Learning Local Objective Functions for Robust Face Model Fitting

• Published in: IEEE transactions on pattern analysis and machine intelligence Vol. 30; no. 8; pp. 1357 - 1370
• Main Authors: Wimmer, M., Stulp, F., Pietzsch, S., Radig, B.
• Format: Journal Article
• Language: English
• Published: Los Alamitos, CA IEEE 01.08.2008; IEEE Computer Society; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Acceleration; Algorithm design and analysis; Algorithms; Applied sciences; Artificial Intelligence; Computational models of vision; Computer science; control theory; systems; Computer Simulation; Computer vision; Concrete; Exact sciences and technology; Face - anatomy & histology; Face and gesture recognition; Face detection; Face recognition; Fittings; Humans; Image Enhancement - methods; Image Interpretation, Computer-Assisted - methods; Image Processing and Computer Vision; Imaging, Three-Dimensional - methods; Layout; Learning; Mathematical models; Model-based coding; Modeling and recovery of physical attributes; Models, Anatomic; Object recognition; Optimization; Pattern analysis; Pattern matching; Pattern Recognition, Automated - methods; Pattern recognition. Digital image processing. Computational geometry; Real time systems; Reproducibility of Results; Robustness; Sensitivity and Specificity; Shape; Studies; Subtraction Technique; Texture; Training data; Vision and Scene Understanding; Model-based coding; Computer vision; Modeling and recovery of physical attributes; Shape; Image Processing and Computer Vision; Vision and Scene Understanding; Computational models of vision; Real-time systems; Object recognition; Texture; Pattern matching; Face and gesture recognition; Image interpretation; Objective analysis; Gesture recognition; Global optimum; Real time system; Annotation; Search algorithm; face and gesture recognition; Model matching; Scene analysis; real-time systems; vision and scene understanding; Facies; Objective function; Pattern analysis; Artificial intelligence
• ISSN: 0162-8828; 1939-3539
• DOI: 10.1109/TPAMI.2007.70793

Model-based techniques have proven to be successful in interpreting the large amount of information contained in images. Associated fitting algorithms search for the global optimum of an objective function, which should correspond to the best model fit in a given image. Although fitting algorithms have been the subject of intensive research and evaluation, the objective function is usually designed ad hoc, based on implicit and domain-dependent knowledge. In this article, we address the root of the problem by learning more robust objective functions. First, we formulate a set of desirable properties for objective functions and give a concrete example function that has these properties. Then, we propose a novel approach that learns an objective function from training data generated by manual image annotations and this ideal objective function. In this approach, critical decisions such as feature selection are automated, and the remaining manual steps hardly require domain-dependent knowledge. Furthermore, an extensive empirical evaluation demonstrates that the obtained objective functions yield more robustness. Learned objective functions enable fitting algorithms to determine the best model fit more accurately than with designed objective functions.