IMPSAC: Synthesis of Importance Sampling and Random Sample Consensus

Torr, Philip H. S.; Davidson, Charles E.

doi:10.1007/3-540-45053-x_52

Cited by 28 publications

(25 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Using these values, there are four potential sets of change-points to evaluate at the third step. However, the change from P ðS 1 0 j0; 2Þ to P ðS 2 0 j0; 2; 3Þ is the same as that from P ðS 1 0 j0; 1; 2Þ to P ðS 2 0 j0; 1; 2; 3Þ since by (15), the probability of the sequence after a change-point at 2 is independent of the sequence before it. Hence the most probable of P ðS 1 0 j0; 2Þ and P ðS 1 0 j0; 1; 2Þ will lead to the most probable of P ðS 2 0 j0; 2; 3Þ and P ðS 2 0 j0; 1; 2; 3Þ.…”

Section: Finding Texture Change-points In a 1d Line Searchmentioning

confidence: 84%

“…In [12], the multi-modal representation used is a piecewise combination of Gaussian distributions and is used to track rapidly moving human figures, while in [13] image corners are used with a discrete, tree-like, representation of the multiple hypotheses. RANSAC [14] based approaches have also been used [5] and numerous extensions to ransac have been proposed, in particular the idea of using RANSAC to initalise a set of particles is employed by [15] as part of a structure from motion system.…”

Section: Previous Workmentioning

confidence: 99%

See 1 more Smart Citation

Multi-modal tracking using texture changes

Kemp

Drummond

2008

Image and Vision Computing

View full text Add to dashboard Cite

Section: Finding Texture Change-points In a 1d Line Searchmentioning

confidence: 84%

Section: Previous Workmentioning

confidence: 99%

Multi-modal tracking using texture changes

Kemp

Drummond

2008

Image and Vision Computing

View full text Add to dashboard Cite

“…Sullivan et al showed in [25] that, random particles can be guided by a variational search, with good convergence when the image differences between frames are low. They used a predefined threshold to switch between the probabilistic and deterministic tracking engines.…”

Section: Related Workmentioning

confidence: 98%

Efficient particle filtering using RANSAC with application to 3D face tracking

Lü¹,

Dai²,

Hager³

2006

Image and Vision Computing

View full text Add to dashboard Cite

Particle filtering is a very popular technique for sequential state estimation. However, in high-dimensional cases where the state dynamics are complex or poorly modeled, thousands of particles are usually required for real applications. This paper presents a hybrid sampling solution that combines RANSAC and particle filtering. In this approach, RANSAC provides proposal particles that, with high probability, represent the observation likelihood. Both conditionally independent RANSAC sampling and boosting-like conditionally dependent RANSAC sampling are explored. We show that the use of RANSAC-guided sampling reduces the necessary number of particles to dozens for a full 3D tracking problem. This method is particularly advantageous when state dynamics are poorly modeled. We show empirically that the sampling efficiency (in terms of likelihood) is much higher with the use of RANSAC. The algorithm has been applied to the problem of 3D face pose tracking with changing expression. We demonstrate the validity of our approach with several video sequences acquired in an unstructured environment. q

show abstract

“…Several methods use sampling of multiple minimum-size sets to estimate common underlying parameters, and find out which elements should be discarded. Among them there are RANSAC [18], MLESAC [19], and IMPSAC [20]. M-estimators look for optimum weighting of each element, instead of just trying to discard outliers [21][22][23].…”

Section: Related Workmentioning

confidence: 99%