Dealing with Noise in Multiframe Structure from Motion

Thomas, J. Inigo; Oliensis, J.

doi:10.1006/cviu.1999.0779

Cited by 17 publications

(12 citation statements)

References 48 publications

(86 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Experimental results confirm that fusing accurate two-image reconstructions leads to increased accuracy [73,100,101] (see also [5,8,27,60,85,86]). When some of the two-image reconstructions are very inaccurate, the fused reconstruction can grow less accurate as more images are acquired [73,100,101].…”

Section: Fusing/kalman Filteringsupporting

confidence: 54%

“…In fact, this is well known to be true for a general complex estimation problem, 8 and there is no reason to believe that the situation differs in SFM. Though the results of [3,5,7,8,56,59,60,73,85,86,88,90,100,101,110] show that SFM fusing can improve intermediate reconstructions that are already reasonably accurate, this intrinsic and general problem with fusing suggests that it may fail when they are not. It is important to understand experimentally the usefulness and limitations of SFM fusing in practical applications.…”

Section: Fusing/kalman Filteringmentioning

confidence: 98%

“…Fusing, including Kalman filtering, is the second traditional approach to MISFM after optimization [3,5,7,8,56,59,60,73,85,86,88,90,100,101,110]. By fusing, I mean any MISFM technique that computes a final multi-image reconstruction from intermediate reconstructions (and estimates of the uncertainties in these) rather than from the image data directly.…”

Section: Fusing/kalman Filteringmentioning

confidence: 99%

See 2 more Smart Citations

A Critique of Structure-from-Motion Algorithms

Oliensis

2000

Computer Vision and Image Understanding

Self Cite

131

View full text Add to dashboard Cite

Section: Fusing/kalman Filteringsupporting

confidence: 54%

Section: Fusing/kalman Filteringmentioning

confidence: 98%

Section: Fusing/kalman Filteringmentioning

confidence: 99%

See 1 more Smart Citation

A Critique of Structure-from-Motion Algorithms

Oliensis

2000

Computer Vision and Image Understanding

Self Cite

131

View full text Add to dashboard Cite

“…The analysis of Hartley et al [19] shows the huge influence of noisy correspondences in the 3D point triangulation, where the authors estimate a noise level of σ = 0.2 in their real world images. Hebert [13] deals with uncertainty in SfM with noise variance up to 1 pixel, being this variation the overall trend in the field [20]. By contrast, our images present a noise level σ = 10.58 (estimated), due to the JPEG-compression artefacts inherent in wireless communication implemented on such low-cost omnidirectional platforms.…”

Section: Prior Workmentioning

confidence: 96%

Improved 3D sparse maps for high-performance SFM with low-cost omnidirectional robots

Cavestcmy

Rodrı́guez²,

Barberá

et al. 2015

2015 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

Citation for published item:gvestnyD F nd odr¡ %guezD eFvF nd wrt¡ %nezEfrer¡ D rF nd frekonD FF @PHISA 9smproved Qh sprse mps for highEperformne pw with lowEost omnidiretionl rootsF9D in PHIS siii snterntionl gonferene on smge roessingD sgs PHISD PUEQH eptemer PHISD uee gityD gD gnd Y proeedingsF xew orkD eX siiiD ppF RWPUERWQIF Further information on publisher's website: Additional information: Use policyThe full-text may be used and/or reproduced, and given to third parties in any format or medium, without prior permission or charge, for personal research or study, educational, or not-for-prot purposes provided that:• a full bibliographic reference is made to the original source • a link is made to the metadata record in DRO • the full-text is not changed in any way The full-text must not be sold in any format or medium without the formal permission of the copyright holders.Please consult the full DRO policy for further details. ABSTRACTWe consider the use of low-budget omnidirectional platforms for 3D mapping and self-localisation. These robots specifically permit rotational motion in the plane around a central axis, with negligible displacement. In addition, low resolution and compressed imagery, typical of the platform used, results in high level of image noise (σ ∽ 10). We observe highly sparse image feature matches over narrow inter-image baselines. This particular configuration poses a challenge for epipolar geometry extraction and accurate 3D point triangulation, upon which a standard structure from motion formulation is based. We propose a novel technique for both feature filtering and tracking that solves these problems, via a novel approach to the management of feature bundles. Noisy matches are efficiently trimmed, and the scarcity of the remaining image features is adequately overcome, generating densely populated maps of highly accurate and robust 3D image features. The effectiveness of the approach is demonstrated under a variety of scenarios in experiments conducted with low-budget commercial robots.

show abstract

“…It is said that known 3D points are provided to obtain the camera calibration values. Oliensis (12) dealt with noise in multi-frame structure. Motion error model is used to estimate 3D points error by the correlations between motion error and structure error.…”

Section: Introductionmentioning

confidence: 99%

Fusion of 3 D Shapes in Multiple View Points to Obtain More Accurate Shape

2000

View full text Add to dashboard Cite

The aim of this paper is to obtain an exact 3D model of a scene using multiple images from different camera positions. The reconstruction of 3D shape using two images has problem such as being weak at noise. Therefore we present methods to reduce noise and to improve the accuracy of 3D shape with multiple images. The system is divided into three stages: firstly, reconstruction of 3D shapes at different camera positions, secondly fusing the 3D shapes to obtain a para-ideal shape, and thirdly removing the outlier shapes and feature points by evaluation function, and fusing the rest of shapes. Even though the corruption of image data by noise is one of the unavoidable problems in any system, this paper shows how well the noise is removed by the proposed algorithms in multiple view points. We demonstrate a significant improvement of recovered shape. Experimental results show that our system performs well to remove noise with robustness. The maximum noise reduction rate is 82% in the real image experiment.

show abstract

Dealing with Noise in Multiframe Structure from Motion

Cited by 17 publications

References 48 publications

A Critique of Structure-from-Motion Algorithms

A Critique of Structure-from-Motion Algorithms

Improved 3D sparse maps for high-performance SFM with low-cost omnidirectional robots

Fusion of 3 D Shapes in Multiple View Points to Obtain More Accurate Shape

Contact Info

Product

Resources

About