NAPSAC: High Noise, High Dimensional Robust Estimation - it's in the Bag

Myatt, D. R.; Torr, Philip H. S.; Nasuto, Slawomir J.; Bishop, J. M.; Craddock, R.

doi:10.5244/c.16.44

Cited by 129 publications

(106 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Different guided sampling strategies have been proposed in literature. NAPSAC (Myatt et al, 2002) is based on the assumption that the probability of finding an inlier adjacent to another inlier is high. This is based on the assumption that inliers of a model tends to lie closely.…”

Section: Methodsmentioning

confidence: 99%

“…If the percentage of inliers is very low, the convergence of RANSAC is generally slow. To deal with these issues, many variants to RANSAC like MLESAC (Torr and Zisserman, 2000), progressive sample consensus (PROSAC) (Chum and Matas, 2005), locally optimized RANSAC (LoSAC) (Chum et al, 2003), NAPSAC (Myatt et al, 2002) etc have been proposed by the computer vision community (Choi et al, 2009). Many of these techniques are applied to 2D vision problems like epipolar geometry estimation, motion segmentation, homography estimation, object recognition, image retrieval etc.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Robust Estimation Technique for 3d Point Cloud Registration

Pankaj

Nidamanuri

2016

Image Anal Stereol

View full text Add to dashboard Cite

The 3D modeling pipeline involves registration of partially overlapping 3D scans of an object. The automatic pairwise coarse alignment of partially overlapping 3D images is generally performed using 3D feature matching. The transformation estimation from matched features generally requires robust estimation due to the presence of outliers. RANSAC is a method of choice in problems where model estimation is to be done from data samples containing outliers. The number of RANSAC iterations depends on the number of data points and inliers to the model. Convergence of RANSAC can be very slow in the case of large number of outliers. This paper presents a novel algorithm for the 3D registration task which provides more accurate results in lesser computational time compared to RANSAC. The proposed algorithm is also compared against the existing modifications of RANSAC for 3D pairwise registration. The results indicate that the proposed algorithm tends to obtain the best 3D transformation matrix in lesser time compared to the other algorithms.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Robust Estimation Technique for 3d Point Cloud Registration

Pankaj

Nidamanuri

2016

Image Anal Stereol

View full text Add to dashboard Cite

show abstract

“…Our method addresses this fact in two ways. In [21] Myatt et al propose to bias random selection towards clusters in a multi-dimensional space. Following this approach the first interest point A is selected randomly.…”

Section: Incremental Computation Of Good Hypothesesmentioning

confidence: 99%

Incremental Model Selection for Detection and Tracking of Planar Surfaces

Prankl

Zillich

Leibe

et al. 2010

Procedings of the British Machine Vision Conference 2010

View full text Add to dashboard Cite

Man-made environments are abundant with planar surfaces which have attractive properties and are a prerequisite for a variety of vision tasks. This paper presents an incremental model selection method to detect piecewise planar surfaces, where planes once detected are tracked and serve as priors in subsequent images. The novelty of this approach is to formalize model selection for plane detection with Minimal Description Length (MDL) in an incremental manner. In each iteration tracked planes and new planes computed from randomly sampled interest points are evaluated, the hypotheses which best explain the scene are retained, and their supporting points are marked so that in the next iteration random sampling is guided to unexplained points. Hence, the remaining finer scene details can be represented. We show in a quantitative evaluation that this new method competes with state of the art algorithms while it is more flexible to incorporate prior knowledge from tracking.

show abstract

“…Some works developing new robust estimators were reported [7]. However, the initial matches are possibly sparse and with high ratio false matches, too early to get the global geometry relationship estimation or even bring on a bad result.…”

Section: Introductionmentioning

confidence: 99%