Time-Frequency Analysis of Color Spectra

Most physics-based color processing methods define color signals as functions of wavelengths. Such a description characterizes the color signal completely from a physical point of view. This representation has however several drawbacks: the description is redundant and important properties are often hidden and difficult to extract from this representation. Other color representations may provide more compact descriptions or descriptions in which the relevant properties are more clearly visible. Other color systems have thus been used, including popular color signal representations based on principal component analysis or Fourier transformations. Most of these methods decompose color signals as linear combinations in a given basis system and they require thus the selection of a basis before they can be applied. In this paper we introduce time-frequency methods as an alternative that avoids the selection of a basis and provides a signal representation that combines advantages of both, the wavelength and the Fourier transform based signal description.