Complex-valued wavelet transform applications in planar shape prototype generation and recognition

The prototype generation has been widely used in industrial design, medical imaging, computer animation, pattern recognition, and computer vision. We propose a new hierarchical tree based algorithm for generating a prototype contour from a class of similar planar solid objects. Our algorithm uses the wavelet moduli of a complex-valued wavelet to extract the contour features. Features of different objects are organized as a binary tree. The root of a subtree corresponds to the prototype contour for the contours of its children. We show that one can update this tree efficiently by only modifying the local subcontours. An experiment of our method to prototype generation and pattern recognition is demonstrated.