l33tmeatwad wrote:First off, how are you removing the interlacing, was it soft telecined and you used force film?
Most important question to answer, indeed. However...
Additionally, since the artifacts (as it appears) in those samples are going vertically, it isn't interlacing (as that would be horizontal), that would be aliasing.
No, the user reported another issue, specifically the chroma. Compare the two images he posted: in the latter, the coloring of her hair is out of place, as it is above her head.
This could be due to a number of different reason, from issue in source, to bad deinterlace, hence why whomever the other user might be, his suggestion of double checking the deinterlacing isn't too bad per se.
It's just that, without an actual m2v sample of the scene, it's hard to tell.
Honestly I don't think the light aliasing shown there is anything to worry about, and I think "fixing" it would introduce more issues, instead. But something should be done about the chroma, if possible.
Secondly, with TTempSmooth() commented out (with the # in front of it) you have NO noise reduction set (outside of the deblocking in mpeg2source).
I don't think ttempsmooth would be the best choice here anyway, and using mpeg2source's cpu parameter to deblock is a bad way to go about it to boot (so I suggest to change the cpu parameter to 0).
For blocking, mosquito noise, and ringing, I'd suggest a combination of smoothd2
, and possibly hqdering
As a reminder, stay low with the parameters, as it's better to underfilter than to overcompensate, unless you're doing a frame-based filtering, in which case you can afford to overfilter individual frames. Keep in mind these are all issuess introduced by the encoder (MPEG-2 is especially bad on this, considering how old and limited of a format it is, particularly with DVD restrictions) since it didn't have enough bitrate to properly store the frames, so clearly some frames are going to be in a worse shape than others and would need a different amount of filtering.
Third, why are you using a derainbow filter? I'm fairly certain those DVDs do not have a rainbow issue (maybe I'm wrong).
I think it was his attempt at fixing the chroma problem he has shown in the screenshots.
Anyway, I'd also get rid of:
ConvertToYV12(): Your source already is i420, so using ConvertToYV12() is a NOP() right now, thus useless.
FastLineDarkenMod(): Please don't do line darkening unless you have extremely valid reasons to ─ keep in mind that recent sources (digital ones as opposed to transfers from analog sources, in most cases) do not need it.
LSFmod(strength=120): Please don't sharpen unless you have valid reasons to ─ if it's because of a blur introduced by the denoising/deblocking, see if using better deblocking/denoising parameters or perhaps contrasharpening afterwards are of more help; if it's due to the source being an upscale, inverse upscale back to the original resolution; if it's because the source looks to be somewhat blurry, it might make sense, but first try to understand whether it was an artistic choice or not (possible in recent sources); in older sources, it most likely makes sense if there is a bit of sharpening needed.
To be honest, most of your questions about cleaning up the source are answered in this guide
here on the org.
Not quite... I haven't really done almost anything about it yet, but the avtech really does need improvements on a few things, avs filters being one of the most prominent issues.