雾化图像的约束进化时频加权滤波去雾算法

Image Defogging Algorithm Based on Constrained Evolutionary Time-frequency Weighted Filtering

传统的基于物理纹理特征的去雾算法不能有效去除雾化图像的边缘雾点,特别在浓雾条件下处理效果不好。提出一种采用约束进化时频加权滤波的雾化图像去雾算法,首先采用分块处理方法,基于3×3模板求解雾化图像的暗原色。以图像暗原色为新的图像特征,设置约束进化条件,对图像雾点进行平滑重排伪魏格纳-维尔-霍夫变换(RSPWVD-Hough)时频分析,提取RSPWVD-Hough时频特征作为加权向量来指导雾化图像的盒子滤波,实现去雾处理,并通过约束进化算法对图像去雾的结果不断进行循环优化。RSPWVD-Hough时频特征能有效反映图像的边缘雾化特征点,故对雾化图像的边缘雾点去雾效果甚好。仿真测试表明,改进的图像去雾算法能在提高处理速度的同时优化图像输出结果,透过率提高明显,在浓雾条件下对图像目标识别等领域有较好的应用价值。

The traditional defogging algorithm based on physical texture feature cannot remove edge fog effectively,and the performance is bad especially in thick fog. An improved image defogging algorithm was proposed based on constrained evolutionary time frequency weighted filtering. The block processing method was taken, and 3× 3 blocks were used to solve the dark color. The dark color was used as the new feature. The constraint evolution condition was set. The smoothing pseudo and RSPWVD-Hough time-frequency analysis were implemented. The RSPWVD-Hough time frequency feature was extracted as the weighting vector to direct image box filter, and the defogging process was realized. The constrained evolutionary algorithm was used to optimize the result. Because the RSPWVD-Hough time-frequency feature can effectively reflect the edge atomization feature points of the image, the edge defogging performance is perfect. Simulation results show that the improved defogging algorithm can optimize the image defogging performance and improve the operating speed. The penetrating rate is improved greatly. It has good application value in image target recognition on thick fog condition.