Estimate the All Vanishing Points from a Single Image

【Author】

Yongyan Yu;Zhijian Wang;Yuansheng Lou;

【Abstract】

Vanishing point is the important precondition for camera self-calibration from a single image. Previously proposed solutions, either relies on voting in the Gaussian sphere space, or iteratively time and again base on Maximal Likelihood Estimator. All these methods expend lots of time, have large error and low level of efficiency. Hence,a new scheme will be presented with a recently proposed algorithm called J-Linkage,in which each edge is represented with the characteristic function of its preference set and vanishing points are revealed as clusters in this conceptual space.First,it estimates all possible vanishing point, and then refines them by Expectation Maximization. Finally, two experiments show that algorithm reduces the number of variables, error measures are done in the image, a consistency measure between a vanishing point and an edge of the image can be computed in closed-form.So that it has a low computation and high precision.

【Keywords】

preference set,consistency measure,Jaccard distance,clustering,refine

References

To explore the background and basis of the node document

Springer Journals Database

Total: 10 articles

  • [1] Rene Vidal;;Yi Ma;;Shankar Sastry, Generalized Principal Component Analysis (GPCA), IEEE Transactions on Pattern Analysis and Machine Intelligence, IEEE Transactions on Pattern Analysis and Machine Intelligence (IEEE Transactions on Pattern Analysis and Machine Intelligence)
  • [2] AGUILERA D G;LAHOZ J G;CODES J F, A new method for vanishing points detection in 3d reconstruction from a single view, Proceed-ings of the ISPRS Commission V,
  • [3] Roberto Toldo;Andrea Fusiello, Robust Multiple Structures Estimation withJ-Linkage, European Conference on Computer Vision,
  • [4] HADAS K;RON M;RENATO K, Vanishing points estimation by self-similarity, IEEE Transactions on Pattern Analysis and Machine Intel-ligence,

More>>

Similar documents

Documents that have the similar content to the node document

More>>