修正大气耗散函数的单幅图像去雾

发布时间：2018-03-13 13:15

  本文选题：图像去雾　切入点：暗原色先验　出处：《中国图象图形学报》2017年06期 　论文类型：期刊论文

【摘要】：目的针对暗原色先验原理对雾霾图像中天空或白色物体等明亮区域透射率估计不足,导致该区域去雾后彩色失真的问题,提出一种基于暗原色先验和引导滤波修正大气耗散函数的单幅图像去雾算法。方法首先,基于暗原色先验模型得到大气耗散函数的粗估计值;其次,构造一个修正函数,纠正暗先验失效的明亮区域的大气耗散函数;然后,对修正后的大气耗散函数和求得的初始传输图分别利用引导滤波进行优化,平滑图像边缘的同时保持图像细节信息;最后,由优化后的传输图和估计的大气光值得到复原图像。结果选取多幅经典图像进行对比实验,并利用峰值信噪比和均方误差衡量去雾结果的失真程度。实验结果表明,本文算法不但在非明亮区域可以得到较好的去雾效果,而且也能使图像中的明亮区域保持原有色彩,相比而言本文算法得到的复原图像整体失真较少;对于大小为460×300像素的图像,本文算法与He方法相比,得到的复原图像峰值信噪比提高了0.600 5 d B,均方误差降低了0.002 6,耗时缩短了29.622 0 s。结论对于雾天包含明亮区域的降质图像,提出了一种修正大气耗散函数的单幅图像去雾算法。实验结果的主观和客观评价表明本文算法对天空或白色物体等明亮区域能得到较好的去雾效果,有效改善了暗原色先验原理对图像中明亮区域造成的彩色失真问题。
[Abstract]:Aim to solve the problem of color distortion caused by dark priori principle in haze image due to insufficient estimation of the transmittance of bright regions such as sky or white objects. This paper presents a single image de-fogging algorithm based on a dark priori and a guided filter to modify the atmospheric dissipation function. Firstly, based on the dark priori model, the coarse estimation of the atmospheric dissipation function is obtained; secondly, a correction function is constructed. The atmospheric dissipation function of the bright region with dark prior failure is corrected. Then, the modified atmospheric dissipation function and the obtained initial transmission diagram are optimized by guided filtering, respectively, to smooth the image edges while keeping the image details. From the optimized transmission map and the estimated atmospheric light, the reconstructed image is worth recovering. Results A number of classical images are selected for comparison experiments, and the distortion degree of the de-fogging result is measured by using the peak signal-to-noise ratio (PSNR) and mean square error. The experimental results show that, This algorithm can not only get better effect in the non-bright region, but also keep the original color in the bright region of the image, compared with the whole distortion of the restored image obtained by the algorithm in this paper. For an image of 460 脳 300 pixels, the peak signal-to-noise ratio (PSNR) of the reconstructed image is increased by 0.600 5 dB, the mean square error is reduced by 0.002 6, and the time is shortened by 29.6220 s compared with that of he method. Conclusion for the degraded image with bright region in fog, the mean square error is reduced by 0.002 6, and the mean square error is reduced by 29.6220 s. A single image de-fogging algorithm with modified atmospheric dissipation function is proposed. The subjective and objective evaluation of the experimental results show that the proposed algorithm can achieve better defog effect for bright regions such as sky or white objects. The color distortion caused by the priori principle of dark primary color in the bright region of the image is effectively improved.
【作者单位】： 西北大学信息科学与技术学院;洛阳师范学院中原经济区智慧旅游河南省协同创新中心;
【基金】：国家自然科学基金项目(61502219) 中国博士后科学基金项目(2015M582697) 国家科技支撑计划基金项目(2013BAH49F02)~~
【分类号】：TP391.41
，

本文编号：1606540