117 replies to this topic

[Darklord]

It's official: I give up.

I wonder what kind of anti-interpolation machine Ikari envisions could perform the task of extrapolating original 540 line fields from interpolated 576p frames of Seven’s HD broadcast? And I wonder how it would differentiate from original detail and interpolated detail, and then restore the interleaved line structure? Maybe it would powered by that awesome pixel fairy dust I’ve been reading so much about?

[pneu]

I'm not saying it would be the original lines (only original lines if it's a 1080p bob conversion back to 1080i, as I've always said)

Depends how Sunrise has been de-interlaced from 1080i50

If they've taken a 540 line field and scaled the colums of pixels down to 720 wide, and upscaled the 540 lines to 576 , then the 576 lines are pretty damn close to the original 1080i50 fields.

On the other hand, if they've bobbed to 1080p and then scaled down the 540 lines + interpolated data, there's a hell of a lot more averaging and weighting of pixels going on and it probably wouldnt work too well

Sorry!  Just couldn't resist!

[pneu]

Darklord, this is going to sound quite bizarre, but now I'm starting to think bobbing 1080i for display at 1366x768 may be technically superior to weaving!  Blasphemy, I hear you say,  but more of the original field data might be kept!

WEAVE:
1920x1080 = 2073600 pixels  
1366x768  =  1049088 pixels
This represents a loss of 1024512 pixels from the original 1080p frame!  That's a lot of information lost in the downscaling process, almost half the detail!  I'm tempted to argue that "50% of the video information is thrown away" for weave, not bob!  On the bright side, the full theoretical resolution of the 1366x768 is closer to being achieved as the source was higher than the display mode.  Oh yeah, and it will look quite nice, obviously

BOB:
1920x540 = 1036800 pixels
1366x768 = 1049088 pixels
This represents a total artificial increase of 12288 pixels, including 228 new fictitious interpolated lines, and a loss of 554 columns of pixels.

Now, if we tried our best to recover those original 1920x540 fields we might get much more of them back (but not all) from the bob de-interlaced 1366x768 version!

Ok, that's enough rambling for now, I made this post only to entertain, not to win an argument

BUT, having said that, I will get to work on presenting some visual evidence to you that will demonstrate how bob conveys all the picture information from both fields, despite not being natively interleaved.

lol

[pneu]

Darklord, on Apr 16 2006, 05:56 PM, said:

I wonder what kind of anti-interpolation machine Ikari envisions could perform the task of extrapolating original 540 line fields from interpolated 576p frames of Seven’s HD broadcast?

...Maybe it would powered by that awesome pixel fairy dust I’ve been reading so much about?

Easy!  Only 36 extra interpolated lines were added (but only for 1080i50 sources such as Sunrise because 7 weaves fields for their 25p HD content before scaling down to 576p), so just go through each line and compare it to the one below, and if it's a match, omit it.  After removing the 36 lines the frames are now slightly out of anamorphic aspect ratio, so scale/strech the 720 horizontal resolution to restore the 16:9 aspect.

Same goes for getting the original 540 line fields out of the 1366x768 bobbed version.  228 new lines were added to the 540 to get it up to 768, so we just need to find them.

1.  If it's used crude bobbing (line doubling), go through each line and compare to the one below and if it's a match, remove.  Or, just remove one in every 540/228=2.36 lines.

2.  If it's used interpolating, it's tough to know which line(s) it came from, so you'd have to rely on just removing one in every 2.36 lines.

3.  If it's added the 228 new interpolated lines to the existing 540 AND blended them with any of the original 540 lines (thus causing a loss off original picture information) , then obviously we can't get the original 540 lines back out.  However the original 540 are still all there on the screen and will work to convey all the information of the original 540 line fields.  Obviously you disagree with this last bit Darklord and I intend to show some evidence.

All that's left is to scale the 1366 back up to 1920 and youve got something close to the original fields (1:1 line mapping with original fields but a loss of horizontal resolution -- unavoidable regardless of the de-interlacing method used).

Darklord, in the meantime why don't you just chuck in a DVD and switch between bob and weave and look at some fine detail, like say, small text in one of the menus, and observe the perception of both fields when using bob.  Then compare with field captures to even further illustrate that it's not the same as looking at 1 field!

[pneu]

btw, had a look at a 50" plasma running bob de-interlaced 1366x768 from native 1080i50 source on 9HD (was the national news studio shots and screen titles that I was looking closely at) and for the life of me couldn't actually see much flickering.  The 9 watermark might've had some flicker on the bottom edge of it but I could barely notice.

Maybe the plasma was applying some additional filtering to reduce the flicker?  Normally bob looks a lot more flickery than that (eg. 576i DVD's bobbed) , but even with all that extra res I was expecting some more flicker than what I saw.  Or, maybe if I actually brought the thing home and spent enough time watching it I might start noticing the flicker?  Who knows.

Sorry for the multiple posts, but to be honest I have nothing better to do with my spare time!  

Sorry if I am wasting your time by making you consider all these arguments but where else and with who else am I supposed to debate this topic in great detail?!

I respect your opinion Darklord and always have (your original article is excellent and I would testify 99.99999% of it , just our opinions of how bob is perceived on the screen are different.  But hey if we all shared the same opinions the world would be a boring place, would it not?

[Darklord]

Ikari,

I’m sorry, but this is now beyond a joke.

It’s become quit clear that you are either:

1. Taking the piss, perhaps with aim of being intentionally provocative.
2. In a "locked mindset", having already made up your mind based upon inaccurate information, and therefore haven’t properly read or acknowledged anything I have had to say in response.
3. One of those weird people that like to argue just for the hell of it without actually believing what you are saying.
4. Stark Raving Mad.

You post about bob being technically superior to weave on a 1366x768 display is so technically absurd I wouldn’t even know where to begin properly deconstructing it. To analyse it and respond to it scientifically I’d be here all night, and quite frankly I’ve wasted enough time on this thread already.

All I will say in closing is this:

As usual you’re completely overlooking what actually makes weave superior, and can’t seem to get your head around the fact that interpolation of odd/even fields destroys the original interleaving effect and worsens image quality regardless of how much original data is present. Do you know what aliasing is for instance? Do you know that ALL bob de-interlaced picture will suffer from it along any diagonal lines?  Why do you think technology like DCDi exists for de-interlaced native interlaced video? Quote – “It’s purpose is to eliminate jagged edges (jaggies) along diagonal lines caused by interpolation.” Do you know that because weave combines fields properly, it doesn’t suffer from any aliasing at all? Let alone effect like moiré or shimmer? (Rampant with all bobbed pictures). I can only assume you don’t know anything about these artefacts if you’re arguing in favour of bob. Maybe you should look them up on the net?

A neat downscale of a complete non interpolated progressive 1920x1080p frame to a complete 1366x768 frame will yield far superior quality to a single interlaced half frame (one field - with every second line missing!!!) interpolated directly to 768p. If you can’t understand why this is, then you need to some proper research, and start doing some real world tests for yourself, with various film and video sources. I’d suggest starting with learning about film and the telecine process itself, and how because film is a progressive medium, it is best viewed when resorted to its original progressive state for viewing. Contrary to what you may think, bob does not do this, or give the perception of doing this. Bob simply treats all sources as a native interlaced format, halving perceived resolution, and presenting the frames with all the interlaced odd/even artefacts intact, with the addition of interpolation artefacts and blur. How anyone could argue in favour of it is just amazing. I’ve certainly never encountered it all my years of discussing A/V technology.

Ikari said:

But hey if we all shared the same opinions the world would be a boring place, would it not?

As you may have seen from many threads here at DBA, I love a good technical debate, and will happily listen to other people’s arguments if they have any technical merit. I’m sorry to say that yours do not. Rather, you’ve obviously looked at basic figures (such as the amount of pixels present per field) and have completely dismissed what actually takes place with interpolation of fields to frames, and in turn how that effects our perception of them. You still haven’t acknowledged that any perception of fields being blended together (as seen in native interlaced form on a CRT) is lost when a field is interpolated/up-scaled and presented one after the other. There is simply no room for debate with this point. BOB DOES NOT PRESERVE PERCEPTION OF INTERLEAVING. By denying this point you’re actually doing harm to any other argument you make, as this effect with bob is 100% common knowledge.

It’s important you understand that your aren’t just “disagreeing with me”. You’re disagreeing with the entire A/V industry. Including such people as Joe Kane, who has dedicated his life to the science of visual technology. Not to mention your views are against every company in the world that specialises in advanced video processing (for both broadcast facilities and consumers). Have you had a good read of the Silcion Optix’ HQV technology page for example? Are you honestly trying to suggest everyone has got it wrong except for you?

And what on Earth are you talking about when you say “Now, if we tried our best to recover those original 1920x540 fields we would actually get much more of them back”. Recover with what?! The process takes place on the display itself. Not to mention its not a case of recovering the data, but presenting it in the best possible progressive manor without artefacts (which as everyone and his grandma knows is down with weave). In any case if you use weave, you are already getting the maximum possible amount of detail. Ah hell, I don’t even known why I’m bothering to respond to this point . I’m sure I'm only wasting my time.

Ikari said:

Easy!  Only 36 extra interpolated lines were added (but only for 1080i50 sources such as Sunrise because 7 weaves fields for their 25p HD content before scaling down to 576p), so just go through each line and compare it to the one below, and if it's a match, omit it.  After removing the 36 lines the frames are now slightly out of anamorphic aspect ratio, so scale/strech the 720 horizontal resolution to restore the 16:9 aspect.

You obviously completely ignored or didn’t understand the long section I wrote above on averaging. Please try to grasp this point: In the majorty of cases interpolation does not just interpolate news lines in the interlaced gaps when scaling. ALL lines are interpolated based on the lines present in the original field as part of the scaling process. In other words the entire frame is interpolated from the original data. Just as when 60hz video (from the US) is converted to 50hz (for Australia) they don’t just cut ten fields off the end. The entire series of fields is interpolated in order to achieve smooth seamless motion.

I also love the way you suggest getting this data back is a feasible and practical possibility with bobbed frames from SevenHD. Are you going to design a piece of dedicated hardware with pixel based motion adaptive analysis of individual fields that can correctly extrapolate original data and then convert it to 1080/50i? How is life up there with the fairies dude?

Ikari said:

btw, had a look at a 50" plasma running bob de-interlaced 1366x768 from native 1080i50 source on 9HD (was the national news studio shots and screen titles that I was looking closely at) and for the life of me couldn't actually see much flickering.  The 9 watermark might've had some flicker on the bottom edge of it but I could barely notice.

Ok now I really know you haven’t been paying attention. For starters WEAVE IS ONLY USEFUL FOR 24/25p FILM SOURCES. All displays have to use either bob or motion adaptive de-interlacing (motion adaptive is very rare for HD due to being so processor intensive) for any native interlaced 50i or 60i content, which includes all studio based material.

You will see bugger all flicker during studio material because it’s near stationery! The interpolated/smoothed frames are also being placed over the top of each other so only very fine detail will flicker up and down. However if you were to see vertical or diagonal movement you would see chronic aliasing, plus shimmer and moiré in detailed parts of the picture. Such effects are inherent to native interlaced scan formats (shot at 50i or 60i) when viewed on a progressive monitor and even motion adaptive de-interlacing cant completely remove them (although motion adaptive de-interlacing will give much better detail when there is no movement). This is why native 50i or 60i video sucks when compared to native progressive 50p or 60p. But that was never under discussion in this thread. What we’ve been talking about is weaving of film sources from 24p or 25p material. It’s so blindingly obvious that weave is the best method of de-interlacing for film sources that it amazes me I’m even having this discussion. It seems most other members here have been sensible enough to just leave your comments alone (I noticed you completely ignored Davo’s comments by the way).

Iakri, I’m sorry but I’m withdrawing from this discussion. You are completely ignoring the evidence and fabricating pseudo-scientific hypothesis’ based on your own personal conjecture. Quite frankly none of it has any merit, and it’s a waste of my time. PLEASE for your own sake and others, do some proper research on the net, and learn more about the tried and tested methods of de-interlacing and scaling that are in use across the globe (and please take into account the different de-interlacing methods that are most appropriate for different sources).

Bob actually has a place as “last resort” fallback method for native interlaced video sources, but should never be used for film sources. The sooner you get your head around this basic concept, the quicker you’ll be on you way to being someone that has a good grasp of the de-interlacing and scaling process.

If after reading everything I have to say, and after doing some proper research on the net, you still argue against any of the above, I suggest checking out this link.

Good luck with your research. I'll be praying for your de-interlacing enlightenment.

[pneu]

Once again you treat me like a noob!

"Go and learn about progressive vs interlace, you need to learn about film and video soures" etc.

I know all the facts.  So do you.  However...

Darklord, on Apr 17 2006, 07:47 PM, said:

In the majorty of cases interpolation does not just interpolate news lines in the interlaced gaps when scaling. ALL lines are interpolated based on the lines present in the original field as part of the scaling process. In other words the entire frame is interpolated from the original data.

Wrong!  I clearly said several times that, if it did bob upconvert to a 1080p frame first, and then scale back down to 768p, 576p etc. then HELL YEAH, you'd be losing tons of picture information.  No denying that, I never denied that, was always aware of that, it's a fact.  But why would they bob to 1080p first, and then re-scale down to 768p?  It's indirect, inefficient, would require more processing than going straight to 1366x768 by adding only a few lines, overhead is increased, AND the picture quality becomes degraded, softened, lower resolution.  Why would any manufacturer spend more on a worse quality picture?  If they did that I would truly be shocked.  If all of the 540 lines were averaged with the fake lines then that is a clear reduction in resolution, obviously.  But I don't think this is what happens in most cases of bobbing.

I'm thinking, your article needs to outline different types of bobbing and explain that one type of bobbing (interpolating the original lines with the fake lines) is the only type that reduces resolution.

[Darklord]

At least most newbs have the courtesy to listen, and to learn a thing or two. You're denying evidence and methods used by the entire A/V industry, and instead going on your own "gut feelings"! Now that is damn funny. 5 minutes on Google would set you straight, but hey Ikari knows better!

Again you're simply denying everything I have to say, and cant back up anything you have to say without giving any evidence whatsoever. Interesting debating method.

And I never said anything about bobbing 1080i to 1080p before converting to 768p!!  Where did you get that crap from? I said that bobbing on a 768p display will average a 540 line field directly to 768p. Rather than just "fill in the gaps" as you seem to think happens. I’d love to know what mathematical algorithm you think is used there given 540 doesn’t translate directly to 768 without any averaging taking place.

At least have the courtesy to read what I’ve written properly! (although you are clearly incapable of taking in new information).

[pneu]

Darklord, on Apr 17 2006, 08:39 PM, said:

I’d love to know what mathematical algorithm you think is used there given 540 doesn’t translate directly to 768 without any averaging taking place.

yep

Take a 1920x540 field, add one duplicate line in every ~21 lines [EDITED: should be one in every 2.36 lines] and remove 554 columns of pixels.

No averaging of original lines with fake lines occurs.

[Darklord]

Here’s one for you Ikari, given you think you’ve already done all the research you need to do on de-interlacing. I wasn’t going to bring this into the article or discussion as it complicates matters further, but given the direction this crazy thread has taken, why the hell not?

Did you know that many processors that use bob de-interlacing discard the second field entirely? In other words they up-convert the odd line field to a frame, then skip to the next odd line field. This technique is often used to avoid any jumping up and down of content, but detail wise looks very similar to bob converting that up-convert both fields (as takes place with all interlaced video sources).

Just in case you doubt what I’m saying here, here’s what Silicon Optix (creator of the famous Teranex processor) have to say about this:

Silicon Optix HQV Technology FAQ said:

Simplest Competitor Approach (Non-Motion Adaptive):

The simplest approach to avoid these artifacts is to ignore the even fields. This is called a non-motion adaptive approach. In this method, when the two fields reach the processor, data from the even fields are completely ignored.

The video-processing circuitry recreates or “interpolates” the missing lines by averaging pixels from above and below. While there are no combing artifacts, image quality is compromised because half of the detail and resolution have been discarded.

More-advanced techniques have been adopted by virtually all standard-definition video processors, but this basic approach is still sometimes used for high-definition signals, due to the increased computational and data-rate requirements of higher video resolution.
With video processors from some competitors, only 540 lines from a 1080i source are used to create the image that makes it to the screen. This is true even for video processors from companies that may have been considered providers of flagship performance in the standard-definition era.

But ermm, yeeah. Field interpolation works great right Ikari? It preserves all the detail right? Who needs weave?

[pneu]

Darklord, on Apr 17 2006, 08:49 PM, said:

Did you know that many processors that use bob de-interlacing discard the second field entirely? In other words they up-convert the odd line field to a frame, then skip to the next odd line field. This technique is often used to avoid any jumping up and down of content, but detail wise looks very similar to bob converting that up-convert both fields (as takes place with all interlaced video sources).

lol what an awful thing to do.  Anything that does that is an embarrassment and an abomination!

I would laugh at anything that throws away half the resolution.   eg. drop field.

Discarding every second field turns 50fps material into 25fps and effectively doubles the shutter speed and therefore should be considered as something else other than bob de-interlacing.  It completely ruins the source material, a straight bob will include all the recorded information at the correct framerate.


EDIT:  Darklord don't you reckon it might be worth making a point in your guide that there are different methods of bob de-interlacing , some a lot better than others, cause some just line duplicate and dont destroy the original field information, others blend the interpolated stuff with the original stuff and make a mess, and some just discard the second field, as you have just mentioned.  Theres a lot going on there, i think it's worth differentiating between the various types of bobbing.

[pneu]

Just to make it clear, the topic we are debating is:


"Bob de-interlacing throws away 50% of the resoluton"


Darklord is for the argument, I am against.

[pneu]

Darklord, on Apr 17 2006, 07:47 PM, said:

A neat downscale of a complete non interpolated progressive 1920x1080p frame to a complete 1366x768 frame will yield far superior quality to a single interlaced half frame (one field - with every second line missing!!!) interpolated directly to 768p.

Duh, of course it will, because youre only comparing 1 field with 1 frame.  Show that second bobbed field immediately after and youre revealing the full image, the full detail.

I realise your response to this is that the interleaving effect is destroyed and no more detail is perceived in the second bobbed field, but you coudln't be more wrong.  Think about where field 1 is in relation to field 2 within the bobbed frames , they still line up in the correct interleaved way.  It's the fake interpolated lines of field1 that are clashing with adjacent non-interpolated lines of the field 2, that causes the bob flicker.  The same goes for field 2.

Darklord, on Apr 17 2006, 07:47 PM, said:

If you can’t understand why this is, then you need to some proper research, and start doing some real world tests for yourself, with various film and video sources

I already said I've compared bob vs weave in the real world and that I can perceive the full detail of both fields in the bobbed version.  What real world tests have you done?  I asked you a few posts ago.

[pneu]

Here's an example nexx produced a while back (you might remember this from the "576p is not HD" thread!)

http://img360.images...ntitled19gd.gif

Observe the curve underneath the Continental logo.  When bobbed it becomes a smooth curve, and there is clearly perception of the detail from both fields.


NOTE: The two bobbed fields are supposed to be alternating 50 times per second, if they aren't doing this then there won't be a perception of extra detail.  Using Firefox, I get it animating 50 times per second but briefly slows down every now and then (CPU lag?), so it's still not quite an accurate pure 50hz alternation between the two.  For some reason when I open in IE it only shows a few fps, so please be sure to view in a browser that animates it at 50fps

[Darklord]

Ikari said:

Just to make it clear, the topic we are debating is:


"Bob de-interlacing throws away 50% of the resoluton"

Darklord is for the argument, I am against.

Actually looking at your multiple posts to yourself, you seem to be arguing with yourself now. You need help dude!

Ikari said:

lol what an awful thing to do. Anything that does that is an embarrassment and an abomination!

I would laugh at anything that throws away half the resolution. eg. drop field.

Did you even read the quote above from HQV that explains why drop field is used? Let me quote the relevant bit for you “The simplest approach to avoid these artifacts is to ignore the even fields”

In other words with bob and film sourced material you have two choices: 1. Upscale all fields to frames, creating horrible jumping up and down of detail and motion artefacts, (along with all the other interpolation nasties) and artificially turn 25p into 50p, or discard every 2nd field so that the original film cadence remains accurate and detail doesnt jumps up and down as much (it still will to a degree due to differing spatial positions of objects in one frame to the next). Bob is horrible no matter which technique is used.

The drop field method is actually the preferred method for film sources as it exhibits fewer artefacts, and there is no discernible difference in detail between it and up-scaling of both fields, for the many reasons I have explained above. With bob you simply don’t get the perception of extra detail when it isn’t onscreen at the same time.

You seem to also conveniently overlook the other inherent negatives of field interpolation with every response you make. I've argued these continuously and I'm yet to hear any sensible response from you about them. All you’re concerned with is the fact that “all the lines are still there” (and often even this isn’t the case as explained above!). If you had actually properly compared bob to weave de-interlacing with film sourced material on a good quality display, you’d know full well that weave looks vastly superior in every way. Remember no one is arguing against this but you!

Ikari said:

Discarding every second field turns 50fps material into 25fps and effectively doubles the shutter speed and therefore should be considered as something else other than bob de-interlacing. It completely ruins the source material, a straight bob will include all the recorded information at the correct framerate.

Ikari, this comment illustrates better than any other that you’re just not getting the technical processes involved. What you have said would only be true if were talking about native 50i video (with 50 fields from different moments in time), not film which is sourced from 24/25p.

1080/50i  when taken from a film source IS 25p! I don’t how much clearer I can make it. It’s simply segmented into two fields that are from the same moment in time. It is sometimes also referred to in the industry as 1080/25psf (segmented frame) because those segments can be easily put back together for proper presentation (weave!) on a progressive display. Bob artificially turns it into 50p (if each field is upscaled) without any benefit to motion. Don’t you see that’s the entire point? On an interlaced display this segments field method can work ok, as you’re essentially perceiving 25 full frames per second presented at 50hz due to the way that native interlaced scan and CRTs work (although progressive presentation via weave is still much better). On a digital display with bob de-interlacing you’re seeing 50 frames a second that have been artificially created from individual fields! That’s why the drop field technique is actually MORE accurate than up-scaling each field. You’re seeing 25 unique motion updates a second (with each frame repeated) just as you’re supposed to!

I also love the way you used the phrase “ruining the source material” in reference to drop field. Bob de-interlacing of any kind destroys line structure, detail, blurs the picture and introduces interpolation artefacts such as shimmering and aliasing. Weave presents the material at full vertical resolution, the correct frame cadence, without any interpolation nasties, and restores the 540i fields to their original 1080p state. This is a case of simple 2:2 pull-down for accurate presentation at 50hz. Even cinemas double each frame and present  films at 48p! To present 50hz films at 50p on a digital display is a perfectly acceptable method of presentation, providing those fields are woven together and repeated. In fact with Blu-Ray we’ll have the option of 48p and 72p on some displays for 100% faithful 2:2 or 3:3  reproduction of 1080p sources. Hopefully native 1080p and native 24p will eventually see the death of these ridiculous kinds of interlaced discussions.

Yes, Ikari I am now treating you like a “newb” and telling you to do some proper research, because in all honesty you are displaying technical ignorance of all de-interlacing processes, and their well established pros and cons. Not to mention you ignore all evidence and technical explanations presented to you. Just to illustrate this clearly for you, I spent 2 minutes on Google and came up with the following quotes from respected sources. There are many more out there to be found. (note: bold emphasis added by me).

www.hometheatrehifi.com said:

Single-Field Interpolation (or “Bob”)
This just involves taking each field and scaling it to a full frame. The missing lines between each of the scan lines in the field are filled in with interpolated data from the lines above and below. Done badly, the screen looks blocky and pixellated. Even done well, the image looks very soft, as image resolution is unavoidably lost. In addition, thin horizontal lines will tend to “twitter” as the camera moves. These thin lines will fall on just one field of the frame, so they will appear and disappear as the player alternates between the odd fields and even fields. This is the most basic deinterlacing algorithm, and the one that almost every deinterlacer falls back on when nothing else will work.

www.hqv.com said:

Simplest Competitor Approach (Non-Motion Adaptive):

The simplest approach to avoid these artifacts is to ignore the even fields. This is called a non-motion adaptive approach. In this method, when the two fields reach the processor, data from the even fields are completely ignored.
The video-processing circuitry recreates or “interpolates” the missing lines by averaging pixels from above and below. While there are no combing artifacts, image quality is compromised because half of the detail and resolution have been discarded.
More-advanced techniques have been adopted by virtually all standard-definition video processors, but this basic approach is still sometimes used for high-definition signals, due to the increased computational and data-rate requirements of higher video resolution.
With video processors from some competitors, only 540 lines from a 1080i source are used to create the image that makes it to the screen. This is true even for video processors from companies that may have been considered providers of flagship performance in the standard-definition era.

Wikipedia said:

Line doubling takes the lines of each interlaced frames (consisting of only even or odd lines) and doubles them, filling the entire frame. This results in the video having effectively half the vertical resolution, scaled to the full resolution. While this prevents mouse teeth, it causes a noticeable reduction in picture quality. This technique is also called bob deinterlacing, because the fields are bobbed up and down.

sigmadesigns said:

Scan Line Interpolation
Scan line interpolation generates interpolated scan lines between the original active scan lines. Although the number of active scan lines is doubled, the vertical resolution is not. The simplest implementation, shown in Figure 2, uses linear interpolation to generate a new scan line between two input scan lines:

And the best one:

www.edn.com said:

An alternative approach involves doubling up the odd scan lines to form an entire frame, then duplicating the even lines to form the next frame (Figure 4b). This approach, often referred to as bob , eliminates many of the motion artifacts but induces its own problems. Objects' horizontal edges that align with one of the original interlaced scan lines, or stripes one scan line thick, appear and disappear from frame to frame, causing annoying flicker. By using only one field's scan lines to create the entire frame, you effectively halve the vertical resolution of that frame. And where field-to-field pixel variations are a result of diagonal edges and not object movement, the edges end up with indistinct, distorted appearances.
More elaborate versions of bob and weave interpolate the missing scan-line information in each artificially constructed frame, either from nearby pixels in the same field's scan lines or from pixels at identical locations in past and future fields (Figure 4c). The number of pixels that the interpolation process uses and the proportional priority given to the information in each of these pixels differentiates the alternatives. The more complex the algorithm, the more logic gates or lines of code you need to execute it and the faster those gates or the processor running that code needs to operate. Also, the more pixels you use in the interpolation process, the more buffer memory you need to hold the pixels' respective scan lines.
The best approach to deinterlacing combines the best aspects of both bob and weave. Motion-adaptive deinterlacing selects a temporal or spatial-interpolation algorithm for moving and stationary objects, respectively (Figure 5a and Figure 5b). The selection occurs either on a field-by-field, pixel-group-by-pixel- group, or, ideally, pixel-by-pixel basis, because different sections of the image often move in different directions and at different speeds. How can you tell whether a pixel is in motion? The answer to this question represents the black-magic proprietary technology that no developer is willing to publicly divulge. The Faroudja division of Sage, for example, touts its DCDI, which, according to the company, works well on diagonal edges.

I imagine you’re still not convinced and never will be.

Ikari said:

I already said I've compared bob vs weave in the real world and that I can perceive the full detail of both fields in the bobbed version.  What real world tests have you done?  I asked you a few posts ago.

Having run an A/V website for many years, as well as worked in the Hi-Fi retail and installation industries, I have tested calibrated and installed countless brands of DVD players, projectors, CRTs, digital displays, processors and so on. Having played with top end dedicated scalers I’ve had the opportunity to directly compare bob with weave (and motion adaptive de-interlacing) on many different types of displays using a variety of film and video source material.

I’m yet to see any example of bob de-interlacing which results in a picture that comes close to weave on any quality display. Its that simple. I see bobbed pictures every day on HD displays (some of which are very good quality in other areas) and the artefacts stand out like a sore thumb. Anyone doubting that weave is the best method of de-interlacing film sourced material needs their head examined.

Ikari said:

Here's an example nexx produced a while back (you might remember this from the "576p is not HD" thread!)

http://img360.images...ntitled19gd.gif

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

[pneu]

Are you even viewing it at the right fps?  

I never said it was superior to weave (except maybe at 1366x768 purely because 1 million pixels arent lost like they are with weave), I'm only arguing that 50% of the information is not lost using bob, and that both fields are perceived, which you deny, yet there it is on the screen.

I was going to make my own example of field vs 2 bobbed fields, but I don't quite have the tools for it (dialup internet, slow cpu), so I used nexx's example.

I'm surprised that you still think I can't tell the difference between film and video based sources, but anyway, no need to argue that as you've already denounced me as a noob

I made it clear what we were debating for other readers because we seem to have covered an awful lot of extra information that is superfluous to the main issue - that bob de-interlacing does not discard 50% of the picture information.

I got nothing against you Darklord but it looks like we'll have to agree to disagree on this one

[cwt]

[quote name='Darklord' date='Apr 17 2006, 09:19 PM' post='388132']
Darklord ; just read a review of the Yamaha DPX 1300 in Widescreen review. It has the realta hqv chip for de -interlacing; as you know one of the best in the world . A quote from Greg Rogers the reviewer [ Most projectors and stand alone scalers convert 1080i video to 720p using interpolation to scale each individual field directly to a 720p frame. The actual vertical resolution of the displayed image is limited to the 540 line vertical resolution  of the interlaced fields. ] Joe Kane writes for this magazine ; Gary Reber vets everything published - Roger's also says [ the vertical interpolation also acts as a filter which reduces the vertical resolution and further softens the image.Its now Rogers fave 720p single chip. Its obvious why Faroudga are behind the times with bob de-interlacing  for 1080i-why YAMAHA and infocus have changed to proper motion adaptive de interlace for video. :blink:

[Darklord]

I just read a similar review of the 1300 at Ultimate AV cwt. Good to know I'm not the only crazy one up at 3:30am reading A/V reviews

I quite liked the Yamaha 1200 so read this review with great interest (and look forward to reading the Widescreen Review one – those guys rock). As expected the HQV processing is a big improvement over the bob de-interlacing Faroudja chipset present on the 1200 (which as you know could apply motion adaptive de-interlacing to SD, but not HD sources). And yes it’s interesting that all the A/V professionals agree that bob de-interlaced 1080 is essentially 540p! But hey try telling that to Ikari! I did and look where it got me…

[cwt]

Darklord, on Apr 18 2006, 03:48 AM, said:

I just read a similar review of the 1300 at Ultimate AV cwt. Good to know I'm not the only crazy one up at 3:30am reading A/V reviews

I quite liked the Yamaha 1200 so read this review with great interest (and look forward to reading the Widescreen Review one – those guys rock). As expected the HQV processing is a big improvement over the bob de-interlacing Faroudja chipset present on the 1200 (which as you know could apply motion adaptive de-interlacing to SD, but not HD sources). And yes it’s interesting that all the A/V professionals agree that bob de-interlaced 1080 is essentially 540p! But hey try telling that to Ikari! I did and look where it got me…

lol bob interlacing is old school . If Ikari is right why did HQV switch   ??

[pneu]

Darklord, on Apr 18 2006, 03:18 AM, said:

And yes it’s interesting that all the A/V professionals agree that bob de-interlaced 1080 is essentially 540p! But hey try telling that to Ikari! I did and look where it got me…

lol, I agree too, 'cause 540p is 1080i, always has been.  Just like 576i is 288p, I consider it to be nothing more, never was anything more.  There is almost always a better solution than bob, I've always said that.  The only thing that's a load of bs is "50% of the picture information is thrown away".  It's a matter of principle, 540 lines per 50th of a second to begin with, 540 lines per 50th after bob, AND the data from both fields is perceived on the screen at the same time, plus artefacts.  Don't believe me?  Here's a better comparison of bob vs weave that I've made from a PAL DVD.

http://members.optus.../COMPARISON.gif

Please ensure it's animating at 50fps on your browser.

Observe the tiny 1 pixel high notch on the first tip of the "N" -- thats 1 pixel from the progressive source (from field 2) represented and perceived in the bobbed version.  Same goes for the tip of the sweep underneath the "R" , observe how it's completely absent from field 1.  There's no way a single field can be compared to bob in terms of resolution.

As for the engineers at HQV, they truly have no idea what they are doing if they think it's acceptable to throw away every second field.  Yes, I'm criticising professionals who have more experience than myself, but you know what, even the professionals get it wrong.   Look at how many pros get it wrong thinking it's acceptable to use drop field on SD shows to make them appear to be film based.  Seven have got it wrong with using bob instead of adaptive motion de-interlacing.  I'm sure there are plenty of other stuffups that you can think of that have been made by industry pros.

I'm all for best quality images, always have been.

[pneu]

And here is your original comment after showing me your biased illegitimate field vs frame animated gif:

Darklord, on Apr 16 2006, 04:54 PM, said:

If that animated gif was comparing a weaved frame to two successive blurry interpolated frames (both frames bobbed from the original fields) the image quality would be no different! In fact it may be slightly worse as some part of the image would be seen bobbing up and down.

hah, compare the bobbed image to the field beside it, it's clearly much better!  Comes close to looking like the weaved version above it.

Darklord, on Apr 16 2006, 04:54 PM, said:

The result is a single blurry frame, and NO PERCEPTION OF FIELD COMBINATION.

Oh yeah, no perception of field combination, compare the bobbed to the weaved image above it, plenty of perception of field combination.

Darklord, on Apr 16 2006, 04:54 PM, said:

What I’m trying to tell you (and I really don’t think I can make this much clearer) is that its completely irrelevant that the lines of original broadcast are there in the next frame, as they no longer appear to be connected to the previous frame.

Wrong again.  The bobbed fields in the animated gif are alternating at 50hz and clearly give the perception of 2 fields, just as a native 50hz interlaced CRT alternates scanlines every 20ms.

[Darklord]

Ikari said:

The only thing that's a load of bs is "50% of the picture information is thrown away".  It's a matter of principle, 540 lines per 50th of a second to begin with, 540 lines per 50th after bob, AND the data from both fields is perceived on the screen at the same time, plus artefacts.

Well at least I got a little “plus artefacts” comment out of you (understatment of the year)! I can’t believe you are actually admitting bob looks nasty! Regardless of any disagreements we have about “discarded resolution” there is certainly no denying that bobbing results in a horrendous image looking at your lovely examples

Look Ikari, here’s the thing. Bob can and often does throw away 50% of resolution, depending on what technique is used, and how you actually think about an interlaced storage/compression method like 1080/50i (which for film is simply a 1080/25p delivery method). What I mean by this is the following.

1. If drop field bob is used, then 50% resolution is gone. Full stop. I know you agree on this much. However you seem to be overlooking that fact that a.) This is a very commonly used method (you couldn’t possibly deny this after reading all the references to it I posted above from reputable sources) and b.) Contrary to your claims this is often a preferable method of presentation than multiple field bob (although weave is far better than both), as it means far less “bobbing” artefacts are present, and it preserves the original 25p frame rate. What would you rather? More detail or a picture that destroys the original film cadence, and jumps up and down so much it’s unwatchable?

2. Even if bobbing all 50 fields to 50p, you are discarding half of the vertical resolution in terms of what is present in one single frame. Whether you like it or not, progressive spatial resolution (i.e. the number before the p) is measured in single frames (not two field like interlaced). Agreed? If it wasn’t then a format like 720p would actually be known as 1440p. You can only take into account the detail present in one frame with progressive scan. Therefore a single progressive frame when de-interlaced from film (where the fields are from the same original frame) should show all the detail from both fields (1080p from a 1080i film source) NOT interpolated 540p frames one after the other with the detail bobbing up and down. Only a mad man would suggest that bob is a more faithful and detailed representation of the original source. It clearly isn’t.

Combine the above with the loss of “perception of detail” through interpolation to a full frame, as I have mentioned numerous times (and will get onto in a sec below) and it’s certainly accurate to say that bob doesn’t preserve detail, and weave is a far superior alternative for film sources.

Ikari said:

Don't believe me?  Here's a better comparison of bob vs weave that I've made from a PAL DVD.

http://members.optus.../COMPARISON.gif

Please ensure it's animating at 50fps on your browser.

Observe the tiny 1 pixel high notch on the end of the "N" -- thats 1 pixel from the progressive source (from field 2) represented and perceived in the bobbed version.  Same goes for the tip of the sweep underneath the "R" , observe how it's completely absent from field 1.  There's no way a single field can be compared to bob in terms of resolution.

Its an interesting comparison, but you’ve completely overlooked several things that make it a completely irrelevant representation of what happens in the real world. Here’s why:

1. You have simply laid 2 odd/even fields over the top of each other in their interleaved form (all the original odd/even line structure intact with the addition of simple line doubling). If anything this is more of a “semi-interlaced” representation, as the original odd/even structure remains intact. Therefore, yes you are still perceiving some of the detail onscreen at the same time. In this case I would actually agree that there isn’t a 50% reduction in detail, but there is certainly still a drop in vertical resolution when compared to weave, and lets no get started on the bobbing artefacts which anyone would agree are horrendous. Hence the reason drop field technique is often favored for film sources. However unlike your example this wouldn’t appear this way in the vast majority of cases on real world progressive displays. Here’s why:

2. You aren’t taking into account the uneven interpolation that takes place on something like a 768p plasma, where separate fields are scaled directly to complete frames. In this interpolation process, the interleaved line structure is destroyed and the frames are laid over the top of each other. This destroys the perception of the fields being woven together, and they are now separate entities.

Ikari said:

As for the engineers at HQV, they truly have no idea what they are doing if they think it's acceptable to throw away every second field.  Yes, I'm criticising professionals who have more experience than myself, but you know what, even the professionals get it wrong.   Look at how many pros get it wrong thinking it's acceptable to use drop field on SD shows to make them appear to be film based.  Seven have got it wrong with using bob instead of adaptive motion de-interlacing.  I'm sure there are plenty of other stuffups that you can think of that have been made by industry pros.

LOL. Ok, now I can’t help but have a chuckle at this one. Again I’m afraid this clearly illustrates that you aren’t doing your research. I cant believe you even typed this.

1. For starters HQV isn’t the company, it’s the brand name for the consumer version of Silicon Optix’ advanced video processing chipset (it stands for Hollywood Quality Video in case you’re interested).

2. Silicon Optix NEVER use any kind of bob processing, drop field or not. What on Earth gave you that crazy idea?!! Why on Earth would cwt and I have been commenting on it’s virtues if it was simply bob?!! Do you honestly think that the world’s most sophisticated video processing would fall back on such archaic inferior single field processing?! On the contrary HQV used “per pixel multi-field analysis motion adaptive de-interlacing”, that takes into account data from several preceding and succeeding fields, and uses an advance combination of weaving and interpolation (plus diagonal interpolation and motion adaptive noise reduction) depending on the source. There isn’t any known cadence it won’t lock onto and then apply the best possible method of de-interlacing. Bob it aint.

If you’d spent even 5 minutes at the HQV website you’d know all this.

The bit you’re obviously getting confused with, is that HQV argue against all forms of bob processing that are often used by their competitors when processing HD video. I even quoted a section from their technology FAQ above that clearly mentions this!! Did you even read it? Here is it again for you. I’ve even included the section about Silicon Optix’ own processing so you can compare the difference in technology for yourself:

HQV Technology Guide said:

Simplest Competitor Approach (Non-Motion Adaptive):

The simplest approach to avoid these artifacts is to ignore the even fields. This is called a non-motion adaptive approach. In this method, when the two fields reach the processor, data from the even fields are completely ignored.

The video-processing circuitry recreates or “interpolates” the missing lines by averaging pixels from above and below. While there are no combing artifacts, image quality is compromised because half of the detail and resolution have been discarded.

More-advanced techniques have been adopted by virtually all standard-definition video processors, but this basic approach is still sometimes used for high-definition signals, due to the increased computational and data-rate requirements of higher video resolution.

With video processors from some competitors, only 540 lines from a 1080i source are used to create the image that makes it to the screen. This is true even for video processors from companies that may have been considered providers of flagship performance in the standard-definition era.

Advanced Competitor Approach (Frame-based Motion Adaptive):

More advanced de-interlacing techniques available from the competition include a frame-based, motion-adaptive algorithm. By default, these video processors use the same technique described above. However, by using a simple motion calculation, the video processor can determine when no movement has occurred in the entire picture.

If nothing in the image is moving, the processor combines the two fields directly. With this method, still images can have the complete 1080 lines of vertical resolution, but as soon as there is any motion, half of the data is discarded and the resolution drops to 540 lines. So, while static test patterns look sharp, video does not.

Frame-based motion-adaptive techniques are now common in standard-definition video processors. However, this is still rare in high-definition video processors due to the computational complexity of even frame-level high-definition motion detection.

Silicon Optix HQV Approach (Pixel-Based Motion Adaptive):

HQV processing represents the most advanced de-interlacing technique available: a true pixel-based motion-adaptive approach. With HQV processing, motion is identified at the pixel level rather than the frame level. While it is mathematically impossible to avoid discarding pixels in motion during de-interlacing, HQV processing is careful to discard only the pixels that would cause combing artifacts. Everything else is displayed with full resolution.

Pixel-based motion-adaptive de-interlacing avoids artifacts in moving objects and preserves full resolution of non-moving portions of the screen even if neighboring pixels are in motion.

“Second Stage” Diagonal Interpolation:

To recover some of the detail lost in the areas in motion, HQV processing implements a multi-direction diagonal filter that reconstructs some of the lost data at the edges of moving objects, filtering out any “jaggies.” This operation is called “second-stage” diagonal interpolation because it’s performed after the deinterlacing, which is the first stage of processing. Since diagonal interpolation is independent of the de-interlacing process, competitors have used similar algorithms with their frame-based de-interlacing approaches.

But hey, you know better than the guys “at HQV” right? The people that make world famous $100,000 broadcast quality processors like Teranex. Even though you haven’t even read about their own technology properly, and argue that bob is an acceptable method of de-interlacing! And of course you know better than all the A/V professionals across the planet that are in 100% agreement that weave is by far the preferred method of de-interlacing for interlaced material sourced from film

Ikari said:

I'm all for best quality images, always have been.

Well that comment is completely contrary to many of the things you’ve stated in this thread. If you were “all for the best quality images” then the word bob would never enter into your vocabulary unless used alongside the word “sucks”.

Now that we’ve got all that out the way can we please drop this completely pointless ludicrous discussion? It’s getting very old.

[pneu]

Darklord, on Apr 18 2006, 11:41 PM, said:

Well at least I got a little “plus artefacts” comment out of you (understatment of the year)! I can’t believe you are actually admitting bob looks nasty! Regardless of any disagreements we have about “discarded resolution” there is certainly no denying that bobbing results in a horrendous image looking at your lovely examples

No disagreement there, bobbing generally looks unpleasant.  There's always a better method than bob, I've always been aware of that.

Darklord, on Apr 18 2006, 11:41 PM, said:

2. You aren’t taking into account the uneven interpolation that takes place on something like a 768p plasma, where separate fields are scaled directly to complete frames. In this interpolation process, the interleaved line structure is destroyed and the frames are laid over the top of each other. This destroys the perception of the fields being woven together, and they are now separate entities.

The example in the gif lays the bobbed frames over one another and clearly there is perception of both fields.

For 768p displays, I believe the interpolated lines would be added at repeated intervals so it does preserve the interleaving structure overall.  I mean if the bobbed fields were so far out of whack there'd be such a crazy amount of flickering going on.  A fine edge of text bobbing up and down could be taking place over several rows.  That doesn't happen.

Or have you observed fine detail jumping up and down over several rows of pixels on a 768p display that bob de-interlaces native 1080i video?

Darklord, on Apr 18 2006, 11:41 PM, said:

Silicon Optix NEVER use any kind of bob processing, drop field or not. What on Earth gave you that crazy idea?!!

Because I read your opinion saying drop field was better to avoid the bob flicker and thought you were inferring that's what the entire article was about.  My mistake.

btw I have actually managed to read quite a few articles on video standards over the last few years so I'm not a totally ignorant slacker

Darklord, on Apr 18 2006, 11:41 PM, said:

If you were “all for the best quality images” then the word bob would never enter into your vocabulary unless used alongside the word “sucks”.

I always said it sucks but how much it sucks is the question.

Darklord, on Apr 18 2006, 11:41 PM, said:

Now that we’ve got all that out the way can we please drop this completely pointless ludicrous discussion? It’s getting very old.

Yes.  I feel like this whole argument was pointless because we already knew from the beginning that adaptive motion de-interlacing is the best for 50fps sources and weave the best for 25fps.

[pneu]

Ok here's an example I made that demonstrates how both fields are perceived when bobbing 1080i to 1366x768.  It shows what's happening at the single pixel level.

http://members.optus...c54/compare.htm

[pneu]

So....it would seem that with both methods of de-interlacing 1080i to 768p, there is a loss of 50% of the original resolution, simply because 768p is 1 megapixel, and 1080p is 2 megapixel.  I mean, we perceive the detail of both fields in the bobbed version but it's still only 768p worth of resolution.

To sum it up I would say that weave & downscale of a 1080i25p source averages the pixels spatially (cos the 1 pixel high alternating white & black in the example have become grey), and bobbing averages them temporally (flashing white and black @ 50hz).

Another thing that occurred to me is that the size of 1 pixel on a HD display is so small that 1 pixel alternating on and off @ 50hz doesn't stand out as badly (certainly not as much as 1 pixel from a 576i source would on a large screen), probably the reason why I can't notice bob flicker with 1080i sources on HD plasmas.

Oh and just so I don't offend anyone, I do recognise that bob is an inferior method of de-interlacing and that any other method will produce a significantly better image.

I'm interested to know what other people think about this.  Adam-O?