基于暗原色优化算法的去雾研究

Research on Fog Removal Based on Dark Primary Color Optimization Algorithms

针对基于传统的暗原色先验去雾算法中,由于某些场景下的雾天图像存在大面积明亮区域无法满足暗原色先验的假设,导致去雾效果不佳。文中就此问题提出了一种改进的去雾算法,基于McCartnet的理论建立大气散射模型,根据暗通道理论粗略估计透射率,之后引入容差参数并设置阈值,重新计算明亮区域的透射率,从而实现对明亮区域透射率的自校正。针对于复原图像色彩较暗的问题,采用改进的线性亮度调整方法来调节图像的亮度。实验结果显示,相较于原算法而言,改进算法可以有效的对大气光值进行估计,降低明亮区域的色彩失真,复原的图像可以保持足够的亮度,同时不丢失图像的细节,视觉效果显著提高。

In the traditional dark-primary prior dehazing algorithm,the fogging effect in some scenes could not meet the assumption of dark primary color a priori,which leaded to poor dehazing;effect.In this paper,an improved dehazing algorithm was proposed based on the above results.Firstly,the atmospheric scattering model was established according to McCartnet’s theory.Then the transmission map was roughly estimated according to the dark channel prior theory.The tolerance parameter was introduced and the threshold was set to recalculate the transmission map of the high light regions.To achieve self-adapted function of transmission map in high light regions.In order to restore the dark color of the image,an improved linear adjustment method was applied to adjust the brightness of the image.The experimental results showed that compared with the original algorithm,the improved algorithm could effectively estimate the atmospheric light value and reduced the color distortion of the high light regions.The restored image could maintain sufficient brightness without losing the details of the image,and the visual effect was remarkable improved.