Illumination-invariant retrieval of multicolored object using sampled spectrum image
✍ Scribed by Kousuke Moriwaki; Kotaro Nakatani
- Publisher
- John Wiley and Sons
- Year
- 2000
- Tongue
- English
- Weight
- 733 KB
- Volume
- 31
- Category
- Article
- ISSN
- 0882-1666
No coin nor oath required. For personal study only.
✦ Synopsis
A method of detecting multicolored objects with illumination invariance is described. We use a new image data format, a sampled spectrum image, which samples values of spectrum of incident light at some wavelengths. To capture the image, we constructed a five-channel imaging system with a supersensitive CCD camera that had interference filters attached in front of the lens. Each interference filter has delta-function-like narrow spectral transmissivity. The ratio of corresponding pixel elements of two sampled spectrum images, which are measured under different illuminations, is independent of the spectral reflectivity of the target colors. The ratio depends only on the spectral distribution of the illumination. Our proposed algorithm can simultaneously detect the multicolored target object and estimate the change ratio of spectrum of illumination at the sampling wavelengths.