View-based active appearance models

Abstract

AbstractÐWe describe a new method of matching statistical models of appearance to images. A set of model parameters control modes of shape and gray-level variation learned from a training set. We construct an efficient iterative matching algorithm by learning the relationship between perturbations in the model parameters and the induced image errors.

Inferring 3D structure with a statistical image-based shape model

“…This single-view approach can be extended to 3D by considering multiple simultaneous views of features [8]. Shape models in several views can be separately estimated to match object appearance [5]; this approach was able to learn a mapping between the low-dimensional shape parameters in each view.…”

Sieving Regression Forest Votes for Facial Feature Detection in the Wild

“…Instead of modelling them together like the classic AAMs [8], several methods have focused on using shape models to regularize the local detections like the Constrained Local Models (CLMs) [22], Branch & Bound optimization [2] and tree structured shape models [30]. Instead of using parametric shape models, non-parametric representations of shape constraint include [4] and a series of cascaded pose regression approaches [12,11,6,5].…”

Multi-camera head pose estimation

“…These can be roughly divided into two categories: 2D view-based and 3D modelbased approaches. Examples of the former include the use of support vector machines (SVM) [16,22], PCA [32], Kernel PCA (KPCA) [8], independent subspace analysis [33], Gabor filters [41,42] and networks [19], active appearance models (AAM) [11], shape-from-shading [9] and 2D geometric heuristics [23], amongst others. Model-based approaches can be tackled from a concise mathematical formulation [15], or if real-time analysis is required, by simplifying the problem using affine transformations [13].…”