Human body modeling and motion analysis from video sequences
Pascal Fua, Grün, Armin, Plänkers, Ralf, Nicola D'Apuzzo, Daniël Thalmann · Infoscience (Ecole Polytechnique Fédérale de Lausanne) · 1998
Equations ( 14) should be amended by the constraint .(16)In order to stay within the framework of least squares observation equations we also express (16) as observation equation, but with an associated very small weight p D for the pseudo-observation D:Equations ( 14) , ( 15) and ( 17) will contribute to (8). CONCLUSIONSWe have presented some basic ideas how we intend to use image data from video sequences for the modeling of a human body under motion.This clearly represents an initial report with some preliminary results, including 3-D determination and tracking of passive marker points, extraction of silhouette data, generation of surface models by image matching, and setting up a framework for joint least squares estimation.We have outlined a technique that allows us to fit a simplified animation model to noisy image data with very limited manual intervention.Because this model is closely related to the complete model we apply to perform animation, these results can be used to initialize this complete model and further refine it using the same data.The capability we intend to develop will be of great applicability in an area such as the generation of feature films for entertainment.Generating and animating sophisticated models requires a tremendous amount of manual labor.While this may be appropriate for big-budget one-off use, the mass market of television entertainment is much more cost-driven and would benefit greatly from using techniques such as those described above.Furthermore, there currently is an inherent limit to the complexity of the animation models: Realism requires complex models, that is, models that are controlled by large numbers of parameters.As this number increases, so does the difficulty of the task faced by the designer.Automating the process will help solve this problem and will allow an increase in realism while reducing the cost.