We present a method that extracts structural features of images. The method is based on both a region-based analysis and a contour-based analysis. The image is first segmented, based on its pixels' information. Color information of each segmented region is performed by using the hue-saturation-value color space. Area of each region is also extracted by counting the number of bound pixels. Location of each region is computed as a center of the region's convex hull. A contour of the region is approximated by a B-spline approximation to obtain its control polygon and curve in the limit. The region's convex hull is obtained from the control polygon. For multi-scale features, we apply Chaikin's algorithm to the control polygon for finer level of control polygons, which could be used in a coarse to fine comparison. Curvature information of the B-spline curve fitting could also be used in the comparison. Our method could be used in many interesting applications including image retrieval, image classification, image clustering, image manipulation, image understanding, pattern recognition, and machine vision. © 2007 Springer.
CITATION STYLE
Mongkolnam, P., Dechsakulthorn, T., & Nukoolkit, C. (2007). Extracted structural features for image comparison. In Innovations and Advanced Techniques in Computer and Information Sciences and Engineering (pp. 13–18). Kluwer Academic Publishers. https://doi.org/10.1007/978-1-4020-6268-1_3
Mendeley helps you to discover research relevant for your work.