Global Image Dehazing via Frequency Perception Filtering
JOURNAL OF CIRCUITS SYSTEMS AND COMPUTERS(2019)
摘要
In outdoor scenes, haze limits the visibility of images, and degrades people's judgement of the objects. In this paper, based on an assumption of human visual perception in frequency domain, a novel image haze removal filtering is proposed. Combining this assumption with the theory of frequency domain filtering, we first estimate the cut-off frequency to divide the frequency domain of the hazy image into three components - low-frequency domain, intermediate-frequency domain and high-frequency domain. Then, by introducing the weighting factors, the three components are recombined together. After the theoretical deduction of frequency domain, the establishment of the actual model and adjusting the cut-off frequency and weighting factors, we finally acquire a global and adaptive filtering. This filtering can restore the details and the contours of the images, which have less noise, and improve the visibility of the objects in hazy images. Moreover, our method is simple in structure and strongly applicable, and rarely affected by parameters. Our algorithm is stable and performs well in heavy fog and the scene changes.
更多查看译文
关键词
Haze removal filtering,human visual perception,frequency domain,image visibility
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要