How would one go about detecting a video splice using ONLY the video frame data? Audio seems to be much easier to use in general for this purpose since you can actually see the break in the waveform in pic related etc. The problem for me is trying to find it when there is no audio to work with, especially when the splice is during black frame transitions.
Any video forensics fags here?
Other urls found in this thread:
clem.dii.unisi.it
twitter.com
Speedrunners are faggots who should kill themselves.
:(
literally no one gives a fuck. So what, some guy faked a speedrun probably to get you and other autismos assblasted so everyone can have a laugh.
Literally no one gives a fuck whether anyone cheated or not in speedrunning, because speedrunning is for faggots that should kill themselves.
I don't think you can use histograms for splice in video. How will you know it's not a valid cut in the video?
You can maybe try to detect some of the visual effects people may use to hide a splice.. that may work..
A screen recording? There is no way because they're going to do the splice during a fade to black or something. If you want shit that can't be faked without NSA-tier efforts, make them submit VHS recordings.
As a professional analyzer of video footage, I can tell when a video was spliced by looking at the pixels and seeing where the encoding algorithm differs based on the compression noise variation.
Seeing irregularities in compression defects maybe. Like how you can see when a jpeg have been edited.
The only things I can think of are accurate transition frame data and consistent in-game data. Frame perfect cuts are easy to perform on transition scenes.
The final encoding is the most recent source of generation loss, so its compression pattern is the strongest. Detecting older compression artifacts may be possible, but difficult and inconclusive.
Thus, the best way to tell from just the video is just the consistency of game-specifics, like menu states or predictable frame animations.
Ultimately, the only way to tell if something is real is with live performance and a controller camera.
This doesn't detect fade-to-black transistions, but it could detect cutscene transistions.
Black levels in games purposefully never reach actual RGB(0,0,0) noise artifacts are still visible with the right monitor, or just a simple frame analysis using photoshop and/or mspaint with the fucking fill bucket tool.
Thanks dude that's pretty cool I didn't know that
Hmm, good point for recordings with a capture card. Not necessarily for PC, though.
The only reason why you would need to detect a frame separation only using a video track is for speedrunning veedogays. The only other purpose I could think of is with a sketchy security footage for an actual court case, in which case OP can go sit on a fucking dick.
Wow, someone actually wrote a paper on what I could do to detect a good fucking shoop on /b/ in 2005 when I was fucking 16.
Not diagreeing with ya lol.
I'll give this a read sometime. But it made me remember that images can be transformed into a 2d signal space, and videos a 3d signal space. Shoops may provide inconsistencies there.
Part of science is formalizing and verifying, if it's assumed to be common knowledge
When was the last time you read a paper about the mathematical molecular functions your brain uses for firing the specific order of electrical signals to enable muscle contractions in the specific areas that allow you to blink?
Why would I care about that?
I'm saying that "when pushed, killing becomes as easy as breathing." -John Rambo
So you're saying women giving birth are the best murderers?
It's just your typical signal processing that you do with audio, but instead of the 1-dimensional signal you have with audio you now have a 2D signal (because video has both X,Y)
