Adaptive deformable model for mouth boundary detection
Ali Reza Mirhosseini · Optical Engineering · 1998
University of SydneyDepartment of Electrical EngineeringSydney, New South Wales 2006AustraliaE-mail: [email protected] LamHong Kong Polytechnic UniversityDepartment of Electrical EngineeringHong KongAbstract. A new generalized algorithm is proposed to automatically ex-tract a mouth boundary model from human face images. Such an algo-rithm can contribute to human face recognition and lip-reading-assistedspeech recognition systems, in particular, and multimodal human com-puter interaction systems, in general. The new model is an iterative al-gorithm based on a hierarchical model adaptation scheme using deform-able templates, as a generalization of some of the previous works. Therole of prior knowledge is essential for perceptual organization in thealgorithm. The prior knowledge about the mouth shape is used to defineand initialize a primary deformable model. Each primary boundary curveof a mouth is formed on three control points, including two mouth cor-ners, whose locations are optimized using a primary energy functional.This energy functional essentially captures the knowledge of the mouthshape to perceptually organize image information. The primary model isfinely tuned in the second stage of optimization algorithm using a gener-alized secondary energy functional. Basically each boundary curve isfinely tuned using more control points. The primary model is replaced byan adapted model if there is an increase in the secondary energy func-tional. The results indicate that the new model adaptation technique sat-isfactorily generalizes the mouth boundary model extraction in an auto-mated fashion.