摘要
针对已有单幅雾天图像复原算法存在的局部方向性结构信息描述不充分,易导致复原后景物局部细节模糊或丢失的不足,提出一种基于方向延伸专家场的单幅雾天图像复原算法.首先基于大气散射模型获得粗略的大气光传输图估计,并在此基础上建立方向延伸专家场模型对大气光传输图进行优化;为避免在图像景物高亮度区域出现失真现象,利用无黑点约束算法对大气光传输图进行约束及调整,再根据图像中景物的梯度先验信息获取位于无穷远区域的像素集,由此估计出大气光强值;最后根据大气散射模型反向求解,得到复原后的场景图像.实验结果表明,该算法不仅可以显著地改善景物的细节信息、提高图像清晰度,并且恢复后的景物颜色更加自然、真实.
Traditional single foggy image restoration algorithms cannot sufficiently describe the local directional structure information of the scene, which tends to make partial details of the scene be vague or missing after the recovery. To overcome this drawback, a new fog-removing algorithm based on orientation extended Fields of Experts was proposed in this paper. Firstly, the atmosphere transmission map was estimated cursorily according to the atmospheric scattering model, and it was further refined by the orientation extended Fields of Experts. To avoid the distortion at highlight areas, the atmosphere transmission map was restricted and adjusted by the no black pixel constraint algorithm. Secondly, the grads apriority rule was used to obtain the set of pixels that are located in the infinite region, so as to estimate the intensity of atmospheric light. Finally, by reversely solving the atmospheric scattering model, the defogged image was obtained. Experimental results show that the proposed algorithm can not only remarkably improve the detailed information of scene, and enhance the degree of clearness of foggy images, but also make the color appearances much more natural and realistic.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2014年第5期782-787,共6页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金(61379105)
安徽省自然科学基金(1208085QF115)
关键词
图像复原
专家场模型
去雾
梯度先验
image restoration
fields of experts model
defogging
grads apriority rule
