fractal sharpening ?

[halsboss]

http://actions.home.att.net/dSLR_Fractal_Sharpen.html

well, well, is there a similar plugin for avisynth ?

[Didée]

My actions, by contrast, use Genuine Fractals in a unique way to actually "see through" the fuzziness and create an image that's truly sharper than the original.
Hmmh ... I've a vague idea what the basic concept of the "unique way" actually is ...


Note: Because these actions work on upsampled images, they can [use a large amount of computer resources and] take a few minutes to complete, depending on the size of your original image. Be patient. Good things come to those who wait!
So, after several minutes: http://img179.imageshack.us/img179/847/fractalsharpenduckqx5.th.jpg (http://img179.imageshack.us/my.php?image=fractalsharpenduckqx5.jpg)


After fractions of a second: http://img161.imageshack.us/img161/6416/duckseesawsm3.th.jpg (http://img161.imageshack.us/my.php?image=duckseesawsm3.jpg)


Not quite the same, sure, but also ~100 times faster ...
If "a few minutes for a single frame" is acceptable, then I'm pretty sure that SeeSaw could be improved accordingly. :p

[jeffy]

Didée, did you use some special settings? Some prefiltering/MVAnalyzes? Could you please show us the script?
Thank you.

[*.mp4 guy]

Crop, convert to yv12, unsharpmask
ImageReader("fractalsharpenduckqx5.jpg")
crop(24, 23, -556, -24).converttoyv12
unsharp(vary=3) (http://img186.imageshack.us/img186/9449/unsharpradius3bn6.png)

Crop, convert to yv12, unsharpmask, with frequency limited limited sharpen faster:
ImageReader("fractalsharpenduckqx5.jpg")
crop(24, 23, -556, -24).converttoyv12
unsharp(vary=3)
s = last
sharp = SeeSaw(s,s).SeeSaw().SeeSaw().SeeSaw().SeeSaw()
diff = mt_makediff(sharp, last).spline36resize(round(width*1.5/8)*8, round(height*1.5/8)*8).dctlimit(use64dct=false, use32dct=false, use8dct=true, dc=1, dct2=0, dct3=0.25, dct4=1, dct5=1, dct6=1, dct7=1, dct8=1).spline36resize(s.width, s.height)
mt_adddiff(last, diff) (http://img186.imageshack.us/img186/7518/frequencylimitedlsfii7.png)

Crop, convert to yv12, unsharpmask, with frequency limited SeeSaw:
ImageReader("fractalsharpenduckqx5.jpg")
crop(24, 23, -556, -24).converttoyv12
unsharp(vary=3)
s = last
sharp = limitedsharpenfaster(smode=4, strength=4000).limitedsharpenfaster(smode=4, strength=4000).limitedsharpenfaster(smode=4, strength=4000)
diff = mt_makediff(sharp, last).spline36resize(round(width*1.5/8)*8, round(height*1.5/8)*8).dctlimit(use64dct=false, use32dct=false, use8dct=true, dc=1, dct2=0, dct3=0.25, dct4=1, dct5=1, dct6=1, dct7=1, dct8=1).spline36resize(s.width, s.height)
mt_adddiff(last, diff) (http://img186.imageshack.us/img186/9720/frequencylimitedseesawkf9.png)

Crop, convert to yv12, unsharpmask, modified LSF, SSSharp (http://img72.imageshack.us/img72/2701/modifiedlsfandsssharpqo9.png)

[Didée]

If that isn't entertainment, what else is.


@jeffy: No, no prefiltering was used. (Though that would've been a good thing , especially deblockig...)

For the settings, iirc the kids riding the seesaw were Ninetynine and Thirtythree:

SeeSaw(Spower=(99+33)/33, Slimit=99, SdampLo=.99, SdampHi=33, Szp=33, bias=33, sootheS=33, sootheT=33-33)

[Soulhunter]

Its funny what they consider as "Already sharp" huh? ;D

[morsa]

BTW, this makes me wonder:

Is there anywhere a complete list of SeeSaw parameter, what they do, the max and min values and a brief explanation of how to use them?

It is not in Avisynth wiki.

[ericab]

is this something we'll ever see in ffdshow ?

[Leak]

is this something we'll ever see in ffdshow ?
A filter that took several minutes for one of the images Didé eposted above?

No, I really don't think so...

np: John Tejada - What Happened To Manners? (Cleaning Sounds Is A Filthy Business)

[ericab]

After fractions of a second: http://img161.imageshack.us/img161/6416/duckseesawsm3.th.jpg (http://img161.imageshack.us/my.php?image=duckseesawsm3.jpg)


so, this didnt take 'fractions of a second' ??

[45tripp]

SeeSaw

search seesaw use

[Dark Shikari]

so, this didnt take 'fractions of a second' ??That is SeeSaw/etc, not fractal sharpening.