Multispectral Image Capture

Project Leader:
Dr. Jon Y. Hardeberg
Other grant holders:
Dr. Ivar Farup,
Mr Peter Nussbaum,
Principal Investigators:
Dr. David Connah,
Dr Ali Alsam.

Project outline

Traditional colour cameras simultaneously capture three images of a scene, one through a red filter, a second through a green filter and a third through a blue filter. These three images comprise the three channels of a single colour image.

While three channel cameras can give pleasing images, the colour at each pixel is dependent upon both the illumination in the scene and the characteristics of the particular camera. This leads to the problem of metamerism, whereby two surfaces that appear identical to a human observer in a given scene will appear different to any given camera; similarly, two surfaces that give the same camera response may appear different to a human.

Metamerism prevents the reproduction of real world scenes such that their appearance is unchanged. This is clearly true, irrespective of the quality of the image display device.

Multispectral image capture aims to circumvent the problem of metamerism by recovering scene properties that are invariant to illumination and device changes, namely the spectral reflectance of the surfaces in the scene. This can be achieved by taking more than just three independent images of each scene, i.e. by sampling the colour spectrum at more than just three intervals.

At Gjøvik University College we have acquired the hardware for capturing such images. We have a high quality cooled CCD camera (the Alta U32 sold by Apogee) and a tunable colour filter (sold by CRI). The filter design is such that the spectral bandpass of the filter can be changed, thus allowing a potentially infinite number of colour channels. The goal of the multispectral image capture project is to use this device to capture accurate colour images.

The project includes both practical and research components. Mr Atheer Wahham, a student from Linkjoping University, wrote software to simultaneously control the camera and filter, and hence fully automate the image capture process. We are also actively researching several different issues with relation to the overall goal of the project. In particular we have published research on methods for recovering spectral reflectance values from images, deciding how many sensors to use in such a system, methods for choosing optimal filter characteristics and optimal calibration charts.

This project is funded by the The Research Council of Norway as part of the “Strategic University College Programme”, who also fund the sister project “Multispectral colour reproduction”.

Selected publications:

  • Jon Y. Hardeberg. Filter selection for multispectral color image acquisition. Journal of Imaging Science and Technology , 48(2), 2004.
  • David Connah, Jon Y. Hardeberg, and Stephen Westland. Comparison of spectral reconstruction methods for multispectral imaging. In Proceedings of the IEEE International Conference on Image Processing , Singapore, October 2004.
  • David Connah, Ali Alsam, and Jon Y. Hardeberg. Multispectral imaging: How many sensors do we need? Proceedings IS&T/SID Twelfth Color Imaging Conference, Scottsdale, Arizona, pp. 55-58, 2004.
  • David Connah, Jon. Y. Hardeberg. Spectral recovery using polynomial models. Proceedings of SPIE -- Volume 5667, Color Imaging X: Processing, Hardcopy, and Applications, Reiner Eschbach, Gabriel G. Marcu, Editors, pp. 65-75 , 2005.
  • Ali Alsam and Jon Y. Hardeberg. Convex reduction of calibration charts. Proceedings of SPIE -- Volume 5667, Color Imaging X: Processing, Hardcopy, and Applications, Reiner Eschbach, Gabriel G. Marcu, Editors, pp. 38-46 , 2005.
  • Ali Alsam and David Connah. Recovering natural reflectances with convexity. Proceedings of 10th Congress of the International Colour Association, Grenada, Spain .
News archive