HD 1080i Test Pattern to determine Vector Adaptive Deinterlacing + others icl. Ticker - Page 10 - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #271 of 283 Old 01-23-2013, 04:17 AM
AVS Special Member
 
dr1394's Avatar
 
Join Date: May 2002
Location: Mizar 5
Posts: 3,140
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 19
Quote:
Originally Posted by madshi View Post

Is there a specification which defines that? Or an EBU recommendation or something? Or is that just your personal experience?

In my experience at least 99% of all PAL content has the correct fields to weave in the same frame. But to every rule there is an exception. As I said, I have a sample of a PAL DVD where the correct fields to weave together are *not* in the same frame, similar to the original Cheese Slices. Also, an IVTC algorithm which claims to support "any cadence" (e.g. for Anime sources) has to be very flexible, and should have no problem finding the correct fields to weave, even if they're spread over two frames, like with the original Cheese Slices. With hard-telecined 3:2 content it happens all the time that the correct fields to weave together are not in the same frame. So supporting this for PAL, too, is an automatic side effect of an "any cadence" IVTC algorithm.

I have no problem with that. My personal experience is the same, 99%. It's just that using the term 2:2 pulldown to describe this sequence is referring to the pathological 1%, and that's why it's confusing. I can definitely see where an IVTC algorithm could find repeat fields somewhere in this sequence (although the fields are not 100% repeats in my example at around frame 140 above because of the horizontal lines).

All I'm saying is that the repeat fields are just an artifact of a mistake in how the pattern was fabricated as a collection of interlaced frames. By having no motion from bottom field to the top field of the next frame, the pattern is depicting motion that is not natural. That is, it can't happen in video of the real world.

I have to agree with madshi that the pattern is flawed.

However, I'd like to see us all work together on this rather than pointing fingers.

Ron

HD MPEG-2 Test Patterns http://www.w6rz.net
dr1394 is offline  
Sponsored Links
Advertisement
 
post #272 of 283 Old 01-23-2013, 05:16 AM
AVS Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 5,420
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 19 Post(s)
Liked: 113
Quote:
Originally Posted by dr1394 View Post

although the fields are not 100% repeats in my example at around frame 140 above because of the horizontal lines

Are you sure? I've only done a quick check, but I'm pretty sure that weaving the correct fields together produces absolutely perfect results, also around frame 140. At least I can't see any sort of artifacts when playing this back with madVR in forced film mode. No double horizontal lines, no missing horizontal lines, no weaving artifacts, no softness, nothing.

Quote:
Originally Posted by dr1394 View Post

However, I'd like to see us all work together on this rather than pointing fingers.

FWIW, blaubart already created a fixed pattern with true video mode interlaced encoding here:

http://www.avsforum.com/t/1157287/hd-1080i-test-pattern-to-determine-vector-adaptive-deinterlacing-others-icl-ticker/240#post_22850842
madshi is offline  
post #273 of 283 Old 01-23-2013, 10:01 AM - Thread Starter
Member
 
blaubart's Avatar
 
Join Date: Apr 2009
Posts: 161
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by madshi View Post

That is incorrect. It works with any *software* decoder. It currently just doesn't work with native DXVA decoding, as I already told you before.

Ok, you got me, I do not use *software* decoders - senseless on most of my computers even on a 'earlier' i7 (as I already told you before).


Quote:
That is not true. I said that the original Cheese Slices should not be used to judge video mode deinterlacing quality, but that they might still be useful for other purposes. You actually agreed with me there.

Hey madshi don't do so as if you didn't know how the internet (and contracts and politics) work.
- At first you say in grandiloquent triades your wish: 'I'm so sorry but Cheese Clices is crap'. (post#233, 242)
- Then after many many words at a point you can be sure that everybody's meanwile sleeping you bring the things that do not suit your plans, the *fine print without blank lines*: 'oh, may be it's still useful BUT... ... ...'

Believe me everybody who reads this a little bit faster will keep his hands off Cheese Slices at all events! And be sure now his onboard Ati HD 3200/3300 is after all of course playing the VA deinterlacing...



Quote:
And you said nobody ever did use the Cheese Slices to judge video mode deinterlacing quality, anyway. Then I pointed you to AnandTech doing it in the past. Then you said, ooops, now you remember, but you always said they used the Cheese Slices wrong.

I really had forgotten this 'An HTPC Perspective' AnandTech article at that time. And you said >The Cheese Slices have been used by many people to judge the quality of video mode deinterlacers.< so I wondered where are so many people using Cheese Slices outside of just checking CCC's deinterlacing options using DXVA??

In the first 'AMD’s Radeon HD 5450' AnandTech article you linked on they did just what Cheese Clices is able to do and then you say
>All of that is factually wrong. It is not heavily interlaced, it's just telecined, which is a relatively mild form of interlacing. It doesn't have *any* interlacing artifacts in it.<

Are you dreaming or what? Only cause you found a trick to progressive it in madVR - it's not weak deinterlacing at all! Today's graphics cards are more powerful thats all! But earlier graphics cards had no chance with it!

Artefacts artefacts what do you mean with artefacts? Aren't that interlacing artefacts in my screenshots above -> you see with any today's graphics card using MotionAdaptive and VectorAdaptive with Cheese Clices using D.X.V.A?

Who the hell is talking about *software* deintelacing? Nobody! Only the CPU/GPU-heating madVR freaks!



Quote:
Furthermore after having compared telecined and video Cheese Slices DXVA deinterlacing results, I wrote:
> Generally, "video" results are quite similar to the "telecined" results, at least
> with my ATI card, which is probably good news because it means that the
> old Cheese Slice tests are still "somewhat" valid.

Which directly contradicts what you just said.

"probably" "somewhat" too late my dear, at that time everybody fast-reading was sure: Cheese Slices is crap..smile.gif



Quote:
Am I? How about this:

http://madshi.net/cheese/filmVA2.png
http://madshi.net/cheese/filmA2.png

The first is Vector Adaptive Deinterlacing, the second is just Adaptive Deinterlacing (which in theory is much lower quality). In these screenshots it looks like Adaptive Deinterlacing is better than Vector Adaptive Deinterlacing. But that's not actually true in real life. With the telecined Cheese Slices, Adaptive Deinterlacing produces one good output frame, then one bad output frame etc. If you look at the good frames, they look better than Vector Adaptive Deinterlacing output, but the bad frames look worse. This is one negative side effect of the telecined encoding. This doesn't happen with proper video encoding.

I've already mentioned this before, but it seems you missed it. Or chose to ignore it...
I had been using Cheese Slices several times in the past, they've been in my "test patterns" folder for years. I've not the faintest idea how you came to the conclusion that I didn't try them at all. But then that's only one of many wrong conclusions you came to recently...

Anyway, I'm done replying to you, until you do your homework and learn the deinterlacing & IVTC basics (you really have some catching up to do there) and until you come out of your defensive corner and argue based on facts, instead of posting false claims about my intentions and making fun of madVR.

Yes you are, i.e. dreaming of software deinterlacing, all day and night..smile.gif

"just" Adaptive Deinterlacing is indeed a fine deinterlacing, Cheese Slices points that out not only 'telecined' but also in the new ones.

Quote:
With the telecined Cheese Slices, Adaptive Deinterlacing produces one good output frame, then one bad output frame etc. If you look at the good frames, they look better than Vector Adaptive Deinterlacing output, but the bad frames look worse.

> "the bad frames look worse" - No, see below.
> "Adaptive "produces one good output frame, then one bad output frame etc." - That's true but who cares? That's only of interest taking screenshots, but running Cheese Slices and Speedy Slices the only difference is that in Cheese the (overall identcally hard) rippling of the thicker, diagonal lines are turning the ripples moving up and down. And of course you took the screenshot of a 'clean' frame..rolleyes.gif do you hate Cheese Slices hmmmm..?



-> the lines in the following pics are very tiny so it's important to see them in original size. The browser must not be zoomed in any direction. In Firefox and Internet Explorer just press Ctrl + 0 (zero).
-> if after Ctrl+0 you still see waves instead of 1 pixel lines just click on the pictures. Should open them now in org size.



Now here the screenshots to show VA vs. Adaptive differences and non existing Cheese vs. Speedy differences.
1) Cheese_Adaptive



2) Speedy_Adaptive (odd)



..not much difference to see except the written above. The upper pic shows no horizontal lines in the gray cause it's an even frame.

3) But now the Speedy_VA, full with blur artefacts (AMD, Nvidia). They will come again below..




Now compared to 'real' TV sports.

4) Adaptive:
The yellow arrows show what normally is the only 'easy' thing to see comparing VA to MA, Adaptive and bob, small ripples in diagonal moving lines. Only this to catch in average 1080i-TV is difficult enough.



5) VA: Of course VA is the smoother one. But above the white arrow points to green grass - decide by yourself which one looks more like grass?

blaubart is offline  
post #274 of 283 Old 01-24-2013, 08:22 PM - Thread Starter
Member
 
blaubart's Avatar
 
Join Date: Apr 2009
Posts: 161
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
To show some difficulties painting test patterns:

Deinterlacing disortions should be obvious visible.
So the Cheese Slices souce patterns are moving mostly 45° alternately 2-pixel-right/up and 2-pixel-left/down, 20 times per square, each 2nd pic moving. This angle and speed shows (at 25 and 29.97 fps) good weaving (and other deinterlacing) disortion.

See also the org. size part: 2 separate vertical lines, good weaving:




Speed Slices (moving 20 times per square, each pic moving) shows the same weaving disortion but is running too fast, no more weaving (and other details) really good visible.
To avoid the 'telecine effect' at Cheese Slices' "good visible" speed I had to change the movement to 1-pixel-right/up and 1-pixel-left/down, 40 times per square, each pic moving. But the now half as wide weaving disortion is hardly to see.

Org. size part: only 1 vertical line, hardly weaving:




Slighter to no deinterlacing disortions not only in weave but also other deinterlacing settings in other parts of the pattern.

So what to do? Quite simple, I let Photoshop write the 2-pixel-right/up 45° step into four pics instead of two so I have the 40 pics per square, no 'telecine effect' any more.. But too simple, no Chance, this wouldn't work cause each two consecutive pics were (and had to be) completely identical!

The result would be just the same what Cheese Slices does. Switch the odd/even lines of the same pic - then 2-pixel-right/up - then switch the odd/even lines of the same pic - and so on.
Test patterns have their own rules..wink.gif

So I can only double later the length of Speed Slices, sorry. Hope it will help some anti-DXVA people?
blaubart is offline  
post #275 of 283 Old 01-25-2013, 02:05 AM
Member
 
DragonQ's Avatar
 
Join Date: Jan 2007
Posts: 74
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
I don't understand what your objection to MadVR has to do with the cheese slices at all. The simple fact is that the original cheese slice videos aren't interlaced (just progressive video in an interlaced wrapper). They are clearly useful in determining which deinterlacing algorithm is being used, but these versions cannot show the artefacts caused by using the different deinterlacing algorithms on interlaced content. This is obvious from the fact that your newer speedy slices show different effects than the originals.

As long as this is made clear, then there's no problem.
DragonQ is offline  
post #276 of 283 Old 01-25-2013, 03:55 AM - Thread Starter
Member
 
blaubart's Avatar
 
Join Date: Apr 2009
Posts: 161
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
I don't understand what your objection to MadVR has to do with the cheese slices at all.
Just read post#233. Until this point nobody was speaking here about madVR at all.

Quote:
but these versions cannot show the artefacts caused by using the different deinterlacing algorithms on interlaced content.
? What do you mean?

Quote:
This is obvious from the fact that your newer speedy slices show different effects than the originals.

By drawing different kinds of lines on a "white sheet of paper" I determine which artefacts come at which point at which moving speed, moving distance and angle.

But moving speed is not everything at all. It is depending on algorithms I didn't write but examine respectively react on them. So the deinterlacing artefacts may also vanish at 2x or half speed or being doubled on some line arrays or may be something quite different happens or no changes. Same with different moving distances and angles.
.
blaubart is offline  
post #277 of 283 Old 01-25-2013, 04:42 AM
AVS Special Member
 
dr1394's Avatar
 
Join Date: May 2002
Location: Mizar 5
Posts: 3,140
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 19
Quote:
Originally Posted by blaubart View Post

So the de-interlacing artifacts may also vanish at 2x or half speed or being doubled on some line arrays or may be something quite different or nothing happens. Same with different moving distances and angles.
.

I guess you're saying that one pixel of motion per field isn't enough but two pixels are too much? In natural interlaced content, 2 pixels seems like not a lot of motion. In this random frame of some golf, the field motion is at least 24 pixels (you can see three 8x8 DCT blocks side by side in the zoom).

golf.png

Zoom of caddies shoulder.

golfzoom.png

Ron

HD MPEG-2 Test Patterns http://www.w6rz.net
dr1394 is offline  
post #278 of 283 Old 01-25-2013, 05:40 AM - Thread Starter
Member
 
blaubart's Avatar
 
Join Date: Apr 2009
Posts: 161
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
I guess you're saying that one pixel of motion per field isn't enough but two pixels are too much?

No, what i pointed out in a weave example further up in post#274 is: one pixel of motion per field isn't enough but two pixels are enough to be good visible in a vertical line array at 29.97 fps.

But additionally the pattern has to feature different deinterlacing disortion also in VA, MA, Adaptive and bob. Cheese Slices as interacting compromise is a collection of many different deinterlacing indicators to show all of them (and more like skip field) in just one single pattern.

p.s.
If I wanted to do a pattern only for weave I would just let a fat object like the caddy in your example move 24 pixels left or right and 1 pixel up to show the weaving better.
blaubart is offline  
post #279 of 283 Old 01-26-2013, 09:38 AM - Thread Starter
Member
 
blaubart's Avatar
 
Join Date: Apr 2009
Posts: 161
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Finished now rendering the new Slices...

Cheese Slices (nothing changed, only to remember - post#1)
-> the souce patterns are moving mostly 45° alternately 2-pixel-right/up and 2-pixel-left/down, 20 times per square, each 2nd pic moving.

'Each 2nd pic moving' is the best compromise to show distinct differences at all AMD's deinterlacing options: VA, MA, Adaptive, bob and weave and other graphics cards' deinterlacing.
Any 'telecine' effect was not my intention, it's only a result of the required 2 pixel jumps at rel. slow speed (rel. to 29.97fps) for good visibility.

-> the lines in the following pics are very tiny so it's important to see them in original size. The browser must not be zoomed in any direction. In Firefox and Internet Explorer just press Ctrl + 0 (zero). -> if after Ctrl+0 you still see waves instead of 1 pixel lines just click on the pictures. Should open them now in org size.





By special order here now 2 experimental "each picture moving" versions, 1-pixel and 2-pixel:

mpeg2_29.97_1pixel_each-moving_no-ticker
-> the souce patterns are moving 45° alternately 1-pixel-right/up and 1-pixel-left/down, 40 times per square, each pic moving.
No ticker, sorry, this would be huge work... (+ a bit buggy, see post #281)





and the 2-pixel 'Speedy Slices' now as long version:

mpeg2_29.97_2pixels_each-moving_Speedy_Slices
-> the souce patterns are moving mostly 45° alternately 2-pixel-right/up and 2-pixel-left/down, 20 times per square, each pic moving.




To remind - all line arrays were designed for the Cheese Slices moves. Changing them in the experimentals sometimes does but not has to result in a logical reaction e.g. new or missing deinterlacing artefacts or no changes. It all depends on the different algorithm's used by VA, MA, Adaptive, bob and weave. Their reaction on speed, jumps, different lines, antialiasing, points, noise etc. may be sometimes confusing as I had to notice drawing the patterns.

I only mention this for better understanding your own tests outside DXVA. Using DXVA I could not see strange things happen with the experimentals:

- narrower 1pixel weaving artefacts are easy to explain (post#274)
- new, wider or missing VA, MA moving artefacts at some line arrays compared to Cheese Slices are results of the different speed and jumps (post#258)
- narrower deinterlacing shadow in Speedy's 'Fluxus' segment is a result of less time to calculate it (post#258)
- also the GPU-usage playing them thru DXVA is nearly the same.

Same things take effect running all Slices thru madVR software deinterlaced by 'forced film-mode' (inverse telecine) deactivated (if CPU+GPU are fast enough). By activating 'forced film-mode' the experimentals are weave deinterlaced. The classic Cheese Slices are 'progressived' in this way.

Howsoever, if you find sth. special on different setups or things will change in the future please leave a note!


At the beginning I recommend testing on DVI computer monitors.
TV's via HDMI mostly try to use their picture enhancements, this could easily cause different kinds of extra distortion.

I tested using MPC-HC + LAV splitter + EVR/madVR + DXVA + these Video decoders:
PDVD 10/12
MS DTV-DVD Decoder
LAV Video decoder -> (here on AMD graphics MPEG2 @ 'DXVA2 (native)' is only supported on 'HD 6xxx' series and higher)
LAV Video decoder -> to test madVR's 'forced film-mode' set LAV to 'DXVA2 (copy back)' (complies software deinterlacing)

About many software deinterlacing 'specialities' I don't know much, but trying my Core i7 720QM + HD5850 M anyway cannot run 1080i judder free, also not a i5 2450M (Intel HD3000), only the i7 2600 + HIS Radeon HD 6970 IceQ Turbo in my gaming PC are fast enough. So it's up to you..?
blaubart is offline  
post #280 of 283 Old 01-27-2013, 12:40 AM
Member
 
DragonQ's Avatar
 
Join Date: Jan 2007
Posts: 74
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by blaubart View Post

? What do you mean?
It's been explained several times. If there isn't movement between every field then the content isn't interlaced, simple as that. It's just progressive video in an interlaced stream.

Your new Speedy Slices are interlaced. Strangely enough though, on an Intel HD4000, every other frame still looks better than the previous frame in certain parts of the image at least.
DragonQ is offline  
post #281 of 283 Old 01-27-2013, 10:49 AM - Thread Starter
Member
 
blaubart's Avatar
 
Join Date: Apr 2009
Posts: 161
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
It's just progressive video in an interlaced stream.
.. and the deinterlacer has the full work to do as if it were interlaced video if DXVA is used. If only every explanation here would have got this..rolleyes.gif



Additional remark to the middle experimemtal above, the 1-pixel moving without ticker:

Advancing single frames in the moving parts by regarding the lower horiz. grey lines and the green sports field - it is always sharp in one frame and a bit difused, unsharp in the next frame. This is obviously caused by TMPGEnc rendering 45° 1-pixel moves because rendereing into single pictures it's the same. Tried many things now, couldn't get it better.
.
blaubart is offline  
post #282 of 283 Old 01-30-2013, 10:14 AM
Member
 
DragonQ's Avatar
 
Join Date: Jan 2007
Posts: 74
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by blaubart View Post

.. and the deinterlacer has the full work to do as if it were interlaced video if DXVA is used. If only every explanation here would have got this..rolleyes.gif
No, you are wrong. If that were true, the results for the interlaced version would be the same as the telecined version. They aren't.
DragonQ is offline  
post #283 of 283 Old 02-01-2013, 05:31 PM - Thread Starter
Member
 
blaubart's Avatar
 
Join Date: Apr 2009
Posts: 161
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
..that simple? Only because you annoying your master in another forum I have to pay for?
blaubart is offline  
Reply Home Theater Computers

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

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
Trackbacks are Off
Pingbacks are Off
Refbacks are Off