Color image enhancement using both chromatic and luminance components

A vast amount of work has been published regarding grayscale processing of digital images. Although some of this work has been adapted for color images, many of the resulting algorithms neglect the correlation that exists between the individual RGB color components. Consequently, they introduce color artifacts. Attempts have been made to decouple the RGB components through color space transformations that isolate the luminance from the chromatic information. Color image enhancement is then considered a two step process, where the luminance and the chromatic components are processed independently. However, the RGB color space only has a finite number of available colors, which limits the attainable chromatic values for any given luminance level. This recoupling of the chromatic and luminance components constrains the independent processing of these two components. This thesis investigates this coupling and how it effects a number of color image processing algorithms. Specifically, new algorithms for color histogram equalization, automatic white balance and color filtering are presented using the C-Y (color difference) color model. Coupling the chromatic components to the luminance produces improved white balanced results and enhances the ability of histogram equalization to increase saturation contrast. Furthermore, several adaptive filters are implemented using the C-Y color space. Because this color model closely matches the human concept of color, the noise is filtered from color components that closely match the perceptually significant quantities. This improves the results for these adaptive filters, but also requires special attention when filtering the hue component. Improper filtering of the hue component can often lead to color artifacts.