File Exchange

image thumbnail

Video Face Replacement

version 1.0.0.2 (4.21 MB) by Mahmoud Afifi
Matlab implementation of our paper "Video Face Replacement System Using a Modified Poisson Blending Technique"

6 Downloads

Updated 14 Mar 2019

View License

This source code is the implementation of the paper under title:
Video Face Replacement System Using a Modified Poisson Blending Technique
http://ieeexplore.ieee.org/document/7024453/
So if you will use this code, please cite the original paper.
Watch the video:
https://www.youtube.com/watch?v=sVP5kSt8LuM


Run the demo of the software using
>>demo
Be sure, the dataset located in the root folder of the source code folder "../dataset".
If you need to apply the software to your own videos, you must:
1- Convert the videos into sequnce of frames, see the dataset.
2- Target video frames should be in the format %5d.jpg and start from frame 00000.jpg
3- Source video frames should be in the format %5d.png and start from frame 00000.png
4- Source video frames should be stabilized first using, stabilizing function
5- Convert the stabilized frames of the source actor into png frames that contain the interesting region of the actor.
Notes:
- The stabilized frames should have alpha channel to be used for extracting the mask.
- Adjust the cropped interesting region frames for getting your desirable results.
- See the dataset "target folders".

Citation
Afifi, Mahmoud; Hussain, Khaled F.; Ibrahim, Hosny M.; Omar, Nagwa M., “Video Face Replacement System Using a Modified Poisson Blending Technique,” Intelligent Signal Processing and Communication Systems (ISPACS), 2014 International Symposium on , vol., no., pp.205,210, 1-4 Dec. 2014

Cite As

Afifi, Mahmoud, et al. “Video Face Replacement System Using a Modified Poisson Blending Technique.” 2014 International Symposium on Intelligent Signal Processing and Communication Systems ({ISPACS}), {IEEE}, 2014, doi:10.1109/ispacs.2014.7024453.

View more styles

Comments and Ratings (9)

@Ashique K: the stabilization process applies the inverse of the tracking affine transformations to the current original frame. Given a sequence of face frames, the stabilization detects the face, applies the tracking process, then inverse the tracking affine transformations to stabilize the face.

Ashique K

Great work.
But how to do stabilization?
Please explain the procedure for stabilization.
(What are the corrections needed in "stabilization.m")

@Xu in the paper we assume it's given. In practice, we generate it using Adobe Aftereffects; however, you can write a Matlab code to manually select the ROI. But remeber, you have to apply the stablization on it before extracting the ROI

Xu

Dear mahmoud afifi, how to generate the facial part images in dataset\3\source ?

Xu

Yasir Codur

I cant download the Dataset from Dropbox, because it not exist. can you fix that

Yu Chi

Updates

1.0.0.2

..

1.0.0.1

.

1.0.0.0

A correct link to the data was updated

1.0.0.0

.

MATLAB Release Compatibility
Created with R2013b
Compatible with any release
Platform Compatibility
Windows macOS Linux