A Psychophysical Experiment Evaluating the Color and Spatial Image Quality of Several Multispectral Image Capture Techniques

©2004, IS&T—The Society for Imaging Science and Technology are evaluated in which the spatially averaged estimated spectra using imaging are compared with spectra measured using small-aperture refl ection spectrophotometers. Spectra are compared using colorimetric, spectral, and metameric criteria.5 As image data, there is an obvious requirement for viewing the spectral data in a rendered form such as display or print. When rendered, the many-channel data are mapped to three signals. That is, the high dimensional spectral data are projected onto the human visual system subspace (or an appropriate linear transformation). Similar to the conversion from spectral refl ectance to tristimulus values using a refl ection spectrophotometer, the accuracy can be quite high. Intuitively, as the number of channels increases, the spectral accuracy and in turn, the colorimetric accuracy is expected to increase. Unlike spectrophotometers, imaging systems are often subject to considerably greater noise. As noise is propagated through an imaging system,6 the rendered image may have poor spatial image quality and be quite objectionable. With the potential large number of channels and complex signal processing, spectral imaging seems susceptible to image noise.7 It is quite possible that the noise propagation during projection can result in rendered images with poor spatial image quality. In addition, as the number of channels increase, there is an Introduction