion in areas with significant depth of field,a new model is introduced.This model combines wavelet,TV,and contourlet regularization techniques with the goal of enhancing both scene transmission and image quality.First,the initial scene transmission of the dark channel is improved by combining wavelet,TV,and contourlet punishments,followed by highboost filtering to enhance edges and details for a clearer scene transmission.Subsequently,a refined transmission is used to generate a preliminary dehazing image.After completing the processing on the obtained image,we further apply wavelet,TV,and contourlet punishments along with highboost filtering to enhance smoothness while preserving local details.Experimental results demonstrate the effectiveness of the algorithm in removing haze from the original image,with both subjective and objective evaluations indicating strong performance.