MLESAC: a new robust estimator with application to estimating image geometry
Computer Vision and Image Understanding - Special issue on robusst statistical techniques in image understanding
Multiple View Geometry in Computer Vision
Multiple View Geometry in Computer Vision
Matching with PROSAC " Progressive Sample Consensus
CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 1 - Volume 01
Guided-MLESAC: Faster Image Transform Estimation by Using Matching Priors
IEEE Transactions on Pattern Analysis and Machine Intelligence
Preemptive RANSAC for live structure and motion estimation
Machine Vision and Applications
Genetic Algorithm SAmple Consensus (GASAC) - A Parallel Strategy for Robust Parameter Estimation
CVPRW '06 Proceedings of the 2006 Conference on Computer Vision and Pattern Recognition Workshop
IEEE Transactions on Pattern Analysis and Machine Intelligence
A Comparative Analysis of RANSAC Techniques Leading to Adaptive Real-Time Random Sample Consensus
ECCV '08 Proceedings of the 10th European Conference on Computer Vision: Part II
| Hi-index | 0.00 |
RANSAC (RANdom SAmple Consensus) is a popular approach for robust model parameter estimation and data point filtering from noise or outlier contaminated data. In a repeating hypothesize-and-verify procedure, RANSAC selects minimal sets of random data points until a certain confidence level that a good model has been chosen is reached. Like all its descendants however, the reselection of already evaluated hypotheses causes worthless recalculations without any gain in information. Without any prior information and under the assumption that the order of data points in minimal sample sets is irrelevant, this paper proposes a new sampling strategy, which avoids the reselection of already chosen data point combinations at the cost of higher memory consumption. Using P-RANSAC, which is a close relative of RANSAC based on samples without identical data points, as an example for enhanced sampling, the increase in efficiency to be expected is theoretically analysed and confirmed in a stereo camera re-localisation experiment. Results demonstrate that for applications with few model parameters and inliers the proposed NRC-RANSAC (Non-Repeating Combinations RANSAC) is able to perform up to 5 times faster than P-RANSAC.