Robust Line Extraction Method Based on Pyramid Hierarchical Structure

The pyramid hierarchical structure is known as a framework which is fit to extract global features from local ones through hierarchically arranged multi-resolution planes. However, there is a problem in which a mapping process between consecutive planes is complicated and not so effective in a line extraction.In this paper we propose a new method devising a new mapping between consecutive planes. We show that the proposed method is robust against noise and effective to extract and describe the line structures. We examine the system performance for noisy images and several kinds of line structures comparing with the Hough transformation method and confirm that the proposed method can extract and describe the line structures. Furthermore, we demonstrate that the vectorization of a property land map as real data can be done effectively with high precision and noise removing is performed automatically.