CML - Cinematography Information & Discussion
sponsored by:-
Images & Text Copyright © CML do not use without permission
|
||
|
1080i or 780P Published : 6th February 2005 A little help please....Gee is there ever some confusion with this issue both with what people say at conferences and what is written if you surf the net. Firstly, I am referring to distribution not acquisition as I don't believe there is any true 720/50p kit out there. Is there anyone who could help give me a really good and accurate explanation of the differences particularly with regard to what you guys in the US see when looking at HD broadcasts of sports, drama, feature films etc. Does sports from 720/50p really look a whole lot different compared to 1080i and a feature film that was shot at 24p but broadcast at 720 or 1080i can the viewer at home see a noticeable difference? If there is a source on the net that I haven't come across yet and you know to be 100% accurate then please let me know. Much appreciated Janet West Janet : Thomson LDK 6000 HD cameras have a 720P/50 mode. Haven't seen a 720P/50 recorder, but I suspect Panasonic has that format in their D5's and maybe in their HDDVCPRO recorders. Visual differences are subtle, but ANY Progressive capture will out perform Interlace when ever motion is present, especially vertical motion. It is a noticeable difference. Kell is a much disputed factoring guide to resolution, but it does favor Progressive. Interlace, when converted to Progressive suffers much more than the reverse, so there are real differences. You are welcome to contact me offline for further discussion. I was part of some very extensive format qualitative testing and comparisons a few years ago for Fox and a couple of U.S. Government imaging agencies, and have experienced the science and the psycho visual differences. George C. Palmer >Does sports from 720/50p really look a whole lot different compared to >1080i and a feature film that was shot at 24p but broadcast at 720 or >1080i can the viewer at home see a noticeable difference? [The writer is expresses confusion in trying to research , on paper, the difference between 1080i, 720p and 24P]. The Short Answer : The vast majority of live broadcast events (such as sports) in the US is broadcast in 1080i. Subjectively, 1080i looks like really, really, good, clear conventional TV. A couple of networks in the US (primarily Fox & ABC? if I recall) are broadcasting 720p. To the trained eye, it does not have interlace artefacts. Subjectively therefore, it looks more like your computer screen : i.e. no "jaggies" during a fast camera pan, and quite a bit sharper than normal TV. Your final question : [to paraphrase] "Does 24P originated material look different once converted to 1080i or 720p?" Answer : Absolutely. Just as film originated material looks different on TV, so does 24P, which to the trained eye looks almost like film, and to the untrained eye, looks like film. "If there is a source on the net that I haven't come across yet and you know to be 100% accurate then please let me know." REPLY : Aha!!! Now we are getting somewhere : The problem with "researching this on the net" is that it can't be done. There is so much blather on these and other pages, where people pontificate endlessly about megapixels, bit rates, etc. etc., where if you just SAW the video, you'd say: "Oh. That’s what that does (looks like, whatever)." You'd say: "I see that. I totally get it now." Trying to describe certain things verbally is very inefficient; often totally ineffective. Your best bet is to stop surfing the net, and find a way to go look. You'll see: 1080i looks like "This", (really good TV) 720p looks like "That" (really good TV without certain artefacts), and 24P looks like "The Other" (almost like film). Lew Comenetz Lew Comenetz wrote : >To the trained eye, it does not have interlace artefacts. Subjectively >therefore, it looks more like your computer screen : i.e. no "jaggies" >during a fast camera pan, and quite a bit sharper than normal TV. This is only the case if you happen to be looking at an HD monitor that displays both formats natively. Very, very few consumer sets do. The vast majority convert everything to 1080i/60 for display. Under these circumstances (which, as I said, are by far the most common) the two look essentially identical to nearly anyone. Mike Most Mike Most wrote : >The vast majority convert everything to 1080i/60 for display. Under these >circumstances (which, as I said, are by far the most common) the two >look essentially identical to nearly anyone. Mike My set does this, i.e., convert everything to 1080i/60 (Mitsubishi 55") and I see a definite difference in detail when watching Basketball and Football in sharpness and depth with 1080 winning hands down, When I have seen the same comparison on a Panasonic Plasma displaying only 720P from both sources the 720P gets the edge. Michael "loves 1080i TV" Bravin Michael Bravin wrote : >My set does this, i.e., convert everything to 1080i/60 (Mitsubishi 55") and >I see a definite difference in detail when watching Basketball and >Football in sharpness and depth with 1080 winning hands down… Even if this is true (and I'm not sure I agree that it is), any differences are much less apparent in 24fps originated entertainment material, whether it's shot on film or HD video - which is what I thought was being discussed, although I could have misinterpreted that. Mike Most M Most writes : >Even if this is true (and I'm not sure I agree that it is), any differences are >much less apparent in 24fps originated entertainment material, >whether it's shot on film or HD video I can easily tell the difference between, for example, CSI-NY (1080) and Lost (720), even on my Powerbook with a 1280x854 screen. CSI is super crisp and detailed, truly amazing actually, and Lost, while one of the cleaner 720 shows, is noticeably soft, especially on wide shots. On a larger monitor or the projector it's even more so. I vote we start calling 720p "Medium Definition". Tim Sassoon Yes I agree, as a delivery format, 1080i displayed on a native 1080i display is as Tim says, "super crisp and detailed". 720p, on a 1080i native display, at least what I see converted on a Dishnetwork model 6000 set top box, (internally converted to 1080i from the 720p being received) is just a step above 480p, and does not come close in HD wow factor when blown up on larger size screens. Also HD through the terrestrial tuner module in the model 6000, I get the same effect when watching KABCs 720p transmissions, (over the air DTV). I don't mind 720p displayed on a 720p native display, that looks fine and can stand on its own as an acceptable "Medium Definition" format. Oh, come on, Tim. You're comparing different productions with different approaches. All of the CSI's (all of Bruckheimer's shows, in fact) have a signature look that depends on contrast and sharpness. "Lost" is shot primarily on location in differing weather conditions (they're on Oahu - the weather changes every 30 minutes) and is more of a dramatic fantasy. Watch "NYPD Blue" sometime. I've seen the 1080 and 720 masters side by side and even I couldn't really tell the difference. Besides, comparing the differing looks of dramatic productions is no way to compare HD systems. In almost every way, Michael's comparison of sports events is a Mike Most Mike Most wrote : >Even if this is true (and I'm not sure I agree that it is), any differences are >much less apparent in 24fps originated entertainment material. Mike, you are right about the 24fps originated material showing much less difference but it is still there ABC vs CBS. But about being a doubting Thomas I invite you over to my house in Burbank to see for yourself. (You bring the beer)and any other CML folks too. Heck maybe we do a CML mixer at Michael's Mike also wrote : >I've seen the 1080 and 720 masters side by side and even I couldn't >really tell the difference. Masters? what does this have to do with watching HDTV? My system is 19MB/s off Dishnetwork and off a terrestrial antenna. When I take home a Sony F500 and pump it through my system it looks a whole lot different. I think we ALL need to make distinctions about what is ideal and what is real. For me at home, the HDTV I own is native 1080/60, my choice after seeing MANY MANY sets and screens. I do like 720P on a Panasonic Plasma or LCD but I liked the Mitsubishi set best all round. Since most people do not own a 720P native set we must focus on how the TV product we are involved in is viewed. Reality is a great equalizer. Michael "not kidding about the invite" Bravin I think that if we limit our initial choices for the best possible image acquisition tools by acquiescing to the dismally compressed and artifacted 19.2 MPEG "drek" (pardon the spelling) seen at home as "good enough" we might as well shoot 480P. Even for television, the archival quality of a show is just as, if not more important than the end use, and if we as professionals limit our "good enough" to that of the end of a narrow pipe, well then maybe we aren't "good enough" as professionals. George C. Palmer >You're comparing different productions with different approaches. All of >the CSI's... have a signature look that depends on contrast and >sharpness... You're absolutely right, but one has to start somewhere, and I pulled those two out because they seem to me to be emblematic, state of the art shows (I like Lost's mix of Gilligan's Island and Jurassic Park). I'll stick with my thesis, though; that there is a perceptual difference, and that 720p should more accurately be labelled Medium Definition (and I'm de-interlacing the 1080 on a progressive monitor, so that's only half-res vertically, but it is OTA ATSC, not cable/sat). Anyone have another 720p show that they feel really shows off the format? I don't get to watch as much TV as I'd like, but wouldn't mind being proved wrong. >When I take home a Sony F500 and pump it through my system it looks >a whole lot different… Equally true both 1080 and 720, and the sooner NTSC fades away, the sooner the DTV product will be taken more seriously. I'm frankly shocked at some of the images local/networks think are acceptable; I can only assume that they're primarily monitoring in SD. Also, one of the perceptual problems of 720p is the amount of SD upconversion mixed into it. While there are SD spots in the 1080 programs, the difference is more obvious. Tim Sassoon George wrote : >about acquiescing to the dismally compressed and artifacted 19.2 >MPEG "drek" (pardon the spelling) seen at home George Janet West’s particularly cogent and original question was : -differences particularly with regard to what you guys in the US see when looking at HD broadcasts of sports, drama, feature films etc. Does sports from 720/50p really look a whole lot different compared to 1080i and a feature film that was shot at 24p but broadcast at 720 or 1080i can the viewer at home see a noticeable difference? We were talking about television at home as it exists today, I hope it didn't come across as a discussion about lowering standards or producing for the lowest common denominator. We know these are your pet peeves, mine too. We were discussing the real differences in viewing 720P and 1080i originated material on TV. What are your thoughts regarding that? Michael Bravin Michael Most wrote: >You're comparing different productions with different approaches I agree. To do a meaningful test, you need to eliminate all the A more accurate comparison of the two as _acquisition_ formats would be to shoot the same subject uncompressed, uprez the 720 with a high quality converter, edit and compress each identically, and then compare the two on a consumer 1080i monitor. The same could be done by downrezzing the 1080 and comparing with 720 on a 1080i (and or 720) consumer monitor. Even then, there are loads of other variables, including lens selection, which could skew the results. I'm fairly certain that none of the consumer TV's have an actual resolution that is close enough to either format to tell the difference. Much of what has been discussed here is the performance of the built in consumer monitor uprez circuit, which I can assure you is a piece of .... Noel Sterrett I promised myself I wouldn't go any further than I already have, but if you are going to start somewhere, it might be a better idea to start by comparing apples-to-apples and differing production styles usually do not yield apples-to-apples comparisons. So you are, of course, welcome to stick by your thesis, but please don't presume to assign labels such as Medium Definitions. Labels, like knick names, tend to stick in peoples minds; and in this case that label is absolutely inaccurate. You should be aware that 1080i is already pshcyovisually deinterlaced by the human eye, and, based on fairly reliable observational and scientific experience, the human visual perception system throws away as much as 40 to 50% of the vertical resolution of interlaced content, and when that content is further resolution converted (down rezzed to 480 or deinterlaced for any other reason, line disseminated, or interpolated) the vertical resolution is further altered, negatively. No matter how 720P is scaled it is still progressive; inerpolation, down conversion, and upconversion have much less effect on the content, and only interlacing has a truly negative effect. This is not my opinion. All of this has been borne out over and over again in testing by Fox and multiple high level imaging studies. Fox even did a test wherein they captured 1080i, 720P, 480P, and 480i (all 60 frame) images simultaneously with match framed shots using cameras all from a single manufacturer. That content was shown in the blind (with no formats discussed or labelled) on multiple multi-format HD monitors all from a single manufacturer, of appropriate HD viewing size, and at optimum viewing distance, to multiple focus groups, some entertainment industry, some "non-industry", and mixed groups, and content was shifted to different displays for different groups to avoid any conjecture or discussion of which monitor or viewing position was the best. Almost unanimously, 720P was selected as the "highest" resolution; second in the hierarchy (almost unbelievably) was 480P30, which contained a 2:1 pulldown (not 60 frame), third was 1080i, and lowest, of course was 480i (digitally captured and recorded). Other tests I have seen were done differently, but all yielded similar outcomes. You're absolutely right, but one has to start somewhere, and I pulled those two out because they seem to me to be emblematic, state of the art shows (I like Lost's mix of Gilligan's Island and Jurassic Park). I'll stick with my thesis, though; that there is a perceptual difference, and that 720p should more accurately be labelled Medium Definition (and I'm de-interlacing the 1080 on a progressive monitor, so that's only I don't get to watch as much TV as I'd like, but wouldn't mind being proved wrong. George C. Palmer For me at home, the HDTV I own is native 1080/60, my choice after seeing MANY MANY sets and screens. I do like 720P on a Panasonic Plasma or LCD but I liked the Mitsubishi set best all round. Since most people do not own a 720P native set we must focus on how the TV product we are involved in is viewed...We know these are your pet peeves, mine too. We were discussing the real differences in viewing 720P and 1080i originated material on TV. What are your thoughts regarding that? George C. Palmer >You should be aware that 1080i is already pshcyovisually deinterlaced >by the human eye, and, based on fairly reliable observational and >scientific experience, the human visual perception system throws away >as much as 40 to 50% of the vertical resolution of interlaced content. If you don't mind me asking - who did the study, who funded it and where were the results published? Jessica Gallant >My set does this, i.e., convert everything to 1080i/60 (Mitsubishi 55") and >I see a definite difference in detail when watching Basketball and >Football in sharpness and depth with 1080 winning hands down Interesting, I must admit I haven't been able to see any difference at all. Are you watching directly off-air or by cable? Maybe my cable system... Err... "smooths out" the differences. Isn't "Joan of Arcadia" shot on Varicam? It would still be broadcast 1080i on CBS so our sets wouldn't be doing the upconversion - it would be interesting to compare that with a drama shot with an F900. >acquisition tools by acquiescing to the dismally compressed >and artifacted 19.2 MPEG "drek" (pardon the spelling) seen at >home as "good enough" we might as well shoot 480P. Agreed, if I bring home a J3 deck and connect it to the component inputs of my HDTV, playing back 30p footage from my IMX camera, it's actually very hard to tell apart from the HD I get on my cable system, and I know THAT ain't right. George Hupka Satellite Via Dishnetwork and off Air both. Joan of Arcadia on Varicam No No No its shot with the F900, It's Columbia TriStar a Sony Pictures Company Michael Bravin >Joan of Arcadia on Varicam No No No its shot with the F900, Hmm, I could have sworn I read somewhere that it was Varicam...Well, there you go, always get that information verified by multiple sources! Just flipped by the baseball game on Fox - it doesn't look as good to me as the 1080i baseball games I saw during the regular season. Not as sharp, and there's something really unattractive going on in the highlights (esp. Yankee uniforms are clipping rather badly, even the highlights on the Boston uniforms are clipping) which I hope can be attributed to the signal taking too many detours on its way to me. There's also a lot more blocking going on than I saw in regular season games, so maybe the Canadian network isn't getting the best feed from Fox..... I'm watching only by digital cable box... George Hupka >> Joan of Arcadia on Varicam No No No its shot with the F900, At their NAB press conference, Sony had one of Joan's Exec Producers talking about how they shot the show. I doubt they would have had him speak if he was shooting Varicam. A good looking show, too, IMO. Jim Feeley Thanks folks for your comments. As some of you know me off-cml site and trust I'm not quite as naive as the posting implied, I was looking for confirmation with the sort of replies you gave. The confusing issue in Europe at the moment is that there are a lot of people who can't differentiate between acquisition HD formats and transmission HD formats/progressive or interlace etc and forget that you might create in one, edit in another and then transmit in another. Although as you all know there is a lot of HD production in Europe there are currently only two HD channels from Euro 1080 who TX 1080i, TF1 in France just announced last week at MIPCOM they will go 1080i and others struggle to decide while being bamboozled by the various manufacturers each with their own camp of thought. Perhaps eventually we will go 1080p so this begs the question what makes the better halfway stage for TX now ... 720p/1080i? My start of this thread was to seek your comments primarily from the US as you have both TX formats, as a lot of folk make decisions on reading untruths (IBC Daily's Monday 13th Sept issue describing the 1920 x 720p format on the front page????), hearsay, and techy theory rather than just what looks the best. After all it's not HD or any related technology come to it that makes for high TV ratings or a box office hit it's a good story and managing the compromises you have to make during the production process and how good it looks when it's finished and broadcast. Janet West I see a noticeable difference between 720p and 1080i (or p). 1080 looks really fine and textured, revealing incredible detail. 720p just looks like sharper television. Sadly, I like the color I see coming out of a certain 720p camera more than I do the most common 1080 camera. I'd really like to see a mixture of the formers color and the latter's resolution. Art Adams, DP [film|hidef|video] >second in the hierarchy (almost unbelievably) was 480P30, which >contained a 2:1 pulldown (not 60 frame), third was 1080i Unbelievably is right. I wouldn't give that person back their car keys until they got in the taxi. >please don't presume to assign labels such as Medium Definitions. I'm just a pixel mechanic who gets to shoot some now and then. I don't dream of exerting such influence. >You should be aware that 1080i is already pshcyovisually deinterlaced >by the human eye That's a world-class typo, there. I don't necessarily accept the intended conclusion as fact. >the human visual perception system throws away as much as 40 to >50% of the vertical resolution of interlaced content. I can hear Larry Thorpe sobbing into his beer at this news - kinda throws a wrench at the whole "24psf is as good as progressive" argument, dontcha think? >That content was shown in the blind (with no formats discussed or >labelled) on multiple multi-format HD monitors all from a single >manufacturer, of appropriate HD viewing size, and at optimum viewing >distance If they weren't shown on proportionally-sized monitors, then it wasn't any kind of test at all. Constant degrees field of view per pixel horizontally. One can easily design conclusions into experiments. I would reserve judgement on counter-intuitive results until one can examine the methodology. http://c2.com/cgi/wiki?AdVerecundiam Lew Comenetz writes : >So it is with this Hi-Def stuff: People are reading and theorizing about >720P 1080i, up-rezz, down -rez, "How many megapixels can dance on >the head of a pin." Its counter productive ..and I like Lost better as a show. At first I thought it was Gilligan’s Island meets Jurassic park, but this week was more like Office Space meets Forbidden Planet. Monsters from the Id filling out TPS reports. Just a few soft shots this week, too. Tim Sassoon Funny ... just before my post regarding Baseball's AL playoffs on FOX, a comment was made about the highlights, like shoulders on uniforms. I was involved with the uplink of that signal, but I only turned around a bunch of digital data, having taken an ASI stream and sent it to the satellite. I had no artistic input myself. However, I used to run video in those trucks. The cameras used there are the Thomson triax based HD cameras. They are able to make very pretty pictures. As you know, the guy running the knobs is the other half of the equation. Can't tell you the times I would get in a truck as a 2nd video guy, and see the SVO turn OFF all the knee circuits. Just a quick hard clip at 103%. Football, baseball, hockey, basketball --- it didn't matter. I actually had people ask me if I was doing particular games as SVO when they'd watch at home, and see detail in shoulders, or clouds in the sky with some shape. Some people did notice... Did the best guys do any more justice to highlight detail?? Not on your life. The most respected sports guys : 1/. Know the game - iris is set almost before the camera pans or pushes in for the tight shot. They're fast because they know what's happening next. 2/. Know the setup - cameras, LPS cams, coaches ALL-22, booths that are becoming more complex, telestrator, etc... 3/. Most of all, they are respected by the crewing people in LA or New York. If a top Senior Video guy has all that, he's working 200~300 days a year. Chances are the highlights you saw at home are about what left the truck. But sports ain't about art ... ?? Would it do any good to write the network, telling them it looked crappy in places and why?? FOX is very careful to check all things technical prior to air - you wouldn't believe the technical checking. The people at MCR take pride in what they do. Perhaps a few comments would help. THEN, how do you begin to retrain a 20 or 30 year veteran? Oh well - in time, maybe. Kevin Stebleton >second in the hierarchy (almost unbelievably) was 480P30, which >contained a 2:1 pulldown (not 60 frame), third was 1080i Sorry this piques your sensibilities, but this was an interesting reality check for all involved with the test. I can assure you that no one went out of their way to find biased, mentally challenged or driving impaired individuals for the test and the test was not done by the blind or for the blind, just with a cross-section of folks ranging from technical to pure viewer types. I can appreciate your sincerely held scepticism, but predisposition, in this case, was interestingly, and surprisingly over-ridden by observational data.. >I'm just a pixel mechanic who gets to shoot some now and then. I don't >dream of exerting such influence. But you did apply the label (Medium Resolution), so I presume this is false modesty. Don't short yourself. You seem to be a very smart and thoughtful guy. >You should be aware that 1080i is already pshcyovisually deinterlaced >by the human eye >the human visual perception system throws away as much as 40 to >50% of the vertical resolution of interlaced content. I presume you meant to suggest a world-class error in proposition, but, nevertheless, pretty much everyone, even Larry Thorpe, now accepts the fact that interlace is inferior to progressive imaging. Whether he is sobbing into his beer or not and whether or not you accept the science or the evidence or not does not change the entirely logical conclusion that interlace is motion resolution challenged. In fact, before Sony discovered 24psf, their official line was that interlace (1080i) was superior , but after 24psf, they touted 24P as the superior high quality (digital cinema) acquisition format >...kinda throws a wrench at the whole "24psf is as good as progressive" >argument, dontcha think? I'm sure that was a temporary lapse on your part, as I'm sure you know that Sony maintains that 24psf IS progressive, not interlaced. >That content was shown in the blind (with no formats discussed or >labelled) on multiple multi-format HD monitors all from a single >manufacturer, of appropriate HD viewing size, and at optimum viewing >distance >If they weren't shown on proportionally-sized monitors, then it wasn't any >kind of test at all. Constant degrees field of view per pixel horizontally. It's good that reasonable people can disagree in theory about just about anything, but since this was a discussion of the relative merits of the two formats as displayed at home, as was the purpose of the test to ascertain, it was reasonable to use multi-format monitors of the same size, in fact, if anything, 1080i should have outperformed all others since the monitors were actually Panasonic HDTV monitors whose native display was interlaced; all progressive playbacks were mezzanine converted in each monitor(and, as a result) interlaced for display purposes, and suffered from the interlacing process. So the conclusions are even more compelling, since interlace had the display "advantage" and because they (the conclusions) didn't rely on opinion, predisposition, or even any particular technical knowledge, only direct observation. George C. Palmer For those of you who question the substance of my views on the subject, I offer the following. While I worked for Philips Broadcast, I was the liaison for what was then their first HD cameras to Naval Research Laboratories, which was the lead "think-tank" imaging agency in a consortium of other agencies (all DOD associated) that was determining a common format for the Department of Defence. For about 10 years they did theoretical and practical HD imaging acquisition tests and demonstrations. Some of the most advanced minds from places such as MIT and CALTECH were brought to bear on the subject through grants from several grant agencies. Their specific data will never be made public, not for any particular reasons of secrecy, but because they never intended to influence the private sector in any way, intended the research to influence only their own conclusions, and because the compilation of their work would probably have hugely multiplied the cost of the project, but I have talked to multiple, credible imaging experts (not manufacturing employees) who unanimously affirm that their 720P vs 1080i image gathering exercises (which used identical cameras which differed only in the sensors employed) confirm what they all understood (academically) to be true; that interlacing introduces line-to-line temporal discontinuities which confuse the human visual processing system to the extent that it actually discards resolution based information that results from that confusion. I don't have access to their research, but I can assure you that these are folks of stature in the world of visual science academia, who actually tested their academic knowledge and whose credibility was adequate to convince the Secretary of Defence (Secretary Cohen, during President Clinton's Presidency) and his staff of the effecacy of progressive imagery. If you go to this link, you can see their statement : http://www.atd.net/HDTV_faq.html I am, frankly bewildered why, when one advocates progressive over interlace, EVERYONE asks them to justify that proposition with scientific evidence, yet the seeming interlace devotees are NEVER asked to provide such proof for their allegations of superiority. I think that if anyone requires pragmatic evidence that progressive is superior, just look at the world of graphics and animation, where progressive is the exclusive creation format and where the use of interlace is anathema and counter-intuitive, and I also haven't seen any actual demand for, or evidence of, any HD image format testing or research that lead to their 1080i choice. In their defence, they were, most likely, strongly assured by their primary manufacturer of choice (Sony), at the time of their choices, that interlace was the best format (Sony later partially recanted that position, when their introduction of psf required them to staunchly support progressive as a superior image capture medium). At the time, however, it was no accident that both Sony and others had made huge commercial and national investments in interlaced HD technology (MUSE). If interlace had been universally accepted, the return on their national, commercial investments would have been magnified even more than it has, subsequently. Fortunately, now all manufacturing countries and companies therein have proven less parochial in their HD format development positions, and have been instrumental in providing HD equipment that also enables implementation of both progressive and interlace broadcast formats. I am sorry, I can't provide you with specific documents, but I would suggest that a good search of motion resolution studies on the net would reveal and confirm the conclusions of those that have taken the time to (in "Role-Model's" words) "test it" for themselves. George C. Palmer Flame on George George George I am sure it warms Sony and Larry Thorpe's heart to know that you are their spokesperson here on CML. Also I wonder who at Sony "discovered" 24PsF as you claim. Oh that's right you can't produce the documentation. You are back to weasel words my friend. Why can't you see this for yourself. Many here on CML are way past being on to you. Why do you so often work so hard to tarnish the respect that I and so many have for your knowledge and experience. Flame Off Michael Bravin Michael Michael Michael: And I thought we were doing so well. Reading carefully leads to accurate understanding of the words. I never presume to understand Larry's mind; never met anyone who could. Don't care how they "discovered" it. The issue was interlace vs. progressive. Larry and Sony's public (not at all hidden) was very clear that interlace was superior right up to the point of the introduction of 24psf. That was all public knowledge and contained in Larry's oft presented white papers. So where is the weasel? I think you need a cleaner mirror, my friend. George C. Palmer George wrote : >Reading carefully leads to accurate understanding of the words. You bet George couldn't AGREE with you more and we ALL can read exactly what you wrote. Back then Larry publicly said that 1080i IS superior to 720P. When 1080 24PsF was DEVELOPED by Atsugi engineers he said that this format was superior for motion picture production. Which based upon how it has played out since then, turned out to be VERY accurate. In spite of the studies you sited (undocumented). So as always the devil is in the WAY YOU say what you say and the WORDS YOU use. Deflecting it back on me is just more of the same. Like I said before many here are on to you. I merely speak up. In spite of this I still respect you as one of our industries most experienced and knowledgeable engineers and pioneers when you aren't tilting at your ever present personal Sony windmill. BTW Larry if you are out there I do not presume to speak for you. Michael Bravin I’ve looked at some recent tests on a 1920 x 1080 50" plasma display. The initial tests seem to show that 720/50P out performs 1080/50i for a given data rate. Whilst I haven’t seen any results yet 1080/50P should out perform 720/50 for the same data rate. Has anyone run any tests against H264 and Microsoft coding for any tests in the area between 720/50P and 1080/50i ? thanks, Dave Blackham >Has anyone run any tests against H264 and Microsoft coding for any >tests in the area between 720/50P and 1080/50i ? "Microsoft"? Are you referring to "VC-9"? Jeffery Haas >I’ve looked at some recent tests on a 1920 x 1080 50" plasma display. >The initial tests seem to show that 720/50P out performs 1080/50i for a >given data rate. It's almost a sure bet that your 50 inch plasma display is not showing 1920 x 1080 material at pixel for pixel. The most I've been able to find in a large plasma is about 1500 to 1600 pixels wide. So how is 1920 x 1080 being resampled for display on your 50 inch monitor? How is 720 being resampled? Dave Stump >I'm sure that was a temporary lapse on your part, as I'm sure you know >that Sony maintains that 24psf IS progressive, not interlaced. There are interlaced (I) sources, interlaced refresh monitors, progressive (P) sources, and progressive refresh monitors. Mixing and matching these will give dramatically different results. Now, I haven't won any Oscars for work at Triple-I, named Pixar, or invented the alpha channel; just a bit of observational experience. 24psf is essentially the same as an integral frame in a film transferred to video. The two fields are of the same frame, contiguous, though processed as fields by the display device. "True" P images are processed and displayed as an integral frame at all points. P is absolutely the ideal way to display images. Here's where the testing you've outlined falls down. You imply that 1080i, 720p, and 480p were blind-tested on multi-format monitors of the same size. The problem with I monitors is that fields are shown sequentially, creating "Twitter" artefacts on images at the scan resolution. It's an analog Nyquist limit, the standard cure being to reduce vertical resolution. The higher resolution the image, the more it would suffer from twitter and moire artefacts on an I display. Thus, a 480p image might look "better" than a 1080i image. But it's an illusion - the 1080i image is being sold short by display artefacts. And if presented on a same-size display, one tends to show only problems, not benefits. You say there is a loss of vertical resolution on I _displays_. I agree, but this is not necessarily true with an I source on a P monitor. For best case, take a held card. Though captured I, if displayed on a P monitor, it would be indistinguishable from a true P image. Worst case would be an active sports shot, with a lot of inter-frame motion. An I image shown on a P monitor would show a lot of I combing; i.e. the two fields superimposed. Regardless of deinterlacing method, half the vertical resolution is lost. But it doesn't matter. The more motion there is in a scene, the less resolution is needed. Check out a frequency plot of a scene with sudden motion. An order of magnitude drop is common in anything fast enough to generate significant inter-field spatial displacement (or, motion blur). Most scenes contain low to moderate motion. With proper processing, only those areas of the frame will lose resolution. Conclusion: Interlacing sucks. But, with the combination of adaptive deinterlacing and progressive display, 1080i looks much better than 720p. Tim Sassoon "interlacing introduces line-to-line temporal discontinuities which confuse the human visual processing system to the extent that it actually discards resolution based information that results from that confusion." I'm wondering what I said in the last post on this subject which stopped the conversation dead? No offence or anything personal was intended by me at all. Be that as it may, I accept the above premise, in as much as it applies to interlaced _displays_, which can wreck progressive images as well. IMHO interlaced footage on progressive displays is really a software problem, and the hardware market is becoming more progressive, not less. Last night I watched and recorded NYPD Blue, in all its 720p glory, the first time I've seen it in "HD". I didn't see an AC listed, but the focus pulls were nothing short of heroic. It looked sharp, clean, and clear, the best thing I've seen so far in 720p. "720P was selected as the "highest" resolution; second in the hierarchy (almost unbelievably) was 480P30... third was 1080i, and lowest, of cause was 480i " That may have been then, with the displays and image source available at the time, but this is now. IMHO, anyone who seriously thinks that 720p challenges 1080i for image quality really needs to look closer at the pictures, on as neutral a monitor as possible. I recorded the transport stream of NYPDB, CSI-NY, LAX, Lost, and the Tonight Show, and looked at them using VLC and EyeTV controllers with Apple LCD monitors. Conclusions : A/. Hands down winner for sharpness and apparent detail was LAX, closely followed by CSI-NY, which suffers from excessive grain and compression artefacts. Worst was Lost. This was true even on a 1280 display; a 1920 display only made it more so. B/. Comparing 1080i de-interlacing (VLC offers Blend, Bob, Discard, Linear, and Mean, EyeTV "Adaptive"), no significant resolution loss was observed, except with Disable and Discard, though Bob had a lot of jitter. Mean looked the best in VLC, though EyeTV's Adaptive was the least obtrusive and sharpest. It failed, though, to properly de-interlace strobe, rather than motion differences (flashing light). Neither does a true reverse telecine, dammit. C/. There were plenty of problems with NYPDB 720p as well. First up, the title sequence was very poorly converted from SD. No attempt had been made to separate fields, and the comb structure was very evident, though doubled, so it couldn't be interpreted properly on any system. I assume this section would look very bad on an interlaced monitor. Not inherent, just a bad post decision. D/. I'm still trying to understand the 720p strategy for cine-expanding. Quite obviously, _three_ frames from the 24fps original are blended into a single progressive frame, in an attempt to simulate 2:3 pulldown, which seems to me to be no net gain from blend de-interlacing. Why the third frame (is it a true frame?), I don't know. By contrast, 1080i could be reverse-telecine to a true 24fps. Obviously, this is a problem of 24fps original, but it's worth discussing. E/. Sharpness of the 720p, while subjectively quite pleasing in motion, is significantly less than any of the 1080i samples when looking at stills, including Jay Leno, (electronic vs. film). When I say significant, I don't mean a small, arguable amount. I mean that it's really obvious. Another point: 720p is treated as though it's greater than half-way to 1080 from SD, when in fact it is not. Line chart 480, 720, 1080 and you'll see that 720 is below the median, not above. Using 486 only makes it worse. So, maybe we should call 720p "Slightly less than Medium Definition" :-) Tim Sassoon >significantly less than any of the 1080i samples when looking at stills I assume the stills would consist of two superimposed 1920x540 samples which would not, if there is any movement in the shot, mesh very well. Alternatively, if the still is a single 1920x540 field resampled to 1920x1080, there would be very little more data than in a 1280x720 sample. Noel Sterrett >I assume the stills would consist of two superimposed 1920x540 >samples which would not, if there is any movement in the shot, mesh >very well. Perhaps I was unclear. Yes, a 1080i image is two 1920x540 images interleaved (not superimposed). But, most images don't display much interframe motion, which is of course is the working basis of MPEG (my gut estimation is that we could consider most material to be 80% progressive). And if the area of motion that does exist is blended between fields, what you have essentially is motion blur, which has been proven acceptable to viewers :-) Ergo, satisfactory display of 1080i is mainly a software problem. If the entire frame is in motion, a line-doubled, not interpolated, 1080i only has 10% more information that 720p, but that would be worst case, worst technique. 24fps originated material should be reversible to perfect 24p with no interpolation. I would encourage everyone take an empirical approach on this. Find and carefully look at some images, and make up one's own mind based on what you see. Tim Sassoon >D/. I'm still trying to understand the 720p strategy for cine-expanding. >Quite obviously, _three_ frames from the 24fps original are blended into >a single progressive frame, in an attempt to simulate 2:3 pulldown No, that's not how it's done. The broadcast format is 720p/60. The pulldown is added on a full frame basis: 3 full frames followed by 2 full frames. This happens during the conversion from the 1080p/24 master, prior to network delivery. At least that's what's been done for the last 2 seasons.. Mike Most >we could consider most material to be 80% progressive That would also imply 20% non-progressive, or simply lower resolution. What happens is that as the shots transition from still (not very often since we're talking about moving pictures) to moving (camera, hands, lips) there is a clearly visible change in resolution. This change, after watching for a while, becomes noticeable, predictable and very distracting. IMHO this is far worse than a consistent resolution throughout. Noel Sterrett >What happens is that as the shots transition from still (not very often >since we're talking about moving pictures) to moving (camera, hands, >lips) there is a clearly visible change in resolution. I'm not seeing that phenomena in terms of de-interlacing, but I constantly see it as an MPEG artifact (bit-rate and encoding are hairball subjects unto themselves), though. One also sees it on material that has been sharpened using certain inter-frame processes. The best way to see the 1080i (and 720p) the way I'm seeing it is to get the MPEG transport stream into a computer, then play it into a full native resolution progressive monitor with software de-interlacing (VLC, etc.). I'm using a Mac PowerBook G4, and a G5, with an Elgato 500 receiver, though I could use Wintel/Linux, and use any ATSC receiver with a Firewire port (and VirtualDVHS, etc). Tim Sassoon P.S there may be some combination of non-computer based receiver and monitor that will do the same thing, but I don't know what it is. Neither of my set-top receivers will de-interlace to a 1080p monitor, AFAIK. P.P.S Tonight's Lost looked really good, but it's still 720p. >I constantly see it as an MPEG artefact What I was referring to is not an MPEG artefact. Earlier, you suggested that, "we could consider most (1080i) material to be 80% progressive". That implies 20% of 1080i material is not progressive (i.e., 540p). You added that "1080i only has 10% more information that 720p". That implies that 1080i (when moving) is essentially equivalent to 720p (moving or still). If 1080i (moving) intercuts well with 1080i (still), why would it not intercut just a well with 720p? Returning to the 80% figure, I must point out that we are making "moving" pictures. Even if the camera is locked down, something is usually moving or the audience soon will. It is what is moving that draws our attention. I really don't care if the background is progressive if the actors (who are rarely still) are interlaced. I think 80% interlaced would be more accurate. So if what is important in 1080i is 80% interlaced (540p), we are left with a 10% difference between 1080i and 720p. I am sure there are those on this list who can see a 10% difference in pixel count. I admit I am not one of them. I have struggled with 1080i (actually 1035i) since the first Sony 700 camera. Just because you can shoot yourself in the foot, apply bandages, take some pain killers and still be able to walk doesn't mean it's a good idea. Noel Sterrett >If 1080i (moving) intercuts well with 1080i (still), why would it not intercut >just a well with 720p? Changing the subject for a second, pro 720p decks like DVCPRO HD can output 720p50/60 recordings as both 720p and should it be necessary, as 1080i. 720p is a wonderful acquisition format for both 720p- and 1080i-based broadcasting for this precise reason. I still think it's sad that film-type content is often broadcast at 1080i60 with pulldown, essentially wasting 20% of precious bandwidth on material which basically gets thrown out by the deinterlacing, pulldown-removal processes in the display or receiver. Completely insane if you ask me. Rune Hansen >So if what is important in 1080i is 80% interlaced (540p), we are left >with a 10% difference between 1080i and 720p IMHO you're taking the absolute worst possible case and making it the norm. My experiences is suggesting a rosier outcome, as playback devices become smarter and more progressive. BTW, one conundrum has been solved, that of the phantom 3-frame pulldown in the NYPD sample. Credit goes to Phil Dubs of Pixelmetric (I'm at SMPTE) for identifying the cause as an overstressed MPEG encoder (many others in the MPEG stream testing community fluffed it :-). Obvious once said. The number of echoes equals the number of P frames in the GoP. We see three, so that's 2 P frames plus the I frame. It went out at 12.16 Mbps, which is pretty healthy, so either NYPDB should stop flinging their camera around, or the encoding needs to improve. I suspect that it was judged better to see edge repeats than for the swish pans to break up into piles of blocks; i.e. a quality floor was set. Tim Sassoon I know the initial question related to HDTV broadcast, but I do find Panasonic's move away from 1080i cameras interesting, such as the earlier AJ-HDC20A camera which is a 1920 x 1080, 3 x 2.2 million CCD's, with a 1000 lines of horizontal resolution. Jacques Nortier Jacques Nortier wrote :
Also interesting is the introduction by Panasonic last month (at the Digital Synthesis Symposium in Santa Barbara) of a new 1080i camera, apparently competing with the Sony HDW750. The new camera, identified as the HDX400, has a 720p front end but, according to Panasonic reps, records and outputs ONLY 1080i. When I asked why no 720p, I was told that the previous non-variable 720p camera did not sell. The price that I heard for the new camera was about $40K. It has three, one-megapixel CCDs and records on 9 micron tracks, like the 1700 deck. It can be configured to pre-record 10 seconds of video. It is apparently not yet on the Panasonic web site or in the press releases. No literature was available at the unveiling. Charles R. Caillouet Jnr Vision Unlimited/LA Hello!, Just allow me to throw some more wood on the campfire here. As soon as the image is moving (which tends to happen in TV) any interlaced image is really only half the vertical resolution as it is advertised at. This is a fact, not an opinion. 480i is really 640*240 as soon as the image is moving 1080i is really 1920*540 (540!) as soon as the image is moving. No, you can not easily rebuild the lines between the lines missing by a clever de-interlacer. That is called interpolation. It is not real information. No, the information of the other field is not easily usable, because we do not know where or what the camera was filming when that field was recorded. Using the other field boils down to guessing. And if guessing is allowed why not also uprez the progressive images then, using sharpening techniques and make them look very snappy...... (I own a high profile professional realtime de-interlacing processor, I know what it can do, it is impressive, but not really good. Lots of visible defects from interlacing are still visible in the result. Consumer grade de-interlacers will do possibly worse.) I like movement in images which is why I think 1280*720 might be better than 1920*540 even though the latter has 12.5% more pixels. I think 1920*540 just has such a torn-up with/hight equilibrium, hasn't it? 1080i could be created by mounting a trembling (60Hz) motor device on a 1920*540/60p camera in a way the motor would generate a one line vertical trembling movement up/down. This would create a similar result as does interlacing. What I mean to tell here is that any interlacing format is actually half resolution progressive video combined with bad registration. If you wish, you could create 1440i (!) (1280*1440/60i) by mounting this same motor on any 720p camera. Interlacing is an artifact we have been watching daily for 50 years now. Be aware of the nostalgia, emotionally linking this artefact to the pleasure you have been getting from watching TV screens for many years. This is subjective, but I'm sure it counts towards the popularity of interlaced video. Especially on Home TV's. Many common folk will probably feel comfortable watching these trembling images. 1080i looks more like the good old TV we have been looking at for over 50 years! Just a bit sharper. Cool! Just like this love for 2:3 pulldown, like the love for 24p strobing. These are all defects!!! Lets get over with any of them whenever technology allows us to get rid of one! I think we will we get used to the lack of each of these very soon! O.K., One could advocate 1080/30p (PsF) allowing the de-interlacer to do a perfect job. True, but then I still do prefer the higher framerate (60p) over the added resolution. On a home TV, hardly ever to be expected to be much larger than 1m20 (4ft) wide in the general home (the homes will not get bigger with the added resolution) and generally looked at a distance of 2.5 meters (8ft) or more, the benefits of 50/60 fps IMHO clearly outweighs the benefits of the added resolution at the lower framerate of 25 or 30. Also, sports and games move so fast they simply can not be filmed easily at 25p or 30p at all. Too much strobing. 1080 will force sport transmissions to resort back to interlacing, tearing down resolution to a trembling 540 lines. 1080i is not 1080, it is a trembling 540. Best!, Kommer Kleijn http://www.kommer.com Kommer Kleijn writes: >the information of the other field is not easily usable, because we do not >know where or what the camera was filming when that field was >recorded As I understand it, smart de-interlacing uses the information in the other field as a starting point, rather than interpolating a new line completely from scratch. Can you or anyone confirm this? Dan Drasin You always loose resolving power with motion no matter how you image it. The issue is how the brain averages it. Sam Wells Kommer Kleijn wrote: >1080i looks more like the good old TV we have been looking at for over >50 years! Just a bit sharper. Cool! With all due respect Kommer, living where you do I seriously doubt that you have actually seen a lot of 1080/60i material, and you certainly haven't seen it broadcast - not unless you've been spending a lot of time in the States. I see it in my home all the time, and I have to disagree with your statement. It does not look like NTSC but "a bit sharper." Anyone who's seen some of the material broadcast on outlets such as HDNet, or broadcasts of HD sporting events - particularly football and hockey - would undoubtedly disagree with your statement as well. It is an entirely different class of image, regardless of the technical facts behind it. If you intended to say that it has a "video look," this is true by definition because it's 60 images per second, and, when shot on interlaced equipment, is usually shot with video gamma. But to denigrate it as something that's worthless is to be blind. I'm no fan of interlace either, but I do know when I'm seeing something with visual impact. >Just like this love for 2:3 pulldown, like the love for 24p strobing. These >are all defects!!! I don't know anyone here or anywhere else who professes a "love" for 3:2 pulldown. It is simply a technical step that is required in order to bring 24fps imagery to a 60 field per second broadcast system. And we are going to have field based broadcasting systems for a long time, whether you or I like it or not, and no matter how much ranting one wants to do in Internet forums. >1080i is not 1080, it is a trembling 540 If you want to be technical about it. Personally, my attitude is : Who Cares? Nobody is telling you to shoot in that format. However, for broadcast on most of the HD suppliers in the US, it's what we have. Deal with it. Mike Most >Just allow me to throw some more wood on the campfire here Interlacing is a throwback and an homage to the 1930's technology of NTSC, and later, PAL. We've been offered a chance to rid ourselves of interlace and yet we cling to it still. I wish we could make a solid and bold step into the twenty-first century and stay there by going "progressive" even if it is only 720p. Jeffery Haas Dan Drasin wrote : >As I understand it, smart de-interlacing uses the information in the other >field as a starting point, rather than interpolating a new line completely >from scratch. Can you or anyone confirm this? Yes. Our in-house deinterlacer uses a couple of 'smart' techniques : One is to only interpolate in areas of movement, leaving static areas alone. Another is to use directional interpolation, guided by detail in the original image. This helps with diagonal edges. Simon Burley Dan Drasin writes : >As I understand it, smart de-interlacing uses the information in the other >field as a starting point, rather than interpolating a new line completely >from scratch. De-interlacing can be done rather well when many fields are used. John Lowry John Lowry wrote : >De-interlacing can be done rather well when many fields are used. John You SAY this, but be careful. I think your expertise and many years of experience in successfully DOING this is no reason that anyone here should just believe you. Why just last week some guys in Europe told me... Michael Bravin Couldn't interlacing, if de-interlacing algorithms were available which didn't introduce artefacts, be viewed as sort of a 2:1 compression? I'm just curious, not trying to make any point so please don't read anything into this question. Jessica Gallant >I'm just curious, not trying to make any point so please don't read >anything into this question. It's sad that we're at a point where this kind of thing needs to be said. Art Adams, DP [film|hidef|video] Jessica Gallant writes that : >Couldn't interlacing, if de-interlacing algorithms were available which >didn't introduce artefacts, be viewed as sort of a 2:1 compression? Absolutely. One of the reasons there's no such thing as uncompressed video; way before it became data, there were all kinds of things which really amount to compression, color-under (heterodyning), YUV, etc. Sampling chrominance at half luminance (4:2:2) is certainly compression. BTW, one thing often overlooked in the 720"p" debate; a LOT of 720p broadcast is SD up-rez, much of which actually looks really good. Sometimes you have to look hard to see which it is. But a large proportion of it is in fields, whether tape or film original. It is what it is - I'm just saying a lot of it just isn't very P. Tim Sassoon John, >De-interlacing can be done rather well when many fields are used. Care to elaborate? Even just a *little* bit? Cheers, Kim Sargenius >Sampling chrominance at half luminance (4:2:2) is certainly >compression. Michael Bravin writes : >>John Lowry wrote: >De-interlacing can be done rather well when many fields are used. >John >You SAY this, but be careful. Michael, Fair point. I will have to show you some time. Regards. John Lowry P.S : I am in no way in favor of interlace which was implemented as a means of conserving bandwidth a long time ago. >De-interlacing can be done rather well when many fields are used. Kim : An interlaced image looks fine when there is no motion. So the problem is motion and understanding that motion. To understand motion requires the analysis of a number of frames (or fields) in the motion image sequence. Almost all of the image processing we do relates to the analysis and filtering of multiple frames. Is that OK for "just a *little* bit" of info? Regards John Lowry John wrote : >Fair point. John I was taking a sarcastic swipe at some of the CML "editorial" comments My humble apologies Michael Bravin Michael Bravin wrote : >> John I'm sorry I was joking, I consider your experience and expertise >to be unparalleled in this area. I got it Michael! And I thought it was pretty funny! Roderick E. Stevens II >Care to elaborate? Even just a *little* bit? You might find this interesting: http://www.teranex.com/products/docs/TeranexMC Some but not too much marketing-speak, and quite a bit of useful information. Also: http://www.sigmadesigns.com/support/deinterlacing.htm A quick Google of the subject will bring you more info than you can shake a stick at. John Lowry writes : >I am in no way in favor of interlace which was implemented as a means >of conserving bandwidth a long time ago. Back in the tube camera era, fast pans could result in slanted images, like a slit focal plane shutter will with film. Interlacing reduces this by half in the same bandwidth. Irrelevant today. Tim Sassoon John, > Is that OK for "just a *little* bit" of info? Yes, much appreciated! Cheers, Kim Sargenius http://www.teranex.com/products/docs/TeranexMC http://www.sigmadesigns.com/support/deinterlacing.htm Thanks Tim! Kim Sargenius Hi Michael, "John You SAY this, but be careful. I think your expertise and many years of experience in successfully DOING this is no reason that anyone here should just believe you. Why just last week some guys in Europe told me..." The trouble is you've been around too many Brits who "get" irony! It made me smile! However, I think that there might be a difference in perception between what was originally started as a discussion on de-interlacing of broadcast transmission within the receiving monitor, "on the fly" if you would, and professional de-interlacing equipment used to upgrade PAL or NTSC footage to 720p standard or higher prior to broadcast. Regds Russell Branch
|
