PROPERTIES OF DOMINANT COLOR TEMPERATURE DESCRIPTOR

摘要

The concept of color temperature is derived from an innate characteristics of the human visual system. It is formulated as a visual feature referring to a kind of perceptual feeling about perceived light. Color temperature has also a physical-based definition, and hence, color temperature of observed scenes and visual objects can be modelled in a mathematical way as a one-parameter characteristics of perceived light. In the Amendment to the Visual Part of the MPEG-7 Standard the Color Temperature descriptor for image browsing has been proposed. To extend the functionality of content-based image search by color temperature we proposed the Dominant Color Temperatures descriptor, which allows a user to perform query by example and query by value searches. The extraction algorithm was originally adopted from dominant color's one, which utilizes vector quantization in 3D color space. We also proposed a second, much faster algorithm based on scalar quantization in one-dimensional color temperature space. In this paper we present a comparison of the two extraction algorithms. We also compare the querying results of the Dominant Color Temperature descriptor and two conceptually related descriptors: Dominant Color and Color Temperature.