FineSharp()

[Boulder]

Do you have a similar idea for encoding stuff? From what I've noticed, most 1080p material can safely be downscaled to 720p but the result is a bit softer than the original when rescaled to 1080p on the TV/BD player/Media player. I wouldn't say that there's any actual detail lost in the process with most movies, it's just reduced sharpness.

[malmsteen81]

Originally Posted by Didée

Also, switching the cstr operation from gaussian-kernel (positive sstr) to box-kernel (negative sstr) can make some gentle difference.

do you mean positive or negative "mode" value? right? i tried to reduce cstr to 0.5 but the difference are subtle. the default values are ok for almost cases.

[Didée]

Oh, yes, sorry. The switch is build into "mode", not into "sstr". My bad memory.

@Boulder:
That's pretty much how I see it. The actual detail loss from 1080p > 720p usually is quite little, and the issue mostly is just edge sharpness when 720p is upscaled during playback. (Given you've made a good 720p rip in the first place, of course.)

[Boulder]

Originally Posted by Didée

That's pretty much how I see it. The actual detail loss from 1080p > 720p usually is quite little, and the issue mostly is just edge sharpness when 720p is upscaled during playback. (Given you've made a good 720p rip in the first place, of course.)

So simply using LSFMod with edgemode=1 could actually be a good solution for compensating the difference between the two resolutions? I have used that in some cases but I'm unsure whether there are better alternatives.

[Didée]

It's quite feasable, but most likely not optimal. (But frankly: I've almost never really used the given LSFmod function. Somehow I always end up with manual code to sharpen).

After being guilty for the original LimitedSharpen, nowadays I don't like the limiting operation too much anymore. It is somewhat valid for broad edges, but comes with downsides upon thin detail, and risk of uglyness if slight halo'ing is alread pesent. Thinking over it, today I'd rather see working with referencing a gaussian blend (to reduce present edge halos) plus a median cut (both together to "empty" flat areas & weak detail), get an edgemask from that, then use limited sharpening gotten from the gaussian blend on masked edges, and SeeSaw-style impulse sharpening on non-masked areas.

Btw, for downsizing I'm often in favour of "bicubic with negative 'b' ", like e.g. bicubicresize(1280,720,-.5,.25). Neagtive b is hideous for upscaling, but comes very nice for downsizing. New rule: "b + 2c = 1 // for upsizing", and "b + 2c = 0 // for downsizing".

[turbojet]

Thanks for finesharp, it's the best sharpener I've used. It has much more of an effect then lsf(mod) while still looking natural without adding artifacts, although it emphasizes artifacts in the source. Unfortunately it's really buried in this thread, Didée do you think it deserves it's own thread like most great scripts get?

Also thanks for mentioning negative 'b' parameters in bicubicresize, it's the first time I've seen them mentioned outside of the avisynth wiki which says negative values are undesirable but it also suggests bilinear over bicubic for downsizing. I've settled on (-1,0.75) for encoding.

[ryrynz]

Yeah, the topic could probably be updated to "Finesharp : The definitive sharpening script for bluray'" or given a new thread. It's not everyday Didée spits out a new sharpening script.

Originally Posted by Didée

Thinking over it, today I'd rather see working with referencing a gaussian blend (to reduce present edge halos) plus a median cut (both together to "empty" flat areas & weak detail), get an edgemask from that, then use limited sharpening gotten from the gaussian blend on masked edges, and SeeSaw-style impulse sharpening on non-masked areas.

Sounds good, I'd like to see it. Could be something that replaces Seesaw & LSF.

[Gser]

Personally I use ResampleHQ(width=x, height=y, srcmatrix="xyz", kernel="Spline64", chroma_kernel="Spline36") to downsize.

[Budtz]

Could some1 point me to a guide for making this work?

[Adub]

Threw FineSharp on the Avisynth wiki, just in case: http://avisynth.org/mediawiki/FineSharp

[Sp00kyFox]

Originally Posted by Didée

Thinking over it, today I'd rather see working with referencing a gaussian blend (to reduce present edge halos) plus a median cut (both together to "empty" flat areas & weak detail), get an edgemask from that, then use limited sharpening gotten from the gaussian blend on masked edges, and SeeSaw-style impulse sharpening on non-masked areas.

sounds interesting. is there a implementation already available for this idead or could you post an example script?

[gioppatta]

Is it usable with a denoiser? what's a good combo?

[THX-UltraII]

Is there newbie step-by-step guide somewhere that helps me config. this script?

What I already have installed on my HTPC:
- Avisynth 2.5.8
- ffdShow rev.4453 x86 (set as RAW output, needed for SVP frame)
- Microsoft Visual C++ 2008 SP1 MFC Security Update
- MPC-HC 1.6.2.4472
- LAV Filters 0.50.1
- madVR 0.84.2

[djmasturbeat]

I keep trying to use this but get the error:
there is no function named "mt_lutxy".

It seems to be that is something from masktools, which I have.
In fact I have tried numerous versions of masktools2, from
incl:
masktools-v2.0a
masktools-v2.0a35
masktools-v2.0a44
masktools-v2.0a48
can someone tell me which version to use specifically (if it matters)
and if that isn't the issue, please, will someone help me out. This looks like a promising tool, but this seems to be some strange error people get randomly with different scripts, and all people usually say is to "have masktools" (which not so helpful with so many builds, and less helpful if people do have mt2).

[Bloax]

Well that's strange, because it definitely works with a48.

Might be another case of Avisynth bugging out hard with plugin loading, because I have witnessed that before. (The solution being long forgotten also.)

[djmasturbeat]

I guess I can try a fresh install of Avisynth
may be I screwed smth up in there, but otherwise I don't get a lot of errors in AVSPm,
I am not too experienced of an encoder nor with filters and avs, so generally I use very few filters to minimize self-inflicted wounds, tho I do on occasion try some out (mostly in testing and tinkering, as I am not experienced enough to know what all the settings are in some of these big mega avsi files).

I guess I will see what happens from fresh avs install, and me backing up scripts and plugins elsewhere.

[@mFY]

how do I use it?it has dll?

[Didée]

@ djmasturbeat: You're sure to use the correct masktools-dll for your Avisynth version?

Avisynth v2.5.8 --> mt_masktools-25.dll
Avisynth v2.6 ----> mt_masktools-26.dll

The issue of "there is no function named XYZ" is known to appear when the -26.dll is loaded under Avisynth 2.5.8.

[jmac698]

Didee, in the german board you posted some years ago shots of VHS, one was a young woman in a uniform, another had train tracks.. but it looked amazingly sharper. What sharpening script was that? I thought it was a miracle

[Didée]

If it was years ago, then maybe it was the iiP dinosaur. But I really can't remember what pictures you're talking about.