Single image dehazing (SID) solves the atmospheric scattering model (ATSM). The ill-defined nature of the SID makes it a challenging problem. The transmission is the prime parameter of ATSM. Hence, accurate transmission is essential for quality of SID. The existing methods of SID estimate the transmission based on priors with strong assumptions (such as dark channel prior). These methods do not recover original colors, structure and visibility due to wrong transmission under invalidity of these assumptions. Therefor, the difference channel (DCH) is proposed to estimate accurate transmission. The DCH non-linearly translates the minimum channel of hazy image into minimum channel of haze-free image, which is used to compute the value of transmission. The DCH is based on an observation that difference of maximum and minimum color channel of the hazy image is negatively correlated with depth. The proposed method is able to recover the details from hazy image in the form of structure, edges, corners, colors and visibility due to the DCH. The accuracy and robustness of the proposed method is proved by comparing the results with known dehazing methods based on qualitative and quantitative analysis using benchmark data sets.

单幅图像去雾 (SID) 解决了大气散射模型 (ATSM)。SID 不明确的性质使其成为一个具有挑战性的问题。传输是 ATSM 的主要参数。因此，准确的传输对于 SID 的质量至关重要。SID 的现有方法基于具有强假设的先验（例如暗通道先验）来估计传输。由于在这些假设无效的情况下错误传输，这些方法不能恢复原始颜色、结构和可见性。因此，提出差分信道（DCH）来估计准确传输。DCH 将模糊图像的最小通道非线性地转换为无雾图像的最小通道，用于计算透射值。DCH 是基于这样的观察，即模糊图像的最大和最小颜色通道的差异与深度呈负相关。由于DCH，所提出的方法能够以结构、边缘、角落、颜色和可见性的形式从模糊图像中恢复细节。通过使用基准数据集基于定性和定量分析将结果与已知的去雾方法进行比较，证明了所提出方法的准确性和稳健性。