Dataset Description
This dataset contains 59-band visible-NIR hyperspectral images from 420nm to 1000nm at 10nm intervals. The images were captured using a monochrome camera and two VariSpec tunable filters, VIS for 420-650nm and SNIR for 650-1000nm, for capturing each hyperspectral image. The hyperspectral image was then converted into a form of spectral reflectance by a calibration process using a colorchecker whose spectral reflectance property is known. The dataset is available only for research purpose and you may not redistribute the dataset.
Features
- The dataset contains 59-band hyperspectral images, including both visible and near-ifrared (NIR) domains, of colorful objects with rich textures, and therefore, is useful for the evaluation of various RGB-NIR/hyperspectral imaging and image restoration tasks.
- The dataset is caliblated using a colorchart and saved in the spectral reflectance form. Therefore, users can simulate RGB-NIR/hyperspectral imaging pipeline with arbitrary camera sensitivity and illumination. Please see the sample code to do so.
Citation
Please cite the following paper if you use the dataset.
- Single-Sensor RGB-NIR Imaging: High-Quality System Design and Prototype Implementation [Project]
Yusuke Monno, Hayato Teranaka, Kazunori Yoshizaki, Masayuki Tanaka and Masatoshi Okutomi
IEEE Sensors Journal, vol.19, no.2, pp.497-507, January, 2019.
Downloads
We used 40 scenes in our IEEE Sensors Journal paper. However, among the 40 scenes, only 16 scenes are publicly available for research purpose and the other scenes are not available due to license issues.
[All scenes (1.70GB)], [Part1 (432MB)], [Part2 (439MB)], [Part3 (436MB)], [Part4 (440MB)]
 Butterfly [mat] |
 Butterfly2 [mat] |
 Chart [mat] |
 Chart2 [mat] |
 Chart3 [mat] |
 Cloth [mat] |
 Cloth2 [mat] |
 Cloth3 [mat] |
 Doll [mat] |
 Doll2 [mat] |
 Fan [mat] |
 Fan2 [mat] |
 Fan3 [mat] |
 Origami [mat] |
 Paint [mat] |
 Spray [mat] |
Spectral Reflectance Data
The follwoing data provides spectral reflectance data of X-Rite Classic and SG ColorCheckers in 400-1000nm wavelengths, which are measured by a StellarNet BLUE-Wave miniature spectrometer. [Download]
 ColorChecker Classic |
 Spectral Reflectance |
 ColorChecker SG |
 Spectral Reflectance |
Sample Code
This is the sample code of basic implementation of the RGB-NIR imaging pipeliene. The code first converts the 59-band hyperspectral data to a camera RGB-NIR image, given camera spectral sensitivity. Then, the camera RGB-NIR image is transformed to the target color space by color correction using spectral reflectance data of the SG chart. You can modify the code to simulate the image under arbitrary illumination and camera sensitivity, depending on your purpose.
See Also
TokyoTech 5-band Multispectral Image Dataset
TokyoTech 5-band Multispectral RAW Dataset
TokyoTech 31-band Hyperspectral Image Dataset
Contact Information
If you have any questions, please feel free to ask Yusuke Monno (ymonno(at)ok.sc.e.titech.ac.jp).
     |