117 replies to this topic

[DavoNogo]

Ikari, on Apr 22 2006, 10:08 PM, said:

To sum it up I would say that weave & downscale of a 1080i25p source<snip>

That's a pretty magical source, if I don't say so myself...

[Champion_R]

Nobody has mentioned Motion Compensated deinterlacing in this thread.

[Trainspotter]

Champion_R, on Apr 23 2006, 03:15 AM, said:

Nobody has mentioned Motion Compensated deinterlacing in this thread.

That's because they were discussing bob vs weave with film material (where I believe motion adaptive deinterlacing is of no value - it’s only of use for interlaced video).

Darklord did mention motion adaptive in his original article:

Quote

Motion adaptive based de-interlacing

This is the most sophisticated type of de-interlacing and as a result is very processor intensive, requiring expensive dedicated chipsets. As such it is currently rare on any HD displays (and even external processors for that matter). Some displays and processors can apply this type of de-interlacing to standard definition sources (480i or 576i) but very few can use it with high definition 1080i. This is expected to change over the next few years.

So how does motion adaptive de-interlacing work? Motion adaptive de-interlacing analyses several successive fields (the more fields it analyses the better it generally is) and looks carefully at differences between those fields. For parts of the image that are motionless or relatively still, it will weave together sections of those successive fields, and for parts of the fields that are under motion, it will use interpolation to fill in the gaps. Think of it as a combination of weave and bob de-interlacing.

If done properly this can result in better resolution (as you can get close to the full vertical resolution when there is little movement) but the results vary wildly depending on how sophisticated the motion adaptive de-interlacing is. The best motion adaptive de-interlacing chipsets are generally ones that can de-interlace on a “per pixel” level which means they are powerful enough to look at differences between fields on an individual pixel basis.

So that about covers the basics of de-interlacing. There are other less common methods of de-interlacing available such as vector based de-interlacing, but for the sake of simplicity I’ve chosen to keep them out of this article.

[pneu]

DavoNogo, on Apr 23 2006, 02:22 AM, said:

That's a pretty magical source, if I don't say so myself...

lol..you know what I mean, 1080i50 thats come from a 25p source, both fields correspond to the same image etc.

btw Davo, you wouldn't happen to have a couple of screenshots of Sunrise would you?  I'd like to do some photoshop jiggery-pokery and see if interlacing them into 1080p results in anything decent
It'd have to be 2 consecutive 576p screen captures with as little movement in between frames as possible, maybe a studio shot with a still background?  Have you got anything like that?

[DavoNogo]

Ikari, on Apr 23 2006, 05:19 AM, said:

btw Davo, you wouldn't happen to have a couple of screenshots of Sunrise would you?  I'd like to do some photoshop jiggery-pokery and see if interlacing them into 1080p results in anything decent
It'd have to be 2 consecutive 576p screen captures with as little movement in between frames as possible, maybe a studio shot with a still background?  Have you got anything like that?

I do, actually...

http://img105.images...sunrise10de.png
http://img105.images...sunrise23yn.png

It was just a random capture I made a few months ago... there wasn't much footage to go by, and there certainly weren't any scenes with zero movement... though you can still try and focus on the static parts like the coffee mug and all that...

I'm telling you this now though, that you'll be wasting your time.. Seven don't just do a simple bob, however on SD sources, you can clearly see aliasing and a softer, blurrier image than on the SD channel... it actually reminds me of some progressive scan DVD players I've analysed.

[madmax]

Darklord, on Apr 18 2006, 01:09 AM, said:

LMAO! That’s your proof? That’s your example? Yep, bob sure looks great mate

ROFLMAO!       Geez, can't wait for Blu-Ray so I don't have to worry about any of this....just have to remember to buy a compatible display.......

[pneu]

Here's the end result:

Sunrise 1080i

Oh and Davo, I believe you when you say standard def shows look worse on LCN70, but do you think it's because they're using crude bob instead of interpolation?

Here's the example from before except using crude bob.  I gotta say crude bob is starting to look like "half res"

[pneu]

So what do you think guys, does it look anything like high def?  I'm not a very good judge of high def as I've only seen it demo'd in shops

Keep in mind the horizontal res is really hurting at 720 pixels.  If it was 1440 it could look a little sharper, I think.  The fields match up pretty well though - here's a shot of the 2 frames interleaved and field 2 shifted to the right to give an idea of how much of the original fields remain in the ED version.

fields

[norpus]

Ikari, on Apr 25 2006, 11:51 AM, said:

So what do you think guys, does it look anything like high def?  I'm not a very good judge of high def as I've only seen it demo'd in shops

Keep in mind the horizontal res is really hurting at 720 pixels.  If it was 1440 it could look a little sharper, I think.  The fields match up pretty well though - here's a shot of the 2 frames interleaved and field 2 shifted to the right to give an idea of how much of the original fields remain in the ED version.

fields

The lettering in your example is worse - it is skewed
If you hadn't changed the shape from 4:3 to 16:9, it might have been easier to tell any diff.

[pneu]

The lettering?  Do you mean the news banner at the bottom having mice teeth?  There's no way of avoiding that as the 2 fields have come from 2 different points in time, it would look the same if the 2 fields had been captured straight from the 1080i50 original version.

Here's the same image in anamorphic format if you think it looks better quality that way.

[DavoNogo]

norpus, on Apr 25 2006, 12:49 PM, said:

The lettering in your example is worse - it is skewed

Can't be avoided as the news ticker thing is ALWAYS moving...

[norpus]

Ikari, on Apr 25 2006, 03:46 PM, said:

The lettering?  Do you mean the news banner at the bottom having mice teeth?  There's no way of avoiding that as the 2 fields have come from 2 different points in time, it would look the same if the 2 fields had been captured straight from the 1080i50 original version.

Here's the same image in anamorphic format if you think it looks better quality that way.

Ok understood now, thanks

[Trainspotter]

There is an interesting thread at AVS about a ‘Perfect Vision’ magazine article that confirmed after proper testing that many TVs throw away 50% of 1080i resolution. http://www.avsforum....threadid=608670

After reading this there is no doubt in my mind that many TVs discard 50% resolution using bob. How anyone could doubt it I don’t know. I'm sure Ikari will still find some way to disagree

[pneu]

Trainspotter, on Apr 28 2006, 03:47 PM, said:

There is an interesting thread at AVS about a ‘Perfect Vision’ magazine article that confirmed after proper testing that many TVs throw away 50% of 1080i resolution. http://www.avsforum....threadid=608670

After reading this there is no doubt in my mind that many TVs discard 50% resolution using bob. How anyone could doubt it I don’t know. I'm sure Ikari will still find some way to disagree

Hah, if you read my post, I said 768p displays will throw away 50% of the res REGARDLESS of the method of de-interlacing used.

I was simply arguing that bob does not discard any more of the res than weave does, when bobbing/weaving down to the same res.  My examples illustrate that.

[Trainspotter]

If you read the link I mentioned above you’ll see a lot of 768 displays were featured in the test. Contrary to your inaccurate information above, many of them threw away 50 percent of the res as they only use one 540i field to create a 768p frame. This was visible in the 1080i test patterns they used. Other displays featured true deinterlacing preserving a lot of the original 1080i picture. 1080p downscaled to 768p keeps over 200 lines of extra information over bobbing 540i to 768p.

I looked at your examples above, and I actually think they hurt your case! All I see with the bobbing example is flashing from black to white, not the black and white lines being onscreen at the one time as they are with weave.

As darklord pointed out to you a gazillion times you seem to think that showing detail in the 2nd frame somehow counts as the detail not being lost from the 1st frame. This is rubbish. The whole point is that unless you see the entire 768 lines at the one time, its pointless. Particularly with film as all the info is from the same film frame, so presenting an extra 540 lines the following frame does not deliver any more detail. The best way to put it is:

Bob on a 768 display -
1/50th of a second – 540 lines (from the first film frame)
2/50th of a second  - 540 lines (from the first film frame)
3/50th of a second – 540 lines (from the second film frame)
4/50th of a second – 540 lines (from the second film frame)

Weave on a 768 display -
1/50th of a second – 768 lines (from the first film frame)
2/50th of a second  - 768 lines (from the first film frame)
3/50th of a second – 768 lines (from the second film frame)
4/50th of a second – 768 lines (from the second film frame)

I know you’ll say that bob shows lines from a different section of the frame (and this is only true if not using drop field bobbing which most dont by the looks of it) but its irrelevant anyway as  you're not seeing those lines on screen at the same time and its all taken from the same film frame. With weave you’re always getting 768 lines in the same time that bob is delivering 540.

In actual fact your examples above demonstrates this well. You’re just thinking about it the wrong way. which from what I have read of this thread has been your problem all along!!! .

Plus as even you admit bob looks nasty and flickers around horribly so why ever use it?

But it doesn’t matter whether you believe any of this anyway. The fact is there are big magazines like The Perfect Vision and websites who have actually taken hundreds of different plasmas, LCD’s, DLP’s and done real world tests with them. As the thread I linked to points out after proper testing there is no denying that many of these displays (720p 768p & 1080p) throw away up to 50% resolution if using bob. No matter what you say you can’t get around this simple fact. Unless you would like to try and convince us all that all the magazines, technicians and home theatre junkies  have it all wrong and you standing on your own with this crazy view are right!

[pneu]

I suspect the animated gifs arent animating at 50hz on your screen, and you aren't perceiving both fields together.  For some reason Internet Explorer won't animate them at 50hz, only Firefox seems to do it right.

I don't know what more evidence I'm supposed to come up with, I've shown 576i bobbed to 576p and no loss of resolution, 1080i bobbed to 768p versus weaved to 768p, and no loss of resolution, a real world example of reversing the whole process (re-interlacing 7's 576p from Sunrise) and getting matching fields back out of it, demonstrating that original field structure is still there.

And for some reason I seem to be giving the impression that I am praising bob and saying it is acceptable, which I am not.

Oh well!

Trainspotter, on Apr 29 2006, 04:15 PM, said:

1080p downscaled to 768p keeps over 200 lines of extra information over bobbing 540i to 768p.

Actually, 1080p downscaled to 768p loses 312 lines for each frame, with bob none of the original lines are lost, 228 lines are added to each field to get it up to 768.

[pneu]

This is the last example I'll post.  If this doesn't prove it, I don't know what will.

1080i bobbed to 768p
1080i weaved + downscaled to 768p

1080i bobbed to 1080p
1080i weaved to 1080p

A single 540 line field


View these at 100% size and animating at 50hz and you'll see bob is clearly not half the res of weave.

With the single field for example, some of the white horizontal lines are missing, the curves on the circles have gaps in them, and the text is hard to read.  The bobbed version exhibits none of this and there is a definite perception of the two fields together.

[Trainspotter]

Quote

”I suspect the animated gifs arent animating at 50hz on your screen, and you aren't perceiving both fields together.  For some reason Internet Explorer won't animate them at 50hz, only Firefox seems to do it right.”

I use Firefox 1.5. I wouldn’t use Internet Exploder if my life depended on it.

The gifs are animating just fine. I can see what you think you’re proving, but in reality you’re only hurting your case.

Quote

” I don't know what more evidence I'm supposed to come up with, I've shown 576i bobbed to 576p and no loss of resolution, 1080i bobbed to 768p versus weaved to 768p, and no loss of resolution, a real world example of reversing the whole process (re-interlacing 7's 576p from Sunrise) and getting matching fields back out of it, demonstrating that original field structure is still there.”

All you’ve shown us with your gifs is that the data is still present in the next frame, which is exactly what you’d expect! What you haven’t illustrated is that the data in any way appears to be on screen at the same time as the last frame. Which it clearly isn’t on a digital display. Otherwise all digital displays would still be capable of interlaced scan wouldn’t they?

Quote

“And for some reason I seem to be giving the impression that I am praising bob and saying it is acceptable, which I am not.”

No but you’ve said you think its better to weave on a 768p display which is to put it mildly, crap. There is a difference of 228 lines per every single frame with weave!

Quote

“Actually, 1080p downscaled to 768p loses 312 lines for each frame, with bob none of the original lines are lost, 228 lines are added to each field to get it up to 768.”

Read above. What you are saying would only be true if the detail in the following frame was relevant to the detail in the last frame. If you’re using that argument, why don’t you say that detail from frame 3 is also perceived onscreen at the same time as frame 1. Field 4 as well perhaps? Where does the madness end?!

As I’ve read from the beginning of this thread you seem to be approaching this argument like its all still on an interlaced CRT. Fields aren’t even possible on a digital display. They are all upscaled to full frames. What you get in one frame is all that matters. In the case of weave that’s the full 768 lines. In the case of bob, that’s 540 lines upsampled to 768. Its pretty simple when you think about it.

Quote

“This is the last example I'll post.  If this doesn't prove it, I don't know what will.

http://members.optus...ajac54/768p.gif
http://members.optus...54/768pfull.jpg

http://members.optus...zajac54/bob.gif
http://members.optus...jac54/1080p.jpg

View at 100% size and animating at 50hz and you'll see it's clearly not half the res of weave.”

Still doesn’t prove a thing, because you cant prove a point that has no merit. You are just quickly alternating between one field and the next. What does that prove? There is no perception of detail on screen at the same time. If there was it wouldn’t be flickering up and down, and your other gif pictures wouldn’t alternate between black and white boxes (which in the real world is considered a failure to represent a 1080i pattern properly if you read the link to the thread I posted). If you were right it would look like a nice rock solid picture with minimal flicker! (like you often get with a still interlaced picture on a CRT).

How do you explain the new Panasonic plasmas that have just hit the market? http://panasonic.com...m?objectID=3125 . They now feature weave de-interlacing to 1080p before downsampling to 768p. Why would Panasonic introduce this feature if the old bob method looked better? Why does everyone that has seen the new 60A Panny next to the older 500A with a 1080i signal say that the new 60A shows heaps better detail and sharpness? Is everyone mad except you?

Dude, let me put it this way. Either we believe you and some animated gifs which prove nothing (they actually hurt your case), or we believe our own eyes and all the home theatre experts of the world who have actually gone out and done real world test on all the digital displays with 1080i test patterns. Read the link above and the quotes below!!! They are unanimous in the opinion that many of these displays throw away up to 50% resolution with 1080i signals.  No offense, but I know who I’m putting my faith in

And just to ram this point home, how about I show you some quotes from the thread at AVS and the article at Perfect Vision -

Many more quotes here - http://www.avsforum....t=608670&page=1

Quote

Merson uses an SMPTE 133 test pattern at 1080i, with and without motion, to simulate HDTV broadcasts on a variety of LCD flat panels, plasmas, RPTV's, LCoS rear projectors, DLP's, and DLP front projectors from local retailers and manufacturers

56% of the 53 HDTV's tested failed the test by processing only 540 of the 1080 lines transmitted in each frame

##############

The article explains that newer chipsets by Pixelworks and Genesis, properly deinterlace 1080i signals, but many of the HDTV's tested still used older chipsets and failed the test

#############

The Perfect Vision has tested a number of HDTVs and has found that many dispose of 1/2 of the vertical, 1080i resolution.

In the units tested, all LG, Sharp, as well as some Sony (older LCDs as well as plasmas), Samsung (DLP) failed and threw away 50% of the vertical resolution. Passing were the newer Sony LCDs, Panasonic plasmas and LCDs, Fujitsu plasma, Samsung LCDs and some others.

##############


For a long time, Sharp's DLP projectors (the Z10000 and Z12000) were the only displays known to properly handle 1080i deinterlacing. When I first bought the Z10000, I was stunned by how much better 1080i properly converted to 720p looked, even when compared to higher resolution displays I'd seen with improper 1080i processing (specifically, the JVC SX21 with an anamorphic lens, native resolution of 1400x1050). My jaw really dropped when I paired that same SX21 with an HD Leeza, which was the only processor that properly deinterlaced 1080i for the longest time. This was before 1080p displays were widely available, and 1400x1050 LCOS was pretty stunning.

#################

Merson's Sept/Oct TPV review of Sharp's 56DR650 DLP (page 85) mentions this alternating full-screen B&W flashing instead of 720 alternating B&W horizontal lines from SMPTE 133. A Faroudja rep told him some of their chips 'weave' the two 540-line fields into one 1080-line frame before 720p downconversion. With this Sharp display and Faroudja chip, though, the 720p downconversion produced the alternating fields bobbing from solid white to solid black screens.

##################

the issue relates to *all* kinds of HD displays and STBs that process 1080i signals, irrespective of their native resolution. So 720p and lower-res displays are just as much at risk as the 1080p-native displays.

If anything, since 1080p displays are a new vintage, they are the *least* likely to suffer this problem since they are more likely to be using the newest scalers.

###################

Should mention, for real newbies, there's a simple alternative: Watching 1080i CRT-based (RPTV, FP, direct-view) displays that needn't deinterlace 1080i. No concerns about passing or failing deinterlacers. Not touting them in the wrong forum, just pointing out 1080i CRT displays show 1080i as it's delivered, at 60 fields per second. Yes, CRT sales are diminishing and fixed-pixel technology is supplanting them--although, from these tests, it looks like deinterlacing hardware/algorithms, required for all-progressive display, haven't advanced quickly enough for much modestly priced fixed-pixel hardware.

Gary Merson, the author of the original Perfect Vision article, has also posted this real world example of weave vs bob. No fabricated animated gifs here. This is a real photo of a TV in action.

Quote

To help explain what you see when watching a display that "bobs" single fields I have attached a section of full resolution versus half-resolution photo to illustrate what happens when your display "bobs" single fields.

The bob_halfres is a portion of the 1920 x 540 pixel photo.
The full_res image is a portion of the 1920 x 1080 pixel photo


I can not upload the full 1920 x 1080 images due to size limitations of AVS uploads.

full_res_magnify-1 (2).jpg

bob_halfres_magnify-3 (2).jpg

Go on Ikari, tell us all again how the entire world has got it wrong and you are the only enlightened one

[pneu]

I honestly don't have the energy to comment on all of the points you've just made.

The anigifs I've provided show what happens with bobbing.  Interpret them however you wish.  If you open the 768p bobbed gif in one tab, and the 768p weaved in another tab, and alternate between the two, paying attention to what detail can be seen in both, and honestly think the bobbed one looks like half the res, then that's fine by me.

[kaanage]

madmax, on Apr 24 2006, 12:38 AM, said:

ROFLMAO!       Geez, can't wait for Blu-Ray so I don't have to worry about any of this....just have to remember to buy a compatible display.......

But if the source is 1080i........
The only truly compatible display for an interlaced signal is a CRT.

[pneu]

...or one of those ALIS plasmas  

[kaanage]

Ikari, on May 16 2006, 01:21 PM, said:

...or one of those ALIS plasmas  

Not really - and only for 1080i is it "native".

Rather than a proper deinterlace, by showing the successive interlaced signal source frames as 540p by discarding the interlacing lines, these plasmas reduce the number of lines from the potentially available 1080 (by proper deinterlacing successive frames) in the source down to 540 for the display's native resolution. It's just that every image line for each frame is being shown which makes for no scaling issues with 1080i signals and results in a pretty good display.

And for non-1080i signals, these displays can be pretty ordinary (depending on whether how well the scaling has been implemented, just like any other plasma).

[:)]

kaanage, on May 17 2006, 11:04 AM, said:

Not really - and only for 1080i is it "native".

Rather than a proper deinterlace, by showing the successive interlaced signal source frames as 540p by discarding the interlacing lines, these plasmas reduce the number of lines from the potentially available 1080 (by proper deinterlacing successive frames) in the source down to 540 for the display's native resolution. It's just that every image line for each frame is being shown which makes for no scaling issues with 1080i signals and results in a pretty good display.

And for non-1080i signals, these displays can be pretty ordinary (depending on whether how well the scaling has been implemented, just like any other plasma).

thats actually not quite right kaanage, if I have read correctly what you have posted.

For 1080i the hitachi alis actually discards lines to form a 1024 picture which it then splits up into two fields that it alternatively displays to form the entire picture.

it might be worth a read of this article I posted.

http://www.dtvforum....showtopic=13604

I think infact they do a great job for 1080i and pretty damn good for SD, 576p as well in my opinion. Given the choice of plasma alis would be top of my list. I understand hitachi have a full 1080 line model on the way aas well which should be very interesting to check out.

[kaanage]

alebonau, on May 22 2006, 06:33 PM, said:

For 1080i the hitachi alis actually discards lines to form a 1024 picture which it then splits up into two fields that it alternatively displays to form the entire picture.

The e-ALIS technology is something I have not read of before - that is 1024p display lines compared to the 1024i of the original ALIS. You learn something every day thanks!

For non-1080i sources, they are still at the mercy of the quality of their scalers, just as normal plasmas are.

[Adam-O]

This thread is golden... especially the first couple of pages!

Darklord and Ikari both have very valid points of view, but are looking at things quite differently.

I read it all when it was originally happening, a month or so ago, but didn't have time to comment back then. Kind of wish I had

Come on guys? Have you resolved everything now?

Darklord, do you still say that these displays throw away 50% of the information contained in the video signal? Ikari, do you still dispute this?



Cheers,
Adam

P.S. Darklord, I think your guide was an excellent introduction to de-interlacing and scaling. Well done!