On the hough transform of multi-level pictures

摘要

The Hough Transform (HT) is a popular tool for pattern extraction on a discrete planar grid. Unfortunately, it cannot work directly with grey-level or other multi-level images, requiring them to be segmented in advance. The Radon Transform (RT) is free of this limitation at the expense of extra computations being required. In this paper, we introduce the so-called Digital Halftoning Hough Transform (DHHT), which is able to process grey-level images without relying upon any assumption on the picture origin, contents or grey-level distribution as in the RT, and hold the HT computation time. The benefit in time, depending on the average image intensity, is estimated quantitatively in comparison with the RT computed in spatial and Fourier domains. Major DHHT techniques are assessed with regard to the approximation quality metrics established in this paper. The DHHT merits are demonstrated on the artificial example of edge detection, along with the real problem of Young's fringe pattern analysis, successfully solved by the DHHT.