ORIENTED PARTICLE SPRAY: PROBABILISTIC CONTOUR TRACING WITH DIRECTIONAL INFORMATION

Contour following is a standard activity in rotoscoping in the digital post production domain. An artist might need to cut out or edit an object separately from its background and it is left to the artist to manually create the cut out. Techniques for automatically tracing the edges of the object exist, but these operate with heavy manual intervention. The most recent technique called JetStream is a considerable advance on manual or semi-automatic tracing, but suffers from a lack of direction information in the image. This paper considers the incorporation of this information and so reworks the principle of density propagation for contour following. The approach is more robust than previous methods although inevitably needs user intervention to incorporate image semantics.

[1]  Eero P. Simoncelli,et al.  Steerable wedge filters , 1995, Proceedings of IEEE International Conference on Computer Vision.

[2]  Timothy J. Robinson,et al.  Sequential Monte Carlo Methods in Practice , 2003 .

[3]  Neil J. Gordon,et al.  A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking , 2002, IEEE Trans. Signal Process..

[4]  Patrick Pérez,et al.  JetStream: probabilistic contour extraction with particles , 2001, Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001.

[5]  Weichuan Yu,et al.  Approximate orientation steerability based on angular Gaussians , 2001, IEEE Trans. Image Process..

[6]  Pietro Perona,et al.  Deformable kernels for early vision , 1991, Proceedings. 1991 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[7]  Neil J. Gordon,et al.  A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking , 2002, IEEE Trans. Signal Process..

[8]  H. Barlow Vision: A computational investigation into the human representation and processing of visual information: David Marr. San Francisco: W. H. Freeman, 1982. pp. xvi + 397 , 1983 .