Extraction of Skeletal Shape Features Using a Visual Attention Operator
The goal of the shape extraction method presented in this paper was to obtain a concise, robust, and invariant description of planar object shapes for object detection and identification purposes. The solution of this problem was chosen in the form of a piecewise-linear skeleton representation of local shapes in a limited number of salient object locations. A visual attention operator, which can measure the saliency level of image fragments, selects a set of most salient object locations for concise shape description. The proposed operator, called image relevance function, is a multi-scale non-linear matched filter, which takes local maxima at centers of locations of the objects of interest. This attention operator allows a simple extraction of vertices for the skeletal shape description by local maxima analysis.
KeywordsObject Detection Shape Feature Local Shape Object Intensity Salient Location