|Title||Robust invariant descriptor for symbol-based image recognition and retrieval|
|Publication Type||Conference Paper|
|Year of Publication||2007|
|Authors||Wong, A., and W. Bishop|
|Conference Name||1st IEEE International Conference on Semantic Computing|
|Keywords||content-based retrieval, Feature Extraction, geometric transformation, Hough transforms, Hough-based transform, image degradation, image recognition, image representation, image retrieval, robust invariant descriptor, symbol-based image recognition|
This paper presents a robust invariant descriptor for symbol-based image recognition and retrieval. A modified Hough-based Transform is used to extract parameter space information (i.e., position data and angular data) from a symbol image to derive an invariant descriptor. The proposed descriptor provides a compact representation of a symbol image that can be evaluated efficiently. The extracted descriptor is highly robust against geometric transformations such as translation, rotation, reflection, and scaling, and image degradation. A series of experiments were conducted using a set of architectural and engineering symbols subjected to geometric transformations and image degradation. The experimental results clearly show that the proposed descriptor can be used effectively for symbol recognition and retrieval.