Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Hardware & Software > Software players

Reply
 
Thread Tools Search this Thread Display Modes
Old 12th May 2015, 08:11   #81  |  Link
Warner306
Registered User
 
Join Date: Dec 2014
Posts: 1,127
If Upscaling Refinement shaders are added to the post-resized image, why do they only impact videos that are sub 1080p?
Warner306 is offline   Reply With Quote
Old 12th May 2015, 08:28   #82  |  Link
Asmodian
Registered User
 
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,047
Quote:
Originally Posted by Warner306 View Post
If Upscaling Refinement shaders are added to the post-resized image, why do they only impact videos that are sub 1080p?
I assume you are watching on a 1920x1080 display? They apply for me watching 1080p on a 2560x1440 screen. Think of them as part of upscaling and upscaling doesn't run when viewing 1080p at 1920x1080.
__________________
madVR options explained
Asmodian is online now   Reply With Quote
Old 12th May 2015, 09:35   #83  |  Link
Warner306
Registered User
 
Join Date: Dec 2014
Posts: 1,127
Ok, one more question. Is SuperRes a form of image doubling, or can it do irregular upscales?
Warner306 is offline   Reply With Quote
Old 12th May 2015, 09:46   #84  |  Link
Shiandow
Registered User
 
Join Date: Dec 2013
Posts: 753
It can handle arbitrary upscales.
Shiandow is offline   Reply With Quote
Old 12th May 2015, 10:26   #85  |  Link
Asmodian
Registered User
 
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,047
I understand SuperRes is more of a post process method than an independent scaler; it runs along with or on top of the other scaling option chosen, it doesn't replace it.
__________________
madVR options explained
Asmodian is online now   Reply With Quote
Old 12th May 2015, 10:38   #86  |  Link
Shiandow
Registered User
 
Join Date: Dec 2013
Posts: 753
Quote:
Originally Posted by Asmodian View Post
I understand SuperRes is more of a post process method than an independent scaler; it runs along with or on top of the other scaling option chosen, it doesn't replace it.
That's more or less correct, although it will still give reasonable results even if you use a really simple scaler such as nearest neighbour, or even if you just feed it a black image. But if you start with a better image you need less iterations, so it's a bit of a trade off.
Shiandow is offline   Reply With Quote
Old 12th May 2015, 10:42   #87  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,027
Personally I'm still not sure about SuperRes. On high quality content, the difference it makes is very minimal, and on low quality content it tends to sharpen the artifacts more than I would like.
__________________
LAV Filters - open source ffmpeg based media splitter and decoders
nevcairiel is offline   Reply With Quote
Old 12th May 2015, 11:21   #88  |  Link
Asmodian
Registered User
 
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,047
I do think I like SuperRes for chroma. The minimal difference on high quality content seems like a good difference as well, at least with animated content.
__________________
madVR options explained

Last edited by Asmodian; 12th May 2015 at 11:32.
Asmodian is online now   Reply With Quote
Old 12th May 2015, 15:33   #89  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,477
Quote:
Originally Posted by Asmodian View Post
SuperRes: This is a post process method. From Shiandow
Sweeet, thanks for digging the link for us as I was really looking forward to explanations regarding this feature because to my eyes NEDI+SuperRes can really match/exceed NNEDI3's PQ but I'm still finetuning settings atm
leeperry is offline   Reply With Quote
Old 12th May 2015, 15:58   #90  |  Link
Shiandow
Registered User
 
Join Date: Dec 2013
Posts: 753
There have been some changes to the algorithm though, although the general idea has remained more or less the same. The steps now look more like the following:
  1. Calculate an initial guess (using a different upscaler)
  2. Downscale and calculate differences with original image.
  3. Scale those differences to the final size
  4. Improve guess by:
    • Softening the image
    • Subtracting differences with the original image
    • Sharpening (optional)
    • Removing aliasing (optional)
    • Removing ringing (optional)
  5. Repeat steps 2-4 several times.
Shiandow is offline   Reply With Quote
Old 12th May 2015, 20:46   #91  |  Link
Warner306
Registered User
 
Join Date: Dec 2014
Posts: 1,127
Quote:
Originally Posted by Asmodian View Post
I do think I like SuperRes for chroma. The minimal difference on high quality content seems like a good difference as well, at least with animated content.
How can you tell it's impacting the chroma? I thought it only applied to the luma, or the luma and chroma together.
Warner306 is offline   Reply With Quote
Old 12th May 2015, 20:52   #92  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 6,726
Quote:
Originally Posted by Warner306 View Post
How can you tell it's impacting the chroma? I thought it only applied to the luma, or the luma and chroma together.
because superres can be used for chroma only.
huhn is offline   Reply With Quote
Old 13th May 2015, 03:35   #93  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,477
SuperRes is chroma only?

Anyway, major kudos to Shiandow once again for the NEDI+SuperRes combo because to my eyes it doesn't have the artificial look of NNEDI3, doesn't look oversharpened and looks very natural/realistic AFAICS. motionDSP's fluff doesn't look all that impressive anymore.

Would be fun to know how the combo fares with the samples in this link but some of those SuperRes apps are far from realtime so this might be quite unfair.
leeperry is offline   Reply With Quote
Old 13th May 2015, 03:58   #94  |  Link
Warner306
Registered User
 
Join Date: Dec 2014
Posts: 1,127
Quote:
Originally Posted by leeperry View Post
SuperRes is chroma only?

Anyway, major kudos to Shiandow once again for the NEDI+SuperRes combo because to my eyes it doesn't have the artificial look of NNEDI3, doesn't look oversharpened and looks very natural/realistic AFAICS. motionDSP's fluff doesn't look all that impressive anymore.

Would be fun to know how the combo fares with the samples in this link but some of those SuperRes apps are far from realtime so this might be quite unfair.
Can someone explain the three different SuperRes options?
Warner306 is offline   Reply With Quote
Old 13th May 2015, 03:59   #95  |  Link
Warner306
Registered User
 
Join Date: Dec 2014
Posts: 1,127
Quote:
Originally Posted by Shiandow View Post
There have been some changes to the algorithm though, although the general idea has remained more or less the same. The steps now look more like the following:
  1. Calculate an initial guess (using a different upscaler)
  2. Downscale and calculate differences with original image.
  3. Scale those differences to the final size
  4. Improve guess by:
    • Softening the image
    • Subtracting differences with the original image
    • Sharpening (optional)
    • Removing aliasing (optional)
    • Removing ringing (optional)
  5. Repeat steps 2-4 several times.
This should be added verbatim to the guide.
Warner306 is offline   Reply With Quote
Old 13th May 2015, 04:09   #96  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 6,726
Quote:
Originally Posted by leeperry View Post
SuperRes is chroma only?
you can use it as the chroma resizer too and don't have to resize luma at the same time. so it can be used for chroma only.

i guess this is more accurate.
huhn is offline   Reply With Quote
Old 13th May 2015, 04:29   #97  |  Link
Asmodian
Registered User
 
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,047
Quote:
Originally Posted by leeperry View Post
SuperRes is chroma only?
SuperRes can be chroma only (check the "activate SuperRes filter" box on the chroma upscaling page).

Quote:
Originally Posted by Warner306 View Post
This should be added verbatim to the guide.
Yes, I really wish I had more characters to use so I could.

What sections should/could be removed or pared back?
edit: I squeezed it in, I hope I didn't remove anything important.

Quote:
Originally Posted by Shiandow View Post
There have been some changes to the algorithm though, although the general idea has remained more or less the same. The steps now look more like the following:
  1. Calculate an initial guess (using a different upscaler)
  2. Downscale and calculate differences with original image.
  3. Scale those differences to the final size
  4. Improve guess by:
    • Softening the image
    • Subtracting differences with the original image
    • Sharpening (optional)
    • Removing aliasing (optional)
    • Removing ringing (optional)
  5. Repeat steps 2-4 several times.
Thanks for the excellent and succinct explanation of SuperRes, I am sure many people with appreciate it.

...and all the work you put into the method itself, of course.

__________________
madVR options explained

Last edited by Asmodian; 13th May 2015 at 05:30.
Asmodian is online now   Reply With Quote
Old 14th May 2015, 07:29   #98  |  Link
Warner306
Registered User
 
Join Date: Dec 2014
Posts: 1,127
Can you read the following post? I think your rendering diagram is wrong. In the post below, madshi explains the input for image doubling is "always RGB." This is not reflected in your diagram: http://forum.doom9.org/showpost.php?...ostcount=29697.
Warner306 is offline   Reply With Quote
Old 14th May 2015, 07:43   #99  |  Link
Asmodian
Registered User
 
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,047
But my diagram does have the input for image doubling as always RGB? The "image doubling" madshi is talking about is the dashed orange box.

I leave out that madVR will use any of the Bicubic opitons if you have it set for image upscaling but I assume most users will not have a bicubic set for image upscaling. Otherwise that post agrees with my diagram.
__________________
madVR options explained
Asmodian is online now   Reply With Quote
Old 14th May 2015, 20:39   #100  |  Link
Warner306
Registered User
 
Join Date: Dec 2014
Posts: 1,127
Quote:
Originally Posted by Asmodian View Post
But my diagram does have the input for image doubling as always RGB? The "image doubling" madshi is talking about is the dashed orange box.

I leave out that madVR will use any of the Bicubic opitons if you have it set for image upscaling but I assume most users will not have a bicubic set for image upscaling. Otherwise that post agrees with my diagram.
So you are saying a color space conversion from RGB to YCbCr is a necessary part of image doubling? This is what I questioned based on that post.
Warner306 is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 23:22.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, vBulletin Solutions Inc.