A Method for Determining the Subjective Dominant Color of an Image Region by Support Vector Regression

A representative color of an image region is widely used in many image processing fields. However, it is difficult to estimate the appropriate lightness and colorfulness because those are varied by highlights and shades in the region. In this study, we propose a method to estimate the subjective dominant colorfulness and lightness of a region by the support vector regression whose input is the dominant hue and the pixel color distribution on the colorfulness-lightness plane in the CIECAM16 space. The experimental result shows that the estimated colors by our new method are closer to the subjective dominant colors annotated manually than colors by a conventional RGB meaning method and our previous method.