Obtaining scene depth and the sky brightness value is the core issue for fog image restoration method . However ,the present haze removal algorithms are all based on certain assumptions to obtain these two parameters and are not effective enough for processing drab and large sky area inland images .By studying the inland waterway video image , this paper combines the histogram multimodal averaging and bit plane decomposition method to get the sky brightness val -ue to split the sky area .Then ,by using the separate area scene depth calculation method ,the image scene depth of any point is obtained .Finally ,according to the atmospheric scattering model ,the haze removal of single inland image is com-pleted .In order to evaluate the proposed methods ,the paper compares the haze removal effectiveness from three aspects including the visible image edge number ratio ,average gradient ratio and image entropy value .Experiments show that the proposed algorithm is effective for the fog inland image .%图像景深和天空亮度值的求取是图像复原方法去雾的核心问题，但目前的去雾算法都是基于一定的假设条件来求取这2个值的，对于色彩单调、天空区域较大的内河图像去雾效果不理想。通过对内河航道视频图像的研究，提出将直方图多峰均值法和位平面分解法相结合的方式来求得天空亮度值并分割天空区域；采用分区域的景深计算方法求得图像任意一点的景深值。然后基于大气散射模型，完成内河雾天单幅图像的去雾处理。为了客观评价去雾后图像的质量，从图像的可见边数目比、平均梯度比和图像熵值三方面进行了去雾效果的比较。实验证明，该算法对内河航道图像有良好的去雾效果。
展开▼
