The USC Andrew and Erna Viterbi School of Engineering USC Signal and Image Processing Institute USC Ming Hsieh Department of Electrical Engineering University of Southern California

Technical Report USC-IPI-550

“The Role of Texture in Computerized Scene Analysis”

by William Baker Thompson

December 1974

An important aspect of image interpretation is the segmentation of a scene into regions corresponding to objects in the scene. The need is demonstrated for incorporating textural information processing capabilities into scene segmentation systems. Visually distinct textural discontinuities often correspond to object boundaries in an image. Such discontinuities may be recognized as adjacent image regions differing significantly in some visually identifiable property. A "perceptual distance function" is developed which estimates the perceived dissimilarity between two image regions. This model is structured to make use of multiple, often redundant, and occasionally contradictory elementary measurements about the nature of the regions involved. In addition, the problem of determining the size of regions over which textural measurements are to be made in a scene analysis system is investigated. A method is proposed for bounding this region size based on information about human textural resolution. Experimental results are presented which demonstrate the success of the analysis techniques.

This report is not currently available in PDF format for downloading. Contact the Signal and Image Processing Institute for information on its availability.