Transforming Facial Images in 2 and 3-D
Shigeru Akamatsu · 1997
We review image processing techniques for manipulating 2-D representations of facial images and show their potential for extension to 3-D. Techniques for the manipulation of 2-D images include those used for: (i) averaging images of different faces to form facial ‘prototypes’, (ii) automated caricature exaggeration of the way an individual face differs in shape and colour from a prototype and (iii) transformations manipulating perceived facial qualities while maintaining information about identity (e.g. modifying the apparent sex or age of a face). Each technique relies on 160 feature landmarks allocated to the original 2-D images (e.g. centre of left eye, tip of nose). These landmarks can also be found in the 3-D colour and depth maps obtained from Cyber-scans. Using 2-D image warping techniques to align colour and depth maps from different faces allows prototypes and caricatures and may similarly allow transformations of faces to be performed in 3-D.