Search

A model was built to predict chromatic noise perception in digital photography. The model includes the orthogonal opponent color space Y C1C2, and sets of optimized contrast sensitivity functions. Past research on opponent colors, contrast sensitivity functions at threshold and suprathreshold matching has been reviewed. Some historical opponent color spaces have been investigated in terms of transformation and application. Three psychophysical experiments were performed to build the model. Through the equi-luminance plane experiment three equi-luminance planes corresponding to three luminance levels were determined. Method of adjustment was applied for subjects to adjust chromatic noise image until it is least perceptible. Based on results from the first experiment, the orthogonal opponent color space Y C1C2 was developed. The Y C1C2 space shows optimal performance compared with OPP space, with the respect of separating luminance information from chromatic channels and vice versa. The threshold experiment measured contrast threshold for the three cardinal axes and the two diagonal axes of the new opponent color space with three frequency bands and three luminance levels. The QUEST procedure was applied for observers to choose which one of the two side-by-side-displayed stimuli has noise. The supra-threshold experiment was to measure contrast sensitivity above threshold. Method of adjustment was used for observers to adjust the noise contrast of the test stimuli to match the contrast of the parallel displayed achromatic anchor stimuli, which had three times threshold contrast. Sets of optimized CSFs were obtained by empirical modeling on experiment data from the threshold and supra-threshold experiments. The five-parameter band-pass CSF was fitted to model achromatic noise. As to chromatic noise, the six-parameter low-pass CSF was optimized to model chromatic noise. The fact that threshold CSFs and suprathreshold CSFs have similar shape suggests one set of CSFs may be applicable for both cases.