How to achieve best results decompressing dv files before restoring

[FLP437]

In the first image I have used levels as you have said but without the dither option so the spikes in the histogram . In your opinion what will be better plan levels or levels with the ditter option not speaking about speed.

[poisondeathray]

In the first image I have used levels as you have said but without the dither option so the spikes in the histogram . In your opinion what will be better plan levels or levels with the ditter option not speaking about speed.

I don't see a huge difference in terms of quality using smoothlevels vs. levels(coring=false, dither=true) when used like that. Nothing that jumps out that says oh definitely pick this one. The only big difference is speed.

But levels is more basic; smoothlevels has other settings you can use , special limiting modes (lmodes), ecurve, ecenter, chroma processing adjustment (by default it's 50, which is between pure YLevels and what levels does) . Some people would use smoothlevels16 for higher precision (but you're not going to see the difference on dv source footage, waste of time IMO)

[jagabo]

The problem isn't that they're less than 720x576, they're much bigger. For example, in this image:


[Attachment 40571 - Click to enlarge]

the video image (outlined in red) is approximately 1178x936.

[FLP437]

It's what I have tried to explain they have been obtained with a screen grabber utility in a 4k monitor 3840x2160 resolution

[FLP437]

For the present problem a clip with 16-255 should I use some of the more fancy options in smoothlevels or plain vanilla use is the correct way

[poisondeathray]

For the present problem a clip with 16-255 should I use some of the more fancy options in smoothlevels or plain vanilla use is the correct way

There is no necessarily "correct" way . If you need some of the other functions, then you have to use smoothlevels or something else . Test it out for yourself and see what you prefer. For typical DV footage, levels is good enough IMO .

[_Al_]

the best way to compare avisynth scripts seem to interleave them, this way and using steps you can really see the differences, you can easily be fooled comparing images side by side on screen and that way you can also introduce lots of errors as well

video1 = AviSource("D:\DV\base.avi", fourCC="DVSD", pixel_type="YV12").Subtitle("AviSource").SmoothLevels(16, 1.0, 255, 16, 235).histogram()
video2 = ffvideosource("D:\DV\base.avi").Subtitle("ffvideosource").SmoothLevels(16, 1.0, 255, 16, 235).histogram()
interleave (video1,video2)

step by frames in VirtualDub or MPC-HC etc.
takeaway that smoothlevel or histogram if you do not want to see that while comparing or ask someone to change those subtitles for you like subtitles("X") and subtitle("Y") and you find yourself what is better looking ... this sort of thing is actually funny stuff, our brain is really weird, if for some reason you already made a decision you would tend to think in that direction does not matter what, so you better not know what you are actually comparing
the only problem in this case is if those videos have different YUV color space etc.

[FLP437]

Thanks _AI_ I will try your approach

[jagabo]

It's what I have tried to explain they have been obtained with a screen grabber utility in a 4k monitor 3840x2160 resolution

You scaled the frames in VirtualDub. If you want someone to analyze the images you need to post them 1:1. Do what you were told long ago: use Video -> Copy Source Frame To Clipboard (Ctrl+1) then paste from the clipboard into an image editor and save as PNG.

The reason you got different scaling with two instances of VirtualDub is because one was using DirectX and the other not. I recommend you turn off DirectX acceleration in VirtualDub to prevent this. Options -> Preferences -> Display -> Use DirectX...

[FLP437]

@jagabo
If the scaling you say I have done is generated by directx and if virtualdub with unique settings generates two instances one with directx and the other without, I assume it can be true .Also the windows dimension of each instance could have not been exactly equal.

If you are saying I have used in virtualdub or avisynth a scalling filter I can assure you I have not, but I cannot force you to believe it and I´m not interested to argue about this detail as we are going anywhere with it and we lose the large picture. The idea with the previously presented images was not that they were images for work but only to be visualized and to detect any defects however I assume I should have posted individual and lossless images. Related to direct virtualdub extraction and PNG I don’t think I have posted any image after the recommendation to do it, so this insistence seems somewhat extemporaneous and not necessary.

To extract a direct PNG from virtualdub I prefer I different approach , select and mark two adjacent frames to isolate one frame, export using image sequence in PNG assuring colorspace is RGB24 or 32 and you are done no need to external picture edit program, if I use the proposed process due to the large workplace I still have to crop to get the original dimension, for me it´s more complicated.

Also the images you select have been the ones processed with your function wave and history the others are the one´s I´m referring always and they are almost raw and are related to avisource and ffmpeg the other only differ by levels using and are more strongly processed.