Suppression of Noise Effects in Color Correction by Spectral Sensitivities of Image Sensors

The accurate transformation of device responses to tristimulus values is important at an early stage in color management and this transformation is usually called color correction. In this paper it is shown that the influence of noise on color correction can be suppressed if the square of the singular values by the singular value decomposition (SVD) of a sensor matrix represented by surface reflectance, SLVA^1/2, is larger than the noise variance, where S is the M × N matrix of M channel spectral sensitivities, L is the N × N diagonal matrix for an recording illuminant, V is the N × N matrix which is composed of eigenvectors of an autocorrelation matrix for spectral reflectance of samples, A is the N × N diagonal matrix with eigenvalues of the autocorrelation matrix along the diagonal and N is the number of the dimension, respectively. A method to suppress the influence of noise on color correction by spectral sensitivities of sensors is presented for the first time.