Abstract | ||
---|---|---|
To explore the possibility of using hyperspectral methods for handling conventional color images(CIs) or panchromatic images (PIs), some researchers propose to simulate HSI by using PI, based on stacking local pixels. Since the simulated data possesses the similar characteristics as HSI, vertex component analysis is then adopted to factorize the simulated data. Since the method only exploits the pixel value, it may face some problems if the gray value of targets varies. To avoid the problem, an improved approaches is proposed to reduce the influence of gray value. Instead of directly stacking the original pixels, the proposed method first extracts five different features of original images, then stack the different fundamental features and thus forming feature data. Such data provides more information than original RGB image, besides, it could provide more robust information than original methods since the application of image features. In the experiments, two different methods will be used to detect clouds, and results show the efficacy of proposed methods. |
Year | DOI | Venue |
---|---|---|
2016 | 10.1145/3028842.3028887 | ICIIP |
Keywords | Field | DocType |
Panchromatic image processing, hyperspectral image simulating, hyperspectral unmixing, target detection | Computer vision,Pattern recognition,Feature (computer vision),Computer science,Panchromatic film,Rgb image,Hyperspectral imaging,Factorization,Pixel,Artificial intelligence,Feature data,Stacking | Conference |
Citations | PageRank | References |
0 | 0.34 | 6 |
Authors | ||
4 |