基于稀疏特征提取的单幅图像去雾  

Single Image Dehazing Based on Sparse Feature Extraction

作　　者：刘坤 毕笃彦[1] 王世平 何林远[1] 高山[1] Liu Kun, Bi Duyan, Wang Shiping, He Linyuan, Gao Shan(Aeronautics and Astronautics Engineering College, Air Force Engineering University Xi＇an, Shaanxi 710038, Chin)

机构地区：[1]空军工程大学航空航天工程学院,陕西西安710038

出　　处：《光学学报》2018年第3期291-299,共9页Acta Optica Sinica

基　　金：国家自然科学基金(61372167;61701524;61773397)

摘　　要：为解决暗通道先验去雾算法在天空区域和大片白色区域色彩失真的问题,提出了一种基于稀疏表示模型和特征提取的单幅图像去雾算法。通过稀疏字典的训练过程,学习雾天图像的稀疏特征,初步优化粗略介质传输图的稀疏系数。根据雾天灰度图像的稀疏特征,进一步精细化介质传输图。逆向求解雾天退化模型,得到去雾图像。实验结果表明,所提算法在天空区域的处理上优势明显,同时恢复出更多的图像细节和边缘信息。To overcome the color distortion in sky regions and large white regions brought by the dark channel prior dehazing algorithm, we propose a single image dehazing algorithm based on sparse representation model and feature extraction. Firstly, the algorithm learns the sparse features of foggy images via training sparse dictionary, and optimizes the sparse coefficients of the rough medium transmission image preliminarily. Then, the algorithm refines the medium transmission image by the sparse features of foggy gray images. Finally, with the converse solution of the degradation model, the algorithm obtains the dehazing image. The experimental results demonstrate that the proposed algorithm has obvious advantages in the processing of the sky area, and at the same time, it can recover more image details and marginal information.

关 键 词：图像处理 图像增强 图像去雾 稀疏表示 字典学习 特征提取 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]