Hierarchical Two View Line Segment Matching Using Wavelet Transform

摘要

This paper presents a novel algorithm for line segment matching over two views. Two image pyramids are built by applying wavelet decomposition to the two images. After defining supporters for a line segment to be the edge points lying close to it, line segments are matched from the coarsest to the finest level. At each level the supporters are matched using the cross-correlation techniques. This method can match line segments over two uncalibrated views, in which the line segments need not be the images of the same section of a 3D line segment. The hierarchical strategy helps to reduce the computational complexity. Wavelet is adopted to build the hierarchical frame for its built-in multi-scale structure and fast decomposition algorithm. Furthermore, it overcomes the flattening-out problem in the traditional multi-scale Gaussian pyramid technique. Experiments on real image pairs are given to demonstrate the effectiveness and the robustness of our algorithm.