Gabor filter-based feature extraction and its use in object shape matching are addressed. For the feature extraction multi-scale Gabor filters are used. From the analysis of the properties of the Gabor-filtered image, we know isolated dominant points generally exist on the object contour, when the filter design parameters are properly selected. The dominant points thus extracted are robust to the image noise, scaling, rotation, translation, and the minor projection deformation. Object shape matching in terms of a two-stage point matching is presented. First, a feature vector representation of the dominant point is used for initial matching. Secondly, the compatibility constraints on the distances and angles between point pairs are used for the final matching. Computer simulations with synthetic and real object images are included to show the feasibility of the proposed method.
CITATION STYLE
Sun, S. K., Chen, Z., & Chia, T. L. (2002). Invariant feature extraction and object shape matching using gabor filtering. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 2314, pp. 95–104). Springer Verlag. https://doi.org/10.1007/3-540-45925-1_9
Mendeley helps you to discover research relevant for your work.