Perceptually Based Color Texture Features and Metrics for Image Processing

Flexible retrieval of image databases is an important problem with application in video editing, photo-journalism, arts and museums, fashion and design industry, etc. Since color and texture are fundamental aspects of human visual perception, there is a need for an "intelligent" system for search and manipulation of color patterns. However, regardless of the application domain, to accomplish retrieval successfully, it is necessary to understand what type of color and texture information humans actually use and how they combine them in deciding whether two patterns are similar. In our previous work, we performed a subjective experiment which yielded perceptual mechanisms important for the comparison among color patterns. In this paper, we focus on the implementation of these perceptual criteria in an image retrieval system via the design of a novel set of features and metrics that correlate with human perception.