AVISynth : Upsizing & DeInterlacing

[Bonie81]

Following is the Raw Clip Interlaced Source 720x480 Image.



How to DeInterlace & upsize to 1980x1280p or (Interlaced) 1980x1280i without loosing details in AVISynth?

[guns1inger]

You can't upscale an SD clip to full 1080 HD (I assume you mean 1920 x 1080) and retain all the detail - you don't have enough detail to work with. Your sample is already soft lacking detail to begin with. Upscaling means creating non-existent pixels. Simple resizing just does straight interpolation. Some super samplers do more with motion compensation and estimation, but even then they are limited. Using a sharper resize filter (e.g. Lancsoz 4) and adding a small amount of noise can add the appearance of details and sharpness, but you can't really add what you don't have.

[Bonie81]

Thanks for the reply.

i just got satisfactory result with Tom Berry's simple resize.
with x264 it's shooted 80+mbps
results are not good but not bad as well.
just playing arounf with various clips to see differences in the results.

[guns1inger]

Another suggestion would be to try the limitedsharpen filter (or function). If speed isn't an issue, you might also look at this thread - post 14 has a script sample.

[manono]

Following is the Raw Clip Interlaced Source 720x480 Image.

It's not interlaced, but progressive with pulldown.

... without loosing details...?

The only way to do it without losing details is to buy the Blu-Ray. However, you might upscale using NNEDI3 followed by sharpening it with LimitedSharpenFaster or some such.

Oh, and why include 20MB of audio tracks in a 78MB file when all anyone's interested in is the video? Demux them next time. And there's no need to ZIP it.

[Gavino]

i just got satisfactory result with Tom Berry's simple resize.
with x264 it's shooted 80+mbps
results are not good but not bad as well.

It's not clear from your post whether you've done this, but you need to use the InterlacedResize() function from SimpleResize.dll, not the SimpleResize() function.

However, based on manono's observation that you have a progressive source with pulldown, you will get better results doing an IVTC before resizing.

[Bonie81]

Merry Christmas & Happy holidays to all of you.

Thanking you all for your inputs.

@Gavino
i hv done this by
Avisource(".\input.avi")
LoadPlugin(".\Plugin\SimpleResize.dll")
#interp = separatefields().selecteven().eedi2()
#tdeint(edeint=interp)
#SimpleResize(1920, 1280)
InterlacedResize (1920, 1280)
interlaced and Deinterlaced. I like the result more with "#"

@manono
The sample here is for some other purpose for Developers this has to be with multiple audio including DTS.

@guns1inger
got new idea from your link and details from the 14th post.
gonna try soon.

Thanks to all of you!

gonna try tdeint.nedi, and from this thread - post 14 soon.
here is with tdeint.eedi2 plus simple resize result.

[manono]

Why are you using AVISource for a DVD sample? I already told you that sample was progressive. If you actually do the deinterlacing in the script, the part that was commented out, you'll find you degraded it by deinterlacing
that which didn't need deinterlacing and you created a jerky playing video with a duplicate frame in every 5. If you make the D2V using Forced Film (and assuming the rest of the movie is like that sample), you'll get a
progressive 23.976fps D2V ready to use in your script with MPEG2Source. And no need for deinterlacing.

interlaced and Deinterlaced. I like the result more with "#"

Of course, since you commented out the deinterlacing lines, I guess you didn't deinterlace after all. Why did you even say it was interlaced to begin with?

[jagabo]

I agree with manono. There's no point in taking a progressive MPG video, performing pulldown to make an interlaced video, then using a mediocre deinterlacer like TDeint() to make it progressive again. Just use DgIndex and Mpeg2Source() to access the progressive frames from the VOB directly.

[Bonie81]

@manono n jagabo

i got ur point.

[jagabo]

That sample image was so discolored, distorted, and blurry I couldn't tell what it was at first. A simple BilinearResize (the least sharp resize filter in AviSynth) would be much better.



Of course, your graphics card or HDTV probably does better than that so there's no point.

That's frame 372 from the VOB sample by DgIndex/Mpeg2Source()/VirtualDub count.

[Bonie81]

@jagabo

nice work.
"That sample image was so discolored, distorted, and blurry"
That's Original sample without any encoding - as is.

I use addframenumber filter in avisynth,
how could you did that in VDub? Any filter to know Frame Number?
ofcourse, I am not much into VDub.

[manono]

how could you did that in VDub? Any filter to know Frame Number?

It gives you the frame number at the bottom. At least VDubMod does, and I assume VDub is the same.

That's Original sample without any encoding - as is.

Eh? The source is a 720x480 DVD, right? And your pic is 1920x1280 (and with bad aspect ratio) as encoded following that crappy script of yours, right?

[manono]

Deleted. It posted the same response twice, for some reason.

[jagabo]

nice work.

Not really. That was an example of the second worst upscaling method (worst would have been nearest neighbor).

That's Original sample without any encoding - as is.

Then something is extremely wrong with your computer.

I use addframenumber filter in avisynth,
how could you did that in VDub?

VirtualDub shows the frame number in the status bar near the bottom of the window.