Abstract

The spectral curves of LGN cells from a famous study¹ were subjected to principal component analysis to determine the minimal set of basis vectors needed to describe the data. Over 92% of the variability in the 441 spectral curves (147 cells measured at three intensity levels) could be described by a unique orthonormal set of three basis vectors. Every cell’s spectral curve could be represented as a linear combination of these vectors, with minimal loss of information. An orthogonal rotation of the vectors, which preserves all Euclidean chromatic distances in the color space spanned by the vectors, led to a reasonable correspondence to psychophysical chromatic basis vectors.² Possible cone input types were suggested by projecting the components and raw spectral curves into tricone space using multiple regression techniques.

© 1986 Optical Society of America