基于引导滤波的单幅图像去雾算法研究

Research on Single Image Dehazing Using Guided Filtering

基于暗原色先验的单幅图像去雾算法较好地解决了估算雾天场景透射率的问题,算法复原的图像清晰颜色自然。但原始算法计算复杂度高、耗时长、无法满足实时性需求。介绍了引导图像滤波的原理,并将其与暗原色先验结合以改进原始算法;同时分析了滤波参数对去雾结果的影响。实验证明改进的算法在保证原算法去雾效果的同时能使算法复杂度大大降低。

The algorithm of single image haze removal using dark channel prior is achieved the dynamic estima- tion of the transmission in foggy scenarios successfully, and the restored image is a natural visibility. But the origi- nal algorithm has a complex computation that can＇ t satisfy the real-time processing. The guided image filtering is elaborated in detail and combine it with dark channel prior to improve the original method. Then an analysis of how the parameters of the filter influence od generates satisfactory results with the dehazing result is given. Experiments and comparisons show that this meth- low computation complexity.