Is That a Smile? Gaze-Dependent Facial Expressions
Author(s): Vidya Setlur, Bruce Gooch.
Proceedings: 3rd International Symposium on Non-Photorealistic Animation and Rendering (NPAR'04), Annecy, France,
2004.
[BibTeX] 
Abstract:
Using artistic methods for manipulating perception as a basis, we present a technique that creates facial images with conflicting emotional states at different spatial frequencies. The foveal and peripheral components of the human visual system tend to interpret the emotional state of the subject differently, adding a degree of elusiveness to facial image. The system first isolates the coarser low spatial frequency components and finer high spatial frequency details from two images with differing facial expressions. Our technique then performs image segmentation with edge detection, and morphs the images. In practice we have found that high spatial frequency elements determine the dominant expression in the resulting image, while the low spatial frequency elements contribute subtlety.