Color is one of the main visual cues in the recognition of an object or image. High-level color descriptors are important as these descriptors provide valuable insight into the understanding of the object and its color, or the content of an image. Color naming is used to select objects (by color), describe the appearance of an object, and generate semantic annotations. A vivid orange color that we see in pumpkins this time of year describes the color so well we can see it in our mind's eye. When someone says "olive green" we envision in our minds eye the color of green olives. These color descriptors not only describe an object but convey an impression of the object or scene. This paper presents a computational model for color categorization. In this work we start with a spectral or trichromatic data set and compute the National Bureau of Standards -ISCC recommendation for color names and notations from the Munsell Book of Color. This develops color vocabularies and an appropriate syntax. Next, we attach to the color name a representation of those values on screen including RGB values. This realized perceptual color representation ties the perceptual name and the color name together. In testing this method, the human color categorization in known color regions in different color spaces were identified accurately. These are also consistent with human observations.
展开▼
