mesh models an object as a collection of polygons in 3D space (including depth information). The surface texture of a 2D or 3D mesh may be compressed as a static texture image (using the DWT) which is projected onto the mesh at the decoder.

These tools have a number of applications including coding and representing synthetic (computer generated) or ‘natural’ 2D and 3D objects in a scene and may be particularly useful for applications that combine natural and computer-generated imagery (Synthetic Natural Hybrid Coding, SNHC). It is possible to use mesh-based coding for compression of natural video objects. For example, a mesh and a static texture map may be transmitted for selected key frames. No texture is transmitted for intermediate frames, instead the mesh parameters are transmitted and the decoder reconstructs intermediate frames by animating (deforming) the mesh and the surface texture. This type of representation has much more flexibility than block-based motion models, since the mesh can adapt to deformations and nonlinear motion between frames. At the present time, however, automatic generation and tracking of mesh parameters is prohibitively complex for practical video compression applications.

The Basic Animated Texture object type includes the Binary Shape, Scalable Still Texture and 2D Dynamic Mesh tools and supports coding of 2D meshes (with uniform topology) and arbitrary-shaped still texture maps. The Animated 2D Mesh object type is a superset of Basic Animated Texture and adds Delaunay mesh topology (a more flexible method of defining mesh triangles) and the Core Profile video coding tools to enable a flexible combination of animated meshes and video objects.

5.8.2 Face and Body Animation

MPEG-4 Visual includes specific support for animated human face and body models within the Simple Face Animation and Simple FBA (Face and Body Animation) visual object types. The basic approach to face/body animation consists of (a) defining the geometric shape of a body or face model (typically carried out once at the start of a session) and (b) sending animation parameters to animate the body/face model.

A face model is described by Facial Definition Parameters (FDPs) and animated using Facial Animation Parameters (FAPs). The default set of FDPs may be used to render a ‘generic’ face at the decoder and a custom set of FDPs may be transmitted to create a model of a specific face. Once the model is available at the decoder, it can be animated by transmitting FAPs. In a similar way, a body object is rendered from a set of Body Definition Parameters (BDPs) and animated using Body Animation Parameters (BAPs). This enables the rendering of body models ranging from a generic, synthetic body to a body with a specific shape.

Applications for face and body animation include the generation of virtual scenes or worlds containing synthetic face and/or body objects, as well as model-based coding of natural face or body scenes, in which face and/or body movement is analysed, transmitted as a set of BAPs/FAPs and synthesized at the decoder. To date, these methods have not been widely used for video coding.

5.9 CONCLUSIONS

The MPEG-4 Visual standard supports the coding and representation of visual objects with efficient compression and unparalleled flexibility. The diverse set of coding tools described in