Dynamic Background Video Forgery Detection Using Gaussian Mixture Model
Video is a valid evidence in court case and therefore its integrity must be proven first. There are several researches regarding video forgery detection such as histogram correlation analysis published by Jie Xu et al. However, this method is too dependent with pixel line placement that will be used...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/36292 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Video is a valid evidence in court case and therefore its integrity must be proven first. There are several researches regarding video forgery detection such as histogram correlation analysis published by Jie Xu et al. However, this method is too dependent with pixel line placement that will be used as an area to calculate histogram correlation from. Therefore, this method needs an improvement in pixel belt placement.
This research used foreground detection method that used Gaussian mixture model to determine areas that is possible for pixel line placement. Based on foreground location acquired, pixel line then placed intersecting dan filling a foreground object. Using this method, the pixel line will be placed in an area that has a lot of changes and therefore maximizing the observation result.
Using foreground detection, histogram correlation analysis method yield improvement both from detection accuracy and localization precision and recall. Furthermore, using foreground detection also opens up a new possibility of histogram correlation analysis method to detect spatially forged video which is proven by the 90% recall rate |
|---|