A multispectral image is one in which each pixel contains information about the spectral reflectance of the scene with more channels of responses than the normal trichromatic system. By this definition, four-color and high-fidelity color printing are forms of multispectral imaging. The number of channels varies from four to several hundreds of bands for hyperspectral images. Because of its rich information content, multispectral imaging has a very wide range of applications, such as art-object analysis and archive, astronomy, camera and television development, computer graphics, color copying, color printing, medicine, and remote sensing. It is one of the most actively researched areas in computational color technology.
A system commonly used to acquire multispectral color images consists of a monochrome digital camera coupled with a set of color filters and an image-processing module for spectrum reconstruction as shown in Fig. 14.1. By using a single sensor array that has high spectral sensitivity across a wide range of electromagnetic wavelengths from infrared (IR) through visible light to ultraviolet (UV), multichannel information is obtained by taking multiple exposures of a scene, one for each filter in the filter wheel. Multispectral color imaging is an expansion of the conventional trichromatic system with the number of channels (or bands), provided by different filters, greater than three.
Another way of obtaining multichannel information is by using a conventional trichromatic camera (or a scanner) as shown in Fig. 14.2, coupled with a color filter by placing the filter between the object and camera (or scanner) to bias signals and artificially increase the number of channels by a multiple of three. The filter is not an integral part of the camera or scanner. This method has been used by Farrell and coworkers for converting a color scanner into a colorimeter. They introduced a color filter between the object and scanner, then made an extra scan to obtain a set of biased trichromatic signals. This effectively increases the number of color signals (or channels) by a factor of three. With the proper choice of color filter, they are able to achieve six-channel outputs from a three-sensor scanner. This approach has been applied to camera imaging. For example, Imai and colleagues have used the Kodak Wrattern filter 38 (light blue) in front of the IBM PROâ3000 digital camera to give six-channel signals with two exposures, with and without the filter.
Online access to SPIE eBooks is limited to subscribing institutions.