Connected line drawing extraction from a polyhedral image

Abstract

The extraction of line drawings with positional information from the perspective view of three-dimentional objects is essential in image analysis and understanding. In this thesis, we proposed a new line drawing extraction algorithm which is based on the locality of the candidate corners found by cornerfinder. Weighted Directional Polarized Hough Transform (WDPHT), which is a new adaption of Polarized Hough Transform(PHT) to line clustering process, simplifies the line clustering process around corners and can acquire accurate line clusters. The grouping operation in line matching process reduces the redundant comparisons between nommatching line clusters. The postprocessing step removes the false corners and false lines, thus guarantees that our algorithm works stably even in noisy environment.