基于天空区域分割和条件约束优化的图像去雾研究

Research on image defogging based on sky region segmentation and conditional constraint optimization

针对雾天环境中成像设备拍摄的景物容易出现图像模糊、目标丢失、成像质量差等而引起图像识别率下降的问题,提出一种基于天空区域分割和条件约束优化的图像去雾方法,解决使用暗原色先验理论技术进行图像去雾而导致的天空区域出现晕圈和失真的现象。对图像进行预处理后,将图像划分为天空区域和非天空区域,用灰度直方图估计环境光,利用已知条件对非天空区域的透射率进行合理约束,通过自适应增加天空区域透射率的方法,避免天空区域出现严重的颜色失真现象,合并之后对这两部分的透射率进行滤波细化。结果表明:对于含有天空区域的雾天图像,使用本文算法去雾后的图像轮廓和细节上都有较大改进,能够达到去雾目的。这对于提高成像设备在雾天环境中的工作效率具有重要意义。

In view of the phenomenon of blurry scene,target loss and poor quality by imaging equipment in foggy environment leading to the decrease of image recognition rate,an image defogging method was proposed based on sky region segmentation and conditional constraint optimization to solve the problems of halo and distortion in the sky region caused by dark channel prior.After preprocessing the image,the image was divided into sky region and non-sky region.The gray histogram was used to estimate the ambient light.The transmittance of the non-sky region was constrained by using known conditions.The method could adaptively increase the transmittance of the sky region,which avoided serious color distortion in the sky region.The transmittance of the two parts were filtered and refined after merging.The experimental results showed that the outlines and details of fog images with sky area were improved greatly after using this algorithm,capable of satisfying the purpose of defogging well.This was of great significance to improve the working efficiency of imaging equipment in foggy environment.