Modeling and Information Processing

EMI - Image Motion Estimation

Objective

To develop and evaluate image sequence processing aimed at the estimation of the apparent motions in the image coordinate system.

Contribution

Begun in the 80s, the estimation of image movements, or optical flow, has become one of the most active fields in computer vision in the last ten years. This development is due to the increase in computing power and the wide distribution of video acquisition resources. The DTIM's work concerns motion information calculation algorithms (tracking of points in sequences, optical flow), on the one hand, and also the evaluation of the performance of these techniques. In both cases, we base our efforts on the image modeling work done in the SR (Super-resolution) project, in particular by using B-spline models. As far as the evaluation of performance is concerned, which is generally done in the literature on synthesized images, we are carrying out a precision study, jointly with the DMSC department, on experiments to measure material deformations by imaging. Finally, another avenue of research is the algorithm-architecture adequation for optical flow estimation, which we are studying with the AXIS department of the IEF (Orsay-Paris XI).

Illustrations

• B-Spline modeling for the estimation of the optical flow
• An iterative Lucas-Kanade algorithm for dense estimation of the optical flow
• Material deformation measurement by image correlation and accuracy assessment (with DMSC)

Last Update: 11 October 2006 - © ONERA 2009 - Terms of use