1. Affine Capture and Retargeting - U think this is easy .. sigh !!
Capturing the affine component of the motion requires the determination of the 6 affine parameters - 4 for rotaion, scaling and shear and 2 for translation. I did the following to determine these -
-
Preprocessing -
-
The cartoon whose motion is to be captured is to be seperated from the background.
-
This was done using a color based region growing segmentation.
-
After this the silhouette of the segmented figure is made and we work with this silhouette.
|
|
Original Frame |
Silhouetted Frame |
-
Correspondence and Estimation -
-
Since a tracker was not used to track points across frames, the point correpondences between each pair of frames had to be clicked manually.
|
|
Frame n |
Frame n+1 |
-
This was followed by an estimation of the affine parameters using standard computer vision techniques.
-
Then the affine matrices were then concatanated to give the affine transformations from the selected key frames to the inbetween frames.
-
Retargeting -
-
Retargeting just involes applying the affine tranformations to the target keyframes to get the inbetween frames of the target animation. The target keyframes have to be similarly preprocessed however.
-
Problems -
-
Things which cannot be captured in a affine transform cannot be transferred to the target - these include out of plane rotations and translations, and local deformations.
-
Out of plane rotations and translations are transferred as per the animators interpretation of the same while making the target keyframes.
-
Note that the method is sensitive to manual clicking though not very much. The clicking is supposed to be intuitive. Minor aberrations in clicking do not harm the process because of the stylized cartoon motion being captured dominates the visual perception.
[ Back ]
Page last updated on 04 October, 2005.
|
AT cse.iitd.ac.in
|
© Parag Chaudhuri , 2009
|
|