摘要
在雾天条件下,由于大气中悬浮微粒的散射作用,经图像传感器获得的图像会出现能见度低、对比度差等退化现象。传统的暗通道先验算法虽然能得到较好的复原效果,但不具备实时性。针对传统去雾算法的缺点,提出一种暗通道结合小波变换的去雾算法。在小波变换的低频分量中应用暗通道先验原理得到初始传输图,再用形态学滤波修正白色区域的传输图,应用图像分割对天空区域传输图修正,最后用引导滤波对其进行细化,将修正后的传输图代入雾天成像模型得到复原后的图像。实验结果表明,修改后算法去雾时间仅为HE的1/20,复原后图像的可见边集合数与HE相差不大,平均梯度约为HE的1.5倍。
In foggy conditions,due to the scattering effect of suspended atmospheric particles,the image captured by the image sensor will appear degradation of low visibility and poor contrast.Although traditional dark channel prior algorithm can get a good restoration result,but the real-time performance is poor.For the shortcomings of traditional defogging algorithms,we propose a new one which combines the dark channel with wavelet transform.First we get the initial transmission diagram using dark channel prior algorithm in low-frequency component of the wavelet transform.Then we modify the white area of the diagram using morphological filtering and correct the sky area of the diagram using image segmentation.Finally we refine it with the guided filtering,and substitute the amended transmission diagram into the fog imaging model to obtain the restored image.Experimental results demonstrate that the defogging time of the modified algorithm is only 1 /20 of that in HE’s algorithm.The number of visible edges sets in the restored image has slight difference to that of HE’s while the average gradient is 1.5 time higher.
出处
《计算机应用与软件》
CSCD
2015年第10期192-195,共4页
Computer Applications and Software
基金
中央高校基本科研业务费专项资金项目(2012202020204)
湖北省自然科学基金项目(2011CDB272)
关键词
图像去雾
小波变换
形态滤波
引导滤波
图像分割
Image defogging
Wavelet transform
Morphological filtering
Guided filtering
Image segmentation
