Regularized Bundle-Adjustment to Model Heads from Image Sequences without Calibration Data
read more
Citations
Accurate Camera Calibration from Multi-View Stereo and Bundle Adjustment
Articulated soft objects for video-based body modeling
Modeling Facial Geometry Using Compositional VAEs
3D Assisted Face Recognition: A Survey of 3D Imaging, Modelling and Recognition Approachest
Robust and Rapid Generation of Animated Faces from Video Images: A Model-Based Modeling Approach
References
The finite element method
Numerical Recipes, The Art of Scientific Computing
A morphable model for the synthesis of 3D faces
A robust technique for matching two uncalibrated images through the recovery of the unknown epipolar geometry
Related Papers (5)
Frequently Asked Questions (11)
Q2. What future works have the authors mentioned in the paper "Regularized bundle-adjustment to model heads from image sequences without calibration data" ?
In future work, the authors will use articulated models to extend their approach to this new application field. However, calibrated stereo pairs or triplets can be used as well. The animation model [ Kalra et al., 1992 ] the authors use can produce the different facial expressions arising from speech and emotions. Because only the 2-D location of these points need to be specified, this can be done very quickly.
Q3. What is the way to automate the fitting process?
Successful approaches to automating the fitting process have involved the use of optical flow [DeCarlo and Metaxas, 1998] or appearance based techniques [Kang, 1997] to overcome the fact that faces have little texture and that, as a result, automatically and reliably establishing correspondences is difficult.
Q4. Why did the authors choose to demonstrate and evaluate their technique?
The authors chose to demonstrate and evaluate their technique mainly in the context of head-modeling because it is the application for which the authors have all the tools required to perform the complete reconstruction task.
Q5. Why is it important to use the facial animation software?
This is important because the facial animation software depends on the model’s topology and its configuration files must be recomputed every time it is changed, which is hard to do on an automated basis.
Q6. How do the authors solve the structure-from-motion problem?
The authors have shown that by incorporating model-based constraints in the framework of bundle-adjustment, the authors are able to effectively tackle the structure-from-motion problem in a case where correspondences are difficult to establish.
Q7. How do the authors simulate the errors that can be expected from their stereo matcher?
To simulate the errors that can be expected from their stereo matcher, the authors corrupt these projections by adding two kinds of noise:1. White noise with variance σnoise ∈ {0.5pixel, 1.0pixel}.
Q8. Why is the f value only approximate?
It is only approximate because, CamCal, in effect, can also trade changes in the value of f against changes of the estimated distance of the camera to the calibration grid it uses.
Q9. How do the authors increase the robustness of the standard procedure?
To increase the robustness of their algorithm, the authors augment the standard procedure in two ways:1. Iterative reweighted least squares.
Q10. How do the authors estimate the positions and orientations of the two images on either side of the central?
To estimate the positions and orientations for the two images on either side of the central image, the authors begin by retriangulating the surface of the generic face model of Figure 3(a,b) to produce the regular mesh depictedby Figure 3(c,d) that the authors call the bundle-adjustment triangulation.
Q11. What is the way to estimate pose from the head?
In the case of head tracking, a generic 2–D face model can be used to estimate roughly estimate pose from appearance [Lanitis et al., 1995].