Towards Realism in Facial Image Transformation: Results of a Wavelet MRF Method

Abstract

The ability to transform facial images between groups (e.g. from young to old, or from male to female) has applications in psychological research, police investigations, medicine and entertainment. Current techniques suffer either from a lack of realism due to unrealistic or inappropriate textures in the output images, or a lack of statistical validity, e.g. by using only a single example image for training. This paper describes a new method for improving the realism and effectiveness of facial transformations (e.g. ageing, feminising etc.) of individuals. The method aims to transform low resolution image data using the mean differences between the two groups, but converges on more specific texture features at the finer resolutions. We separate high and low resolution information by transforming the image into a wavelet domain. At each point we calculate a mapping from the original set to the target set based on the probability distributions of the input and output wavelet values. These distributions are estimated from the example images, using the assumption that the distribution depends on the values in a local neighbourhood of the point (the Markov Random Field (MRF) assumption). We use a causal neighbourhood that spans multiple coarser scales of the wavelet pyramid. The distributions are estimated by smoothing the histogram of example values. By increasing the smoothing of the histograms at coarser resolutions we are able to maintain perceived identity across the transforms while producing realistic fine-scale textures. We use perceptual testing to validate the new method, and the results show that it can produce more accurate shifts in perceived age and an increase in realism.

Miscalibrations in judgements of attractiveness with cosmetics

“…With traditional averaging methods, composites from multiple faces produce smoother skin texture and a more attractive appearance (Little & Hancock, 2002). Although the use of wavelet MRF texture transform is intended to reduce or remove this possibility (Tiddeman et al, 2005), we examined the issue empirically. We cropped a 114 × 114 pixels patch of skin from the right cheek of each model in her 0% no cosmetics image, her 50% sequence image, and her self-applied 100% cosmetics image.…”

“…JPsychomorph uses a wavelet Markov random field (MRF) method for interpolating realistic, fine-grain textures (Tiddeman, Stirrat, & Perrett, 2005). Using this method, high-resolution information such as colour and texture in a transformed image is calculated by assuming that the new pixel distribution is dependent on the values in the local neighbourhood of landmark points.…”

The impact of attachment on preschool children's preference for parent-resembling faces — A possible link to sexual imprinting

“…To this end, about 150 points around the characteristic facial features (e.g., around the nose, eyes, mouth, etc.) were designated with the image manipulation software Psychomorph (TIDDEMAN, BURT and PERRETT 2001;TIDDEMAN, STIRRAT and PERRETT 2005); these were used as reference points during the averaging of faces. Stimuli were made by transforming same-sex preschool composites 50% in shape in order to resemble either the subject's parents or an unknown individual.…”

A Compositional and Dynamic Model for Face Aging

“…From these composite faces, they extract average aging patterns and perform face aging by superimposing the color and texture difference onto the young face. In a similar approach, Tiddeman et al [34], [35] studied the prototype images for different age groups. The differences between prototypes at different age groups were defined as the axis of age transformation.…”