Unable to decimate accurately

[thetoof]

I have a crappy VHS rip that I can't decimate well. I think it's because of the high amount of noise that makes the detection harder for the filter, but I'm not sure... I tried using selectevery(5,1,2,4,5) and it works for some cycles, but some other need selectevery(5,1,2,3,5). So, I was wondering if there was any way to use a denoised reference clip (with a fast denoiser) or if I should decimate after denoising (if the high amount of noise is the problem), but I'd prefer to do it before since I'm using a VERY slow denoiser and it'd save me a lot of encoding time.

I don't think it needs deinterlacing since there's no combing, so maybe I should use FixBlendIVTC to remove the blending?? I'm not familiar with this filter, so I'm pretty much guessing what I just said.

The file is an uncompressed YUY2 clip, so an uploaded sample would be HUGE. Here (http://s168.photobucket.com/albums/u173/thetoof/?action=view&current=3-Colorcorrection.jpg)'s a frame where you can see the noise I'm talking about. As a reference, this (http://s168.photobucket.com/albums/u173/thetoof/?action=view&current=5-Chromadenoise.jpg) is what it looks after denoising.

[Zep]

I have a crappy VHS rip that I can't decimate well. I think it's because of the high amount of noise that makes the detection harder for the filter, but I'm not sure... I tried using selectevery(5,1,2,4,5) and it works for some cycles, but some other need selectevery(5,1,2,3,5). So, I was wondering if there was any way to use a denoised reference clip (with a fast denoiser) or if I should decimate after denoising (if the high amount of noise is the problem), but I'd prefer to do it before since I'm using a VERY slow denoiser and it'd save me a lot of encoding time.

I don't think it needs deinterlacing since there's no combing, so maybe I should use FixBlendIVTC to remove the blending?? I'm not familiar with this filter, so I'm pretty much guessing what I just said.

The file is an uncompressed YUY2 clip, so an uploaded sample would be HUGE. Here (http://s168.photobucket.com/albums/u173/thetoof/?action=view&current=3-Colorcorrection.jpg)'s a frame where you can see the noise I'm talking about. As a reference, this (http://s168.photobucket.com/albums/u173/thetoof/?action=view&current=5-Chromadenoise.jpg) is what it looks after denoising.

ok now this one you should do a search cause I know there is a ton of threads about it including a thread were me and neuron2 tried a ton of stuff on clips which I also sent to him. (he wrote the decomb filter and a lot more great filters)

basically you need a decimation that floats (pretend the pattern is ocean swells rising and failing and you are the decimation boat and you will get what I mean lol) you want to compare just 2 frames always and if the difference between them is less than your threshold then delete one else keep them both. The problem of course is audio sync since it is not straight forward 1 in 5 drop etc... the final framerate needs to be adjusted so that the time length matches the time length before the decimation.

I have always felt a blend would be great for this since it could look at a run of frames then drop the for sure dupes then blend a smooth run of the rest. I did a test a long time ago where I compared 2 frames at a time dropped 1 if it was VERY VERY close to the other and kept both if not, then I frame doubled the whole thing then frame doubled it again. then ran normal decimation on it which ended up smoothing the video perfectly because there were now a lot of in between frames to chose from so decimation kept about half original frames before the double double but the rest it chose doubled frames to keep instead of back to back original frames which jumped a good bit at times and made it look jerky. The problem was audio. I ended up having to stretch it but that was easy and worked fine lol. yes I could have changed the video framerate also (Note this was all an experiment I would not suggest doing it unless nothing else works lol)

[egrimisu]

what is this doing :selectevery(5,1,2,4,5) ?
is this deleting thoes frames first the fifth, then the next , then the second ...?

[Leak]

what is this doing :selectevery(5,1,2,4,5)?
RTFineM (http://avisynth.org/mediawiki/SelectEvery#SelectEvery)

I'm pretty sure it should be "SelectEvery(5,0,1,2,4)" anyway - the 5 at the end is out of range...

All it does is for every cycle of 5 frames (which are numbered 0 to 4) return frames 0,1,2 and 4, i.e. leaving out frame 3.

np: Blamstrain - Harju pt. 2 (Exosphere)