Probabilistic Generative Modelling

Larsen, Rasmus; Hilger, Klaus Baggesen

doi:10.1007/3-540-45103-x_114

Cited by 2 publications

(1 citation statement)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the original formulation a threshold at 95-98% was used on the cumulated variance described by the principal components. Alternatives include 1) "elbow" identification in scree-plots of covariance matrix eigenvalues [18]; 2) statistical comparison of covariance matrix scree-plot with scrambled data covariance scree-plots [19]; 3) probabilistic PCA modelling using AIC, BIC or cross validation [20]. For the compression AAM models the following methods have been applied.…”

Section: Discussionmentioning

confidence: 99%

Wedgelet Enhanced Appearance Models

Darkner

Larsen

Stegmann

et al.

2004 Conference on Computer Vision and Pattern Recognition Workshop

View full text Add to dashboard Cite

Abstract-Statistical region-based segmentation methods such as the Active Appearance Model (AAM) are used for establishing dense correspondences in images based on learning the variation in shape and pixel intensities in a training set. For low resolution 2D images correspondences can be recovered reliably in realtime. However, as resolution increases this becomes infeasible due to excessive storage and computational requirements. In this paper we propose to reduce the textural components by modelling the coefficients of a wedgelet based regression tree instead of the original pixel intensities. The wedgelet regression trees employed are based on triangular domains and estimated using cross validation. The wedgelet regression trees are functional descriptions of the intensity information and serve to 1) reduce noise and 2) produce a compact textural description. The wedgelet enhanced appearance model is applied to a case study of human faces. Compression ratios of the texture information of 1:40 is obtained without sacrificing segmentation accuracy notably, even at compression ratios of 1:150 fair segmentation is achieved.

show abstract

Section: Discussionmentioning

confidence: 99%