Fast Line and Object Segmentation in Noisy and Cluttered Environments using Relative Connectivity
Proceedings of the Conference on Image Processing, Computer Vision, and Pattern Recognition, 2011.
International Conference on Image Processing, Computer Vision, and Pattern Recognition (IPCV), Las Vegas, USA, 18.-21. Juli 2011
In applications such as 3D plane segmentation of road traffic environments using u/v-disparity-histograms, line extraction is a key component and has to be as fast and precise as possible. Hough Transform is a good way to detect straight lines but specific line segments limited by start and end points are still to be determined. The Line Patterns Hough Transform (LPHT) introduced by Yip
directly delivers potential start and end points using the principle of relative connectivity. But this approach poses some challenges, too. We modified his idea to use Standard Hough Transform (SHT) together with relative connectivity for a fast and robust line segment extraction even in environments strongly affected by noise and clutter. Furthermore, we demonstrate the benefit of modified LPHT and relative connectivity for object segmentation in noisy Synthetic Aperture Radar (SAR) or infrared (IR) data.