Abstract: Since the dark channel prior algorithm has dull restoration effect and too long processing time, a fast dehazing algorithm for single image based on relative transmittance estimation was proposed. On the basis of the analysis of relationship between field depth under haze condition and minimum image of color channel (RGB) images, a preliminary transmittance was estimated through the relative amount of field depth, and then it was adjusted with an improved mean filter. At last, the clear image could be recovered by the atmospheric scattering model and the brightness was enhanced to improve its visual effects. The estimation of transmittance in this paper is simple and effective, the restored images are clear and natural, and have high detail visibility and scenery layering. The experimental results show that the proposed algorithm has great improvement in image dehazing quality and computational time, which is propitious to achieve real-time application.

Key words: image dehazing, dark channel prior, transmittance, image restoration, image enhancement

摘要： 针对暗通道先验算法中恢复效果偏暗以及运算时间过久的问题,提出一种基于相对透射率估计的单幅图像快速去雾算法。该算法在分析雾霾条件下场景深度与最小值图像关系的基础上,依据景深相对量初步估计透射率,利用改进的均值滤波器作精确化调整,最后根据大气散射模型复原清晰图像,并通过亮度增强改善其视觉效果。该算法对透射率的估计简单、有效,复原图像清晰、自然,并且具有较高的细节可见度和层次感。实验结果表明,该算法在去雾效果和处理速度方面均有很大改善,有利于实现实时性应用。

关键词: 图像去雾, 暗原色先验, 透射率, 图像复原, 图像增强