Measures of Similarity Between Objects Based on Qualitative Shape Descriptions

• Published in: Spatial cognition and computation Vol. 13; no. 3; pp. 181 - 218
• Main Authors: Falomir, Zoe, Gonzalez-Abril, Luis, Museros, Lledó, Ortega, Juan Antonio
• Format: Journal Article
• Language: English
• Published: Philadelphia, PA Taylor & Francis Group 03.07.2013; Taylor & Francis
• Subjects: Applied sciences; Artificial intelligence; Categories; Computation; Computer science; control theory; systems; conceptual neighborhood diagrams; Construction; Descriptions; Exact sciences and technology; interval distance; Invariants; Pattern recognition. Digital image processing. Computational geometry; qualitative representation; shape; Similarity; similarity measure; Translations; similarity measure; Geometrical shape; Invariant; Computer vision; Similarity; shape; conceptual neighborhood diagrams; Digital image; qualitative representation; interval distance; Efficiency; Morphological analysis; Size effect; Qualitative analysis; Metric
• ISSN: 1387-5868; 1542-7633
• DOI: 10.1080/13875868.2012.700463

A computational approach for comparing qualitative shape descriptions (QSDs) of objects within digital images is presented. First, the dissimilarity of qualitative features of shape is measured: (i) intuitively using conceptual neighborhood diagrams; and (ii) mathematically using interval distances. Then, a similarity measure between QSDs is defined and tested using images of different categories of the MPEG-7-CE-Shape-1 library, images of tiles used to build mosaics, and a collection of Clipart images. The results obtained show the effectiveness of the similarity measure defined, which is invariant to translations, rotations and scaling, and which implicitly manages deformation of shape parts and incompleteness.