In a previous work Color Res. Appl., 32 (2007, in press) we found that a noise-free RGB camera coupled with color filters provided significantly better recovery of radiance and reflectance in natural scenes than an RGB camera alone. The recovery of spectral radiances and reflectances was derived using a direct pseudo-inverse algorithm instead of a linear-model approach. The advantages of this direct- mapping method are its simplicity and low computational cost in terms of its performance. In this work, we have tested the method with noisy data. The results suggest that the addition of a color filter when using an RGB digital camera affected by high noise level does not improve the spectral and colorimetric quality of spectral recoveries out of the training set. The results also indicate that a preprocessing of color images is needed before direct pseudo-inverse algorithm is implemented to recover adequately spectral data from noisy signals.
reflectance multispectral imaging systems illuminants natural scenes
J. L. Nieves E.M. Valero C. Plata J. Romero S. M. C. Nascimento K. Amano D.H.Foster
Deportamento de Optica, Facultad de Ciencias, Universidad de Granada, Granada,pain Deportamento de Optica, Facultad de Ciencias, Universidad de Granada,Granada, Spain Deportamento de Optica, Facultad de Ciencias,Universidad de Granada, Granada, Spain Deportamento de Optica, Facultad de Ciencias, Universidad de Granada, Granada, Spain Department of Physics, Minho University, Campus de Gualtar, Braga, Portugal Sensing, Imaging, and Signal Processing Group, School of Electrical and Electronic Engineering, Univ