Abstract

The use of the Internet as a means to socialize and to communicate is now widespread. In real life we have learnt to ‘read’ faces and guess feelings. However, in cyberspace it is not as easy to conduct similar behaviour. The use of expressive images to accompany text may provide some of the missing visual cues. In this paper, we describe an expressive image generator for generating a set of expressive images from a single default image. For each expression category the generator can generate images with three different intensities. Our method requires users to provide only one default image, six control points and three control shapes. A series of experiments is presented that tested the recognition rates of the expressions depicted in the generated images. The results demonstrate that more than 60% of the participants correctly recognized the synthesized human facial expressions in the middle- and high-intensity expression categories and more than 70% correctly recognized the cartoon expressions in all intensities.