Color in Context: from perception to image enhancement

For a long time color has been analyzed as a physical property of light and/or objects, without taking into account the characteristics of the human vision system (HVS).
The HVS, during the vision process, continuously performs various adaptation mechanisms, among them color and lightness adaptation. These adaptation mechanisms are responsible of the chromatic appearance of an object, which strongly depends on the scene characteristics, such as the light sources and the chromaticity and the mutual positions of the objects. This fact affects the analysis and the reproduction of digital images, acquired or made by synthesis.
Strictly related with lightness adaptation, is the "tone reproduction" problem, whose purpose is to give a robust and effective solution to a correct reproduction of high range lightness and color gamuts on limited range devices such as a monitor or an ink jet printer. In fact, the HVS operates at very different luminance levels (from 10 exp -6 to 10 exp 8 lux), while a display can reproduce only a restricted range (10 exp 2).
These and other problems have originated a new research field called "Color appearance" which scientific paradigm emphasizes the importance of visual appearance, extending the classical approach which considers only visual stimuli and light sources photometric and colorimetric properties, leaving in the background the perceptual properties of the HVS. In fact, pursuing color enhancement for commercial applications, several researchers investigate HVS computational methods to implement automatic color correction algorithms, with the purpose to produce pleasant images rather than correct image by a colorimetric point of view. It is clear that the notion of pleasantness, even if not quantifiable, is strictly correlated with principles of visual appearance.
A famous model of the HVS color perception is Retinex, due to Land and McCann. This theory is 40 years old, but has a renewed interest. It has been studied by many researchers and several different algorithm have been developed.
Starting from the Retinex approach, but with a completely different computational model, an alternative algorithm has been recently developed by our group. It is called ACE, for Automatic Color Equalization.
It is not a complete model of the HVS, since it performs only some of its mechanisms. Like the HVS, ACE is able to adapt to varying lighting conditions, and to extract visual information from the environment efficaciously. Its computational approach merges the "Gray World" and "White Patch" global equalization mechanisms, and adds local adaptation effects, by taking into account the spatial distribution of color information. ACE has shown promising results in solving the color constancy problem and performing an effective image dynamic data driven stretching. The promising results and its characteristics suggest other types of application.