This is a simple Matlab source code of FVC, where many features are discarded from this code like morphing, inverse completion, and keypoints.
Read the paper:
Watch the video:
Afifi, Mahmoud, et al. “Fast Video Completion Using Patch-Based Synthesis and Image Registration.” 2014 International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS), IEEE, 2014, doi:10.1109/ispacs.2014.7024452.
In the paper, we generate it using image melding technique (Siggraph 12). However, you can get it even using Patch completion in Photoshop. If the first frame of the video doesn't contain the unwanted objext, you don't need to do anything, just use the first frame as a clean frame. Also, you are supposed to use short videos. If you need to complete long videos, it requires the forward/backward completion that is described in the paper but not supported in this Matlab code.
How to get the clean image?