PVR deinterlacing - Intel
#46
(2018-07-20, 22:27)FernetMenta Wrote:
(2018-07-20, 21:40)nhsman Wrote:
(2018-07-19, 07:32)FernetMenta Wrote: sw deinterlacing should look the same. In order not to compare apples and oranges, can we agree to do all further tests on Linux? Just files, no tv headend.

Please post a screenshot of the same scene.   
Hi, I post it when I get a minute. But having sw deinterlacing is not the point for me. I need hw deinterlacing working properly due to high cpu usage with sw decoding enabled. I just quickly compared hw and sw deinterlacing on linux and sw looks much better than hw but still not perfect.    
 All my questions have a single purpose: spotting the issue. Even if sw deinterlacing is not important to you, a correct answer is important to narrow down the problem. If sw deinterlacing does not work, we need to search in a totally different area than just vaapi does not work.  
Here is the screenshot done on SW deinterlacing. It looks similar to yours so I would guess that SW deinterlacing works.

Image
Reply
#47
Would it be end of support for this issue?
Reply
#48
And now: how does MCDI look different to this screenshot? Can you make another screenshot?
First decide what functions / features you expect from a system. Then decide for the hardware. Don't waste your money on crap.
Reply
#49
Please zoom it or display on the big screen and look at the edges of her face or hair, there are some artifacts not present on previous screenshot.

Image


Here is better example of what I mean, please zoom it a little and look at the man or the boy, it looks just partially deinterlaced:

Image
Reply
#50
Does it get better if you choose "bilinear" instead of Lanczos3? (it should get more blurry but are the "rough" edges gone?)
First decide what functions / features you expect from a system. Then decide for the hardware. Don't waste your money on crap.
Reply
#51
(2018-07-25, 22:00)fritsch Wrote: Does it get better if you choose "bilinear" instead of Lanczos3? (it should get more blurray but are the "rough" edges gone?)
Just tried bilinear - same artifacts, softer picture:

Image
Reply
#52
(2018-07-25, 22:46)nhsman Wrote:
(2018-07-25, 22:00)fritsch Wrote: Does it get better if you choose "bilinear" instead of Lanczos3? (it should get more blurray but are the "rough" edges gone?)
Just tried bilinear - same artifacts, softer picture:

Image 
 And finally: VAAPI + "Deinterlace" <- how does that look like? (I'd like to see in which states those "artifacts" are introduced. For VAAPI + Deinterlace a different code path is used to "copy" the decoded pics and deinterlace them in SW).
First decide what functions / features you expect from a system. Then decide for the hardware. Don't waste your money on crap.
Reply
#53
Hi Here is the screenshot taken with VAAPI + "Deinterlace" enabled:

Image

I would say it's less artifacts but the picture is a little bit soft.
Reply
#54
I am not sure about what's happening here. Where do those pixel errors come from. Maybe we have a lot of those and upscaling makes is worse.


Image
Reply
#55
Shader precision issue? Edit: More concrete, this looks like something with sampling goes wrong. Texture lookup? Lookup surface?
First decide what functions / features you expect from a system. Then decide for the hardware. Don't waste your money on crap.
Reply
#56
(2018-07-26, 20:22)fritsch Wrote: Shader precision issue? Edit: More concrete, this looks like something with sampling goes wrong. Texture lookup? Lookup surface?
 Looks more like a decoder issue caused by missing refs or the like. Missing refs can cause quality issues bound to regions of a frame while deinterlacing issues would show on the entire frame.
Reply
#57
What I wonder is, that nobody reported an issue with progressive content yet.
First decide what functions / features you expect from a system. Then decide for the hardware. Don't waste your money on crap.
Reply

Logout Mark Read Team Forum Stats Members Help
PVR deinterlacing - Intel1