Liu and Zhang present a comprehensive and systematic survey of techniques for computational vision systems. Human vision systems are complex enough, but face image processing is particularly challenging because creating believable and engaging systems still requires substantial investigation and development. For example, the need for computers to recognize a user’s emotional state has application across many important areas. These include human-computer interaction, surveillance, teleconferencing, and entertainment technologies.
This book is divided into three parts, each interesting and informative: “Face Representations,” “Face Modeling,” and “Applications.” Typically, each section of each part contains a comprehensive description with equations and figures. Part 1 covers shape models, meshes, and surfaces. Part 2 presents modeling with sensors, images, environment maps, and morphing. Part 3 covers facial animation, appearance, tracking, and gaze. Knowledge of mathematical modeling is required to understand or implement the equations and algorithms.
I found Part 3 to be the most accessible because it deals with applications in common use, such as talking heads and avatars, in which facial expression synthesis is pivotal for believable interactions. Expression synthesis is complex and involves surfaces, motion, lighting, and textures. The authors discuss the results of their experiments and provide the equations used.
Other interesting sections are on eye-gaze correction for videoconferencing; a conversational agent; and the face-based human interaction proof system, which uses the Turing test.
Whether you are developing facial representations, modeling appearance, working with applications, or surveying the subject, this book is likely to have something for you and is a good resource to have on hand.