NEW: A journal version (almost final) from TPAMI, March 2007, can be found here (pdf, 30MB).
A shorter version from CVPR 2004 is here (pdf, 3.5MB).
We present a method for space-time completion of large space-time holes in video sequences of complex dynamic scenes. The missing portions are filled-in by sampling spatio-temporal patches from the available parts of the video, while enforcing global spatio-temporal consistency between all patches in and around the hole. This is obtained by posing the task of video completion and synthesis as a global optimization problem with a well-defined objective function.The consistent completion of static scene parts simultaneously with dynamic behaviors leads to realistic looking video sequences. Space-time video completion is useful for a variety of tasks, including, but not limited to:
NEW: An implementation inspired by this work for filling images is available here.
NEW: These are new results -- improved since the video demo in CVPR'2004. See details in the journal version.
NEW: Added raw uncompressed data for some of the following examples. These videos can be used in publications provided the above paper is cited.
| Removal of a beach umbrella Raw data may be found here (zip, 9MB). |
|
| Removal of the jogging lady
Raw data may be found here (zip, 5MB). |
|
| Reconstruction of three missing frames in a sequence | ![]() |
| Removal of the waving lady
Raw data may be found here (zip, 32MB). |
|
| Retouching of old Chaplin movie with simulated large "coffee stain" |
|
| Full field-of-view of stabilized video.
The input was taken from this nice work |
|