Deriving Perceptually-Based Texture and Color Features for Image Segmentation, Categorization, and Retrieval

Northwestern University, Evanston IL

Investigators

Abstract

The rapid accumulation of large collections of digital images has created the need for efficient and intelligent schemes for image retrieval. Since humans are the ultimate users of most retrieval systems, it is important to organize the contents semantically, according to meaningful categories. This requires an understandingof the important semantic categories that humans use for image classification, and the extraction of meaningful image features that can discriminate between these categories. Recent research efforts have addressed the first problem, but the second remains quite elusive. This research effort is aimed at addressing this second problem, that is, the extraction of low-level image features that can be correlated with high-level semantics and used to capture the semantic meaning of an image. The key to this research is the development of a new methodology for segmenting images, based on perceptual models and principles about the processing of texture and color information. This involves the identification of semantically important, spatially adaptive, low-level color and texture features that can be combined algorithmically to obtain image segmentations that convey semantic information. The same perceptual models and principles can be used to relate the features of the segmented regions (color and texture features, as well as segment location, size, and boundary shape) to semantic concepts that can be used for content-based image retrieval. An integral part of this research is the design and execution of subjective experiments in order to obtain some key parameters for the color and texture features, as well as for linking low-level image features to image semantics.