Inria / Raweb 2004
Project-Team: EVASION

Search in Activity Report, year 2004:


Project-Team : evasion

Section: New Results

Motion capture from video

Participants: Fabien Dellas, Laurent Favreau, Alexandre Perrin, Lionel Reveret.

This research project aims at capturing motion from the automatic processing of video to provide information for 3D animation of characters, such as humans or animals. Unlike several approaches in computer vision research, the goal is not to recognize activities, but rather to acquire robust geometric hints to control animation. Three main projects are currently under investigation: facial animation, deformation of skin surface and motion of animals.

Motion capture of facial expression

Participants: Alexandre Perrin, Lionel Reveret.

Previous results showed that 3D animation of facial expression can be described with a compact set of linear modes. This parametric reduction guarantees robustness when applied to the tracking of facial motion from video. However, it does not take into account the motion due to expressions such as smiling. The goal of this project is to combine several methods of facial animation into a coherent framework. As preliminary results, a method has been implemented to allow integration between linear models for speech production and pseudo-muscles models for expressions (see section 7.3).

In order to allow the processing of large video data such as video documentaries or video streaming from a webcam, previous works on motion capture have been adapted to be computed on GPU. This has implied a reformulation of the algorithms so that the approach could be implemented on the specific architectures of pixel shaders.

Motion capture of animal motion

Participants: Laurent Favreau, Lionel Reveret.

The motion of animals is still a challenging problem in 3D animation, both for articulated motion and deformation of the skin and fur (see Figure 17). The goal of this project is to acquire information from the numerous video footage of wild animals. These animals are impossible to capture into a standard framework of motion capture with markers. There are several challenges in the usage of such video footage for 3D motion capture : only one 2D view is available, important changes occur in lighting, contrast is low between the animal and foreground, etc. Currently, a method has been developed to first extract a binary silhouette of the animals and then, to map this silhouette to pre-existing 3D models of animals and motion thanks to a statistical prediction.

This work has been published at the Symposium on Computer Animation 2004 (SCA'04) [18]. It will be continued in collaboration with the University of Washington at Seatle.

Figure 17. Motion capture of animal motion

Motion capture of skin and muscles motion

Participants: Fabien Dellas, Lionel Reveret.

The modeling of precise deformation of skin due to muscles and bones motion is still challenging. In order to accuratly extract the 3D motion of the skin, a dedicated suit has been built. It consists of stretchy fabric textures with a regular checker board pattern. The recording under three different viewpoints allows to reconstruct in 3D the surface of the skin during a bulging of the muscle for example. One of the contributions has been also to show that statistical criteria could be determined to evaluate if the tracking of the surface is giving non reliable results. This preliminary work has been published at the Workshop on Modelling and Motion Capture Techniques for Virtual Environments [16].