2-D OBJECT RECOGNITION USING HIERARCHICAL BOUNDARY SEGMENTS.
Yuhua Luo, R.T. Chin, Charles R. Dyer · Rare & Special e-Zone (The Hong Kong University of Science and Technology) · 1985
The use of image pyramids and hierarchical models for 2-D object recognition is described. The image containing objects to be recognized is stored as a pyramid (layers of increasing spatial resolution) that represents the image at multiple levels of resolution. Objects are modeled as graphs, each containing nodes arranged in a hierarchy of corresponding spatial resolution. With the image pyramid and the model graph, recognition becomes a model-driven, coarse-to-fine matching procedure involving the matching of model nodes to the corresponding layers of the image pyramid. Experiments have been performed using scenes that contain different types of objects.