Exploring the Complexities of Color Spaces and Perceptual Brightness
Unlike the RGB color space, CIELab and its modern variants, such as CIECAM02 and Oklab, aim for perceptual uniformity. This means that changes in numerical values should correspond to perceived changes in color based on human perception studies. The goal is to create equations that align closely with observed data across different axes.
Helmholtz-Kohlrausch Effect Explained
The process inherently involves some error. Ideally, a color with a lightness value of L=50 should appear twice as bright as one with L=25. However, with highly saturated colors like red, this isn’t always the case across these color spaces, illustrating the Helmholtz-Kohlrausch effect. For instance, a panel of colors with the same lightness value may still appear differently, with red often seeming more vivid.
Recent research has modeled this effect and applied an additional transform, building on work by Fairchild MD and Pirrotta E in the 90s. The resulting “Predicted Equivalent Achromatic Lightness” provides the lightness of gray matching perceived lightness, useful for desaturating images accurately. Traditional desaturation methods provide a [0..1] value rendered as an RGB gray, while the L_EAL value gives a precise perceptual gray for desaturated images.
Implications for Image Processing
The significance of this adjustment is evident in practical applications. A tool for desaturating game screenshots revealed that red assets appeared unusually dark even when using CIELab. This discrepancy occurs because CIELab doesn’t account for the Helmholtz-Kohlrausch effect, undervaluing red’s saturation impact on lightness. Without this adjustment, red assets might be mistakenly brightened during design evaluations.
Unfortunately, no perceptually uniform color spaces currently include these transformations in their final output space. The search for such advanced color spaces continues, as their integration could significantly enhance image processing accuracy.