Spectral measurements of two-dimensional color images

In this work we propose a prototype of the spectral vision system, which can be used Ca measure a color spectrum and two-dimensional spectral images. We first designed a low-dimensional broad band color filter set with a constraint of positive spectral values by the unsupervised neural network. Then we constructed a compact size optical setup for the spectral synthesizer, which can be used to synthesize the light corresponding to the spectral characteristics of the color filter. In the optical setup we implemented the color filters by the use of the liquid crystal spatial light modulator (LCSLM). In our experiments we illuminated a sample of a real world scene by the synthesized lights and detected the intensity images of the filtering process by the CCD-camera. The intensity images correspond to the optically calculated inner products between the color filters and a sample. The data obtained from the filtering process is only a few monochrome images and therefore convenient for storing and transmitting spectral images. From the detected inner products we reconstructed the sample's color spectra by the use of inverse matrix. We present experimental results of measuring a single color spectrum and two-dimensional spectral images.