Aging Faces : Learning Facial Growth Models

Developing computational models for human faces that account for different facial appearances due to factors such as illumination variations, head pose variations, varying facial expressions, occlusions etc. has long been of interest in the computer vision and psychophysics communities. Human faces convey significant information pertaining to individuals such as their identity, gender, age group, ethnicity etc and further facial expressions often help identify the emotional state of individuals. Hence perception studies suggest that human faces are associated with high psychosocial importance and identify attributes such as facial attractiveness, facial age etc. as factors that regulate interpersonal behavior. From a face recognition perspective, numerous algorithms have been developed to perform still-image based and video-based face recognition in the presence of illumination and head pose varitions [5]. In this paper, we develop computational models that characterize facial aging effects commonly observed during formative years (0 to 18 years) and in turn help perform face recognition across age progression. During formative years, while shape variations in the cranium are prominent with increase in age, textural variations in the form of wrinkles and other skin artifacts are minimal. Hence, we address the proposed facial growth model as a craniofacial growth model : a shape transformation model that characterizes facial growth observed during formative years by means of drifts observed in facial features with age. While developing facial growth models, taking into consideration the following aspects is crucial to the success of the model in characterizing facial aging effects :