Determining region color by using maximum colorfulness

Color naming has been widely used as an object attribute or an image region feature. However, estimating appropriate color names is difficult even for single-color objects because the lightness and colorfulness of pixels typically vary depending on the highlight and shade. This study proposes a simple and robust method to estimate the color name of a given region. The proposed algorithm uses the color distribution of the image region. Specifically, this method first estimates the primary hue of the region in the CIECAM16 color space. Next, the colorfulness distribution of pixels with the primary hue is examined. It is used to determine the representative colorfulness and lightness. Finally, categorical color naming is performed in the CIECAM16 color space. Compared with the mean color of the region, the proposed method can produce more saturated colors. This method can be applied to various applications such as image retrieval, image indexing, object attribute recognition, non-photorealistic rendering, and color transformation.