Name
Abstract | ||
|---|---|---|
Ubiquitous visual surveillance is critical to public security. Unfortunately, adverse weathers, especially haze, degrade visual surveillance quality evidently, so dehazing is commonly used to limit the interference of haze. Unlike traditional dehazing methods that use various patch-based priors, non-local dehazing employs color index and regularization to estimate and refine initial transmission, respectively. However, currently non-local dehazing has not made the most of pixel neighborhood relation, so the edge details cannot be preserved powerfully. Since the gradient represents the difference between the adjacent pixels, the non-local dehazing algorithm is enhanced by color gradient in this paper. The color index and color gradient are jointly clustered to improve the accuracy of initial transmission. Finally the haze is removed according to the transmission refined by guided filter. The experimental results show that the proposed non-local dehazed algorithm enhanced by color gradient can effectively maintain the edge details and improve the performance of dehazing. |
Year | DOI | Venue |
|---|---|---|
2019 | 10.1007/s11042-018-5673-6 | Multimedia Tools and Applications |
Keywords | Field | DocType |
Non-local dehazing, Color gradient, Guided filter | Computer vision,Pattern recognition,Computer science,Color index,Regularization (mathematics),Pixel,Interference (wave propagation),Artificial intelligence,Prior probability,Public security,Haze,Color gradient | Journal |
Volume | Issue | ISSN |
78.0 | 5 | 1573-7721 |
Citations | PageRank | References |
0 | 0.34 | 15 |
Authors | ||
3 | ||