Color Refinement for Natural Scene Images

A new method for selective color refinement of natural scene images is proposed. It aims to perform a post processing visual improvement of selected chromatic classes. This application is suitable for consumer devices, where either physical or computational constraints lead to obtain poor color rendition. The proposed solution improves the accuracy of color reproduction for three selected chromatic classes, i.e. skin, vegetation, sky, which have the most impact on the human visual system. The entire process is carried out in a cost effective way, by an automatic chromatic classifier driven by an accurate statistic characterization of a large image database, followed by an adaptive and intensity preserving color enhancement process. Many consumer applications can take advantage from this approach: digital still cameras, color display drivers and other color reproduction systems.