VideoHelp Forum
+ Reply to Thread
Results 1 to 13 of 13
Thread
  1. Member brassplyer's Avatar
    Join Date: Apr 2008
    Location: United States
    Search Comp PM
    Is there such a thing as an app that will minimize loss of image clarity when upscaling/upconverting to a larger size? For example upconverting 16:9 720x480 to HD resolution size?

    Thanks
    Quote Quote  
  2. How are you upscaling now?

    Anything wrong with using nnedi3_rpow2 followed by some sharpening?
    Quote Quote  
  3. Member budwzr's Avatar
    Join Date: Apr 2007
    Location: City Of Angels
    Search Comp PM
    Convert it to 2K, adjust contrast/brightness, downsize to 1080. If that's not good enough, no there's nothing.

    In Vegas, you would set the project to 2K, then render out to 1080. Then change the project to whatever you need. Vegas will automatically scale the preview to project settings
    Quote Quote  
  4. Member brassplyer's Avatar
    Join Date: Apr 2008
    Location: United States
    Search Comp PM
    Originally Posted by manono View Post
    How are you upscaling now?
    Using resize in Virtualdub.

    Anything wrong with using nnedi3_rpow2 followed by some sharpening?
    I don't have the vaguest idea. Never heard of it.

    So, I can rephrase your answer as "try nnedi3_rpow2".
    Quote Quote  
  5. How about this 4:3 DVD frame:

    Name:  dvd.jpg
Views: 153
Size:  37.4 KB

    Upscaled to 1440x1080:

    Click image for larger version

Name:	upscale.jpg
Views:	94
Size:	105.5 KB
ID:	22963

    That was done with nnedi3_rpow2() and a few other filters. Oh sorry, the colors are different because I forgot to use rec.709 to make the upscaled jpg file (I just opened the encoded video with VirtualDub and exported from there).
    Last edited by jagabo; 22nd Jan 2014 at 22:44.
    Quote Quote  
  6. Originally Posted by brassplyer View Post
    So, I can rephrase your answer as "try nnedi3_rpow2".
    Sure. I was just surprised you had never heard of it as it's well-known as the best AviSynth upsizer. Or try Video Enhancer by Infognition. Except it's not all they claim it to be:

    http://www.infognition.com/VideoEnhancer/
    Quote Quote  
  7. Member budwzr's Avatar
    Join Date: Apr 2007
    Location: City Of Angels
    Search Comp PM
    On the upscaled image, I see smoothing. In this case, the upscale became better than the original. But the color palette is limited. What about a 480 still from a DVD?
    Quote Quote  
  8. Originally Posted by budwzr View Post
    What about a 480 still from a DVD?
    That was a 480 still from a DVD and the same frame from the upscaled h.264 video.
    Quote Quote  
  9. Member budwzr's Avatar
    Join Date: Apr 2007
    Location: City Of Angels
    Search Comp PM
    When I set it to 300ppi lot of artifacts appear. Besides, this is the only thread worth posting in.
    Quote Quote  
  10. Member
    Join Date: Mar 2008
    Location: United States
    Search Comp PM
    I found this script on the 'net while I was nosing about in various places. I've never tried to do this before, but it gave surprisingly
    good results. I had a demo 1080p video on my PC (it's a bunch of fish swimming around). I first encoded to DVD using HCenc,
    then recreated 1080 from the dvd using the script (it ran so slow on my PC, I only encoded the first 200 frames)

    Here it is, if anybody's curious. I was unable to find one of the functions, Framecache().
    I just commented it out. It worked, but was extremely slow.

    LoadPlugin("F:\Documents and Settings\Dave Xnet\My Documents\Downloads\dgmpgdec158\DGDecode.dll")
    DGDecode_mpeg2source("F:\Documents and Settings\Dave Xnet\Desktop\movies _temp\samsung.d2v", info=3)

    degrainmv(200)
    nnedi3_rpow2(rfactor=2,cshift="spline64resize",psc rn=false)
    spline64resize(1440,1080,0,5,0,0)
    ColorMatrix(mode="Rec.601->Rec.709")
    noise()

    function noise(clip clp)
    {
    m1 = mt_luts(clp,clp,"avg",mt_square(1),"y x 0.98 * < 0 y x 1.02 * > 0 255 ? ?",u=1,v=1)
    cgrain = blankclip(clp,color_yuv=$808080).addgrainC(0,10,0, 0).temporalsoften(1,0,255,255,2)
    grain = clp.grainfactory3(g1str=20,g2str=30,g3str=50,temp_ avg=100).mt_makediff(cgrain,y=2,u=3,v=3)
    AGmask = m1.mt_inpand().removegrain(19,-1).mt_lutxy(clp,"x y 200 + 455 / 0.5 ^ *",u=1,v=1)
    mt_merge(clp,grain,AGmask,luma=true)
    }

    function degrainmv(clip clp,int "thSAD"){
    thSAD = default(thSAD,150)
    clp
    framecache(7)
    sclip = msuper()
    vf1 = manalyse(sclip,isb=false,delta=1,overlap=4,blksize =8,search=5,pelsearch=2,truemotion=true,sadx264=0)
    vb1 = manalyse(sclip,isb=true,delta=1,overlap=4,blksize= 8,search=5,pelsearch=2,truemotion=true,sadx264=0)
    vf2 = manalyse(sclip,isb=false,delta=2,overlap=4,blksize =8,search=5,pelsearch=2,truemotion=true,sadx264=0)
    vb2 = manalyse(sclip,isb=true,delta=2,overlap=4,blksize= 8,search=5,pelsearch=2,truemotion=true,sadx264=0)
    vf3 = manalyse(sclip,isb=false,delta=3,overlap=4,blksize =8,search=5,pelsearch=2,truemotion=true,sadx264=0)
    vb3 = manalyse(sclip,isb=true,delta=3,overlap=4,blksize= 8,search=5,pelsearch=2,truemotion=true,sadx264=0)
    mdegrain3(clp,sclip,vb1,vf1,vb2,vf2,vb3,vf3,thSAD= thSAD)
    mt_merge(clp,last,clp.blur(1.5).mt_lut("x 80 > 255 x 3 * ?"),luma=true)
    }
    Quote Quote  
  11. Degrain followed by added noise? I don't get it. Added noise alone sure, but why denoise then.

    And can't that Spline64Resize call go directly into NNEDI for the proper centering at the same time as the final resize?
    Quote Quote  
  12. Member
    Join Date: Mar 2008
    Location: United States
    Search Comp PM
    I have not analyzed the script yet, but I believe it is supposed to compensate for too much smoothness being introduced.
    Here's some of the output, enlarged from 720x480 DVD, and the x264 settings.

    x264.exe" --preset veryfast --crf 18 --level 4 --keyint 100 --ref 3 --bframes 2 --vbv-bufsize 25000 --vbv-maxrate 20000 --sar 4:3

    Also, the source m2v.
    Attached Files
    Last edited by davexnet; 23rd Jan 2014 at 16:29.
    Quote Quote  
  13. Originally Posted by davexnet View Post
    I believe it is supposed to compensate for too much smoothness being introduced.
    Yes, it removes grain at the original frame size (so it doesn't get bigger), then adds grain back at the end to reduce posterization artifacts.
    Quote Quote  



Similar Threads