Mtf Of Blu-ray/hd-dvds

[betty boop]

must defintly enjoying your contributions to this thread both bris & hosko, with your techical inputs on the topic.

mo you are right re blade runner. the hidef version still to come, i'm anticipating eagerly.

re fifth element, the original release was using the wrong codec, though now re-released using same codec as hd dvd and a much better effort . really though there will always be stuff ups like that. this one embarrasing for blu ray. but shouldn't be used to hang the format or its capabilities.

[MLXXX]

perhaps deficiencies in the HTPC hardware/software?

But I don't really see your point anyway. Are you hinging your entire case on your experience with King Kong that could be the result of any failing of your hardware or your eyesight?

I provided a link to an AVS thread showing many images. I indicated I had confirmed the accuracy of the image on the left of the first post. I suggest you look at the AVS thread. It has been there for some considerable time -- plenty of opportunity for videophiles to criticize the accuracy of the capture technique.

As for my own eyesight, it is better than 20/20 [6/6 in metric] and I routinely oberve good visible resolution in certain free to air 1920x1080 material such as House.

And some scenes of my HD-DVDs show much better resolution than that particular scene from King Kong.

I will address other points raised by yourself and others, after I've had a chance to go through the material; unless others have already addressed the issue.

And if anyone with a Blu-ray or Hd-DVD player wants to test their disk against the posts in the AVS thread they are welcome to do so and report whether their player yields a better picture.

By the way, one way of testing visible resolution is -- using good quality image processing software -- to downsample to half the size and then upsample back to the original size. If the processed image is almost the same as the original picture it suggests the original picture has low visible resolution (or conceivably that the original scene was of simple non-detailed objects). I use paint shop pro to digitally subtract similar photos, and then raise the gamma in the diference image to highlight where the dfferences are. A soft [e.g. out of focus] photo can yield virtually no differences when downrezzed to 960x540 and then uprezzed back to 1920x1080.

[Owen]

I have been away and have not read the posts in this tread, so this is not in response to anyone.

I present this as a hopefully easy to understand explanation of the limitations in HD.

What is resolution?

Some may say it’s the number of pixels in an image; however that is not the case.

Resolution by definition means being able to resolve.

The visible resolution in an image is not directly related to the number of pixels. For example we can upscale a 720x576 image to 1920x1080. The resulting image has 1920x1080 pixels (pixel resolution) but can never have more then 720x576 visible resolution; as there was never more then 720x576 information to start with.

So, now that we have established that visible resolution can be much lower then the pixel resolution would indicate we can move on.

All film and video cameras are affected by a thing called Modulated Transfer Function or MTF for short. MTF describes the “Sharpness” or detail carrying ability of the device or capture system.

All parts of the chain from lens, film and film scanner or digital camera sensor (CCD), digital anti aliasing filters, video compression, video playback system and display all have an MTF. What we see on our TV screen is the sum of all the MTF losses in the chain.

Some people think if we point a digital camera at a scene, a single detail seen by the camera that is as small as a single pixel in the cameras CCD or sensor will be faithfully captured and sub sequentially displayed with pixel perfect precision on a display with the same number of pixels as the camera sensor. Unfortunately nothing could be further from the truth.

For a start all digital capture devices (camera and scanner sensors) are affected by Nyquist sampling limitations, which means the images produced have about half the visible resolution of the sensor.

For example, take a 1920x1080 video camera. The output of this camera will have no more then about 960x540 visible resolution (the Nyquist limited resolution or limit) due to the need to filter the output of the sensor above the Nyquist limit to avoid artifacts and aliasing in the output image.

To get an image of 1920x1080 visible resolution from a camera it must have a sensor with at least 3840x 2160 pixels. Even then, the image will have lost significant sharpness at 1920x1080 due to MTF losses in the lens and anti aliasing filter.

If you want to better understand Nyquist sampling theory I suggest you research on the site linked below.

Moving on, go to this site:

http://www.normankoren.com/Tutorials/MTF.html

Scroll down to the section titled “Image sharpness and detail”.

Look at the side of the page and you should see a picture with black and white vertical bars, this is a resolution test pattern used to test cameras.

At the top of the image you can see that the black and white bars are perfectly defined with dead sharp edges, all the way to the resolution limit at right. This is the original test image before it was photographed and is what some people expect the output of a high res digital camera to look like.

Now look at the sections further down and you can see the affect of MTF losses caused by the camera lens, film, film + lens and subsequent digitization and digital sharpening.

As you can see the final result has little resemblance to the original test pattern, and we have not introduced video compression losses yet.

Now I pose the question, what would you consider “fully resolved” resolution?

Obviously the “Original bands” at the top are “fully resolved” as they are perfect and have an MFT of 100% or 100% pixel to pixel contrast. If we go down to the “Film only” section we can see that although the image is heavily blurred we can just make out the smallest vertical bars at right. Since the bars are just visible with an MTF of about 10% this would still be classed as resolved, even though it is massively blurred compared to the original.

Film shot on a Panavison camera is generally acknowledged to have about 4k resolution, or 4 Mega pixels total, however at the upper limit of that 4k resolution all we get is barely perceptible detail with about 10 MTF, just like in the “Film only” example. Is this “fully resolved” not in my book its not, and I have no doubt others will agree.

Note. I only used the “Film only” example because it has resolution extending all the way to the right. The “Film + lens” result is too limited in res to illustrate my point.

If we look at the “Lens only” section, we can get some idea what an image with an MTF of about 50% looks like at the pixel resolution limit. This is the sort of results we might expect from digital 1920x1080 images derived from a Genesis digital cinema camera.

The linked .pdf white paper linked below provides interesting info on the REAL resolution of a Panavison cinema film camera. MTF graphs at are at the end of the document are in Lines per picture height, an expression used for both horizontal and vertical resolution. This is equivalent to pixel resolution that we are all accustomed to.

At a horizontal resolution of 1920, cinema film has an MTF of only 10% which is effectively nill, and represent virtually no visible detail.

Vertical resolution is better, with an MTF of about 57% at 1080, which is useable but not remotely close to fully resolved.

And download these. Copy and past URL into a new browser.

www.etconsult.com/papers/Technical%20Issues%20in%20Cinema%20Resolution.pdf

www.cst.fr/IMG/pdf/35mm_resolution_english.pdf

www.digitalpraxis.net/zippdf/scene-to-screen.pdf

These white papers provides more interesting info on the limitations of film as display in a typical cinema. Visible resolution is very ordinary and no where near 1920x1080.

More interesting info here confirming the limitations of film.

Be aware that graphs in cycles/mm need to be converted to lines per picture height to be comparable with resolution as we know it in the domestic environment.

For more interesting reading go here:

http://www.digitalpraxis.net/

Look at the left hand side of the page and select “Technical Papers”

Download DI Guide and Digital Film Scene to Screen .pdg files.

I cant find any useful MTF data on the Genesis digital cinema camera, but digital cameras of similar pixel count tend to behave in a similar manner.

There is the test data on the Canon EOS-1Ds Mark II 16 Mega pixel SLR, with 4992x3328 image sensor, very similar in sensor resolution to the Genesis digital cinema camera.

http://www.imaging-resource.com/PRODS/EOS1...1DS2IMATEST.HTM

MTF data at bottom of page.

Resolution at 50% MTF 1908x1630 without digital sharpening.

Resolution at 50% MTF 3084x2909 with digital image sharpening.

MTF at Nyquist limit (2496x1664) = 11 to 17% without image sharpening.

MTF at Nyquist limit (2496x1664) = 38% with image sharpening.

MTF at Nyquist limit (2496x1664) = 11 to 17% without image sharpening.

This camera, without artificial sharpening to boost its performance provides “Full HD” resolution (1920x1080) with only 50% MTF. Fine details with have only half the contrast they should have and cannot be considered “fully resolved”.

For comparison, here is an interesting and easy to comprehend article on the Sony 1080p 24 HD cameras

bssc.sel.sony.com/Professional/production/productsite/files/24PTechnicalSeminar2.pdf

The Sony 1920x1080p 24 HD camera with lens has an MTF at it’s Nyquist limit (960) of about 25%, about mid way between the unsharpened and sharpened results for the Canon EOS SLR at it’s Nyquist limit. This seems typical of digital cameras.

All the info I have provided confirms that 1920x1080 FULLY RESOLVED, is not a possibility at this point in time, even from the best available source.

The only way to get 1902x1080 resolution fully resolved, (with 100% MTF) is with computer generated graphics. Nothing shot with a digital or film camera can have 100% MTF at full HD resolution.

Resolution specifications without accompanying MTF data are worthless, so be very careful when resolution is quote. It’s the MTF rating that will tell you how sharp the image will look.

As pointed out in my links, digital video cameras tend to be sharper then film because that have better MTF in the very important mid resolution band, where the human eye is more sensitive. A typical 1920x1080 video camera looks very sharp yet has very limited resolution.

I hope this explains my stance on the limitations of HD. Sometime we get caught up in semantics over terminology, which clouds the real issues.

Regards to all.

Owen

Edited to correct inoperative links.

[momaw]

however with most master these days coming from a minimum 2k the resolution in the master far exceeds that of "true HD" and hence is actually scaled down. The flaw with your argument is it seems based on the notion of upscaling and the belief that cameras capture less than "true HD" resolution. The reality is far more information is discarded than is kept when the final blu-ray or HD DVD is made.

[Owen]

Sorry, who where you responding to.

If it was me it makes no sense.

[momaw]

Sorry, who where you responding to.

If it was me it makes no sense.

of course it doesn't

[Owen]

Don’t be like that, explain your case.

I said nothing about up scaling or 2k resolution.

I showed that the original film negative has very poor MTF at 1920x1080, about 10% horizontal and 57% vertical, not even close to fully resolved 1920x1080. 10% modulation isn’t worth a cracker.

Are you saying all the Panavision MTF data is inaccurate?

[betty boop]

however with most master these days coming from a minimum 2k the resolution in the master far exceeds that of "true HD" and hence is actually scaled down. The flaw with your argument is it seems based on the notion of upscaling and the belief that cameras capture less than "true HD" resolution. The reality is far more information is discarded than is kept when the final blu-ray or HD DVD is made.

exactly momaw, all the cameras I have provided info on have greater than 2k res. owen it really would be worth your while reading the links to info provided and also the very insightfull info provided by hosko who very clearly knows what he is talking about and a very clear understanding of the technological capabilties of film making & production.

it is without a shadow of doubt that the technology used to capture & transfer onto hd dvd & blu ray is more than capable of 1920x1080. all my links confirm this, hosko with his understanding has confirmed this too. not sure what more you would want.

to refute it is not possible to capture either by film or digital at more than 1920x1080, let alone to have cg material @ 1920x1080 on hd dvd & blu-ray is just a crazy argument.

[momaw]

Don’t be like that, explain your case.

I said nothing about up scaling or 2k resolution.

I showed that the original film negative has very poor MTF at 1920x1080, about 10% horizontal and 57% vertical, not even close to fully resolved 1920x1080.

Are you saying all the Panavision MTF data is inaccurate?

Hence why I said:

The flaw with your argument is it seems based on the notion of upscaling and the belief that cameras capture less than "true HD" resolution.

The problem is film negative isn't 1920x1080 nor are these digital cameras recording such a low res.

[brisvegas1]

to refute it is not possible to capture either by film or digital at more than 1920x1080, let alone to have cg material @ 1920x1080 on hd dvd & blu-ray is just a crazy argument.

Cheeky devil now you are just trying to stir.

Owen has specifically made neither of those points - suggesting he has is just mischevious. His argument is more nuanced than that..

[Owen]

I never said that 1920x1080 or more was not attainable, what I have said is that the MTF or modulation at that sort of resolution is low and does not represent fully resolved,

all the MTF graphs I can find confirm this.

How can a pixel data that has less then half the pixel to pixel contrast of the original image be considered fully resolved?

You can have 4000x2000 pixels, but if they only have low MTF, there is little to be gained.

[betty boop]

Don’t be like that, explain your case.

I said nothing about up scaling or 2k resolution.

I showed that the original film negative has very poor MTF at 1920x1080, about 10% horizontal and 57% vertical, not even close to fully resolved 1920x1080. 10% modulation isn’t worth a cracker.

Are you saying all the Panavision MTF data is inaccurate?

we have seen nothing from you inregards the genesis camera or anyother camera which is easily able to do 2k and above.

we have nothing from you to refute any of the links i've provided. did you read the one about the resolution of film.

by the way normenkoren is just a guy into still consumer cameras. nothing to do with film cameras were talking here. we are also not talking about tv cameras.

[betty boop]

Hence why I said:

The problem is film negative isn't 1920x1080 nor are these digital cameras recording such a low res.

I just don't think he seems to be able to understand that, and has obviously read any of the info posted or linked to in this thread, just a pointless argument really now.

[Owen]

I said I could not find any MTF data for a Genesis, but I did find data for a Canon SLR with similar pixel count.

I also said that most digital cameras seem quite similar in their MTF characteristics at the Nyquist limit, be they still or video.

The 16 Mega pixel (16k) Canon is good but it still cannot “fully resolve” 1920x1080, the MTF is no where near 100% at that resolution, no camera can do that.

Unless you can show at least one example of a cinema or video camera that can provide 100% MTF at 1920x1080 or more, I don’t see how you have a case.

More pixels dont count if they dont have usefull data.

I am not saying the Genesis is not good, it should have higher MTF and hence sharpness then anything else at the moment, but its far from perfect.

[momaw]

I said I could not find any MTF data for a Genesis, but I did find data for a Canon SLR with similar pixel count.

I also said that most digital cameras seem quite similar in their MTF characteristics at the Nyquist limit, be they still or video.

The 16 Mega pixel (16k) Canon is good but it still cannot “fully resolve” 1920x1080, the MTF is no where near 100% at that resolution, no camera can do that.

Unless you can show at least one example of a cinema or video camera that can provide 100% MTF at 1920x1080 or more, I don’t see how you have a case.

More pixels dont count if they dont have usfull data.

I am not saying the Genesis is not good, it should have higher MTF and hence sharpness then anything else at the moment, but its far from perfect.

I think you are missing the point entirely that the recordings are not made at 1920x1080 but at a much higher resolution. Thus with a good compressionist the resultant image can be downscaled to a fully resolved 1920x1080 image. And this is precisely what they do.

[brisvegas1]

I just don't think he seems to be able to understand that, and has obviously read any of the info posted or linked to in this thread, just a pointless argument really now.

That is really not what he is saying. It is much more like - for a high spatial resolution image of a real world scene (i.e. not CG) eg. 1920x1080 you need to capture at much higher resolution.

i.e. it isn't just nice to have 4k - for scenes with lots of fine detail eg. fur, hair etc - you need 4k capture in order to have pixel in a 1920x1080 image that are independent of each other.

So far you guys are in agreement - where you differ is in estimation of camera capabilites.

Owen acknowledges that the camera captures 4k frames, he just questions if each pixel is independent of each other at that or lower resolutions.

The only way to test this is objectively with a test pattern - Owen claims to have seen reports backing up his view. Others say no.

Simple 2k resolution by the laws of math/physics/information theory would not be enough - at 4k it starts to become possible, but would be even better at 6k.

Somewhere between 2k and 6k owen stops being right and starts being wrong - if the optics/ccd (capture) is up to it.

I can make my little cybershot capture a 4k image - but it would be crap and not have any real detail. The genesis camera may have great optics and a fantastic sensor - but in the absence of evidence proving otherwise, owen is entitle to be a little sceptical. If for no other reason than information theory stating that what you propose is towards the ideal end of what is possible.

That is not to say that the image from a genesis camera is crap - far from it. Objectively and subjectively they are great.... miles better than what we are used to. And HD is a great medium in which to experience it... but you are talking about a stickler with owen. someone who took apart a perfectly reasonable screen to give it a little extra oomph... of course he is going to care about a slight and possibly perceptible softening of a HD image over what will soon be possible.

Most people wouldn't pick it / wouldn't care if they did - but it is precisely folks that do care that drive innovation.

[Owen]

The point I have repeatedly tried to make is that no matter how many pixels the camera captures it cannot fully resolve (100% MTF) at 1920 lines per picture high horizontally or 1080 lines per picture height vertically, neither can a 16 Mega pixel digital SLR that captures in 4992x3328.

A digital camera cannot capture anything at the single pixel level. MTF is zero at the native resolution of the sensor.

MFT is typically 25% at HALF the native pixel resolution of the sensor, and falls off rapidly there after.

25% MTF is a long why from “fully resolved” which requires 100% MTF.

The best camera I have been able to find data on manages about 50% MTF at 1920x1080 lines per picture height. Thats good but you cant call it fully resolved.

[MLXXX]

This is all I have time for in this thread tonight:-

...

Here's some interesting comments from Josh Zyber (I'm sure you all know who he is)

Josh Zyber: A long-time movie buff and collector of discs from laserdisc to DVD. Trained at film school.

The comments are not expressed in a technical way but include the observation that DVDs derived from electronic sources look sharper than film derived ones. I believe this is because the MTF is maintained relatively high as the Nyquist limit is approached in a TV camera.

I note this:

Obviously a digital video source will retain its full resolution, but

although it would appear to be dependant on the very analogue lens system,

the MTF of the lens is more often than not, not the limiting factor (unless

you have a really cheap and nasty lens).

The MTF curves of a studio high definition camera lens separately, and when combined with the camera, can be seen at page 6 of this Canon pdf if anyone is interested.

The video camera sensor uses optical filters that specifically limit the resolution as Nyquist is approached. This allows the TV camera to offer a sharper picture. It will be noted that at the Nyquist limit, the MTF of the lens is about 78%, and of the camera overall around 20%. Twenty percent will cause a little bit of aliasing but viewers may prefer to put up with that if the picture looks a bit sharper.

The tradition of moviemaking is different. There was no concept of 1920x1080 sampling, so there was no optical filtering to limit the camera performance. The negative contains higher spatial frequencies that could generate moiré patterns.

Perhaps this is why filmmakers appear to have decided to be conservative in the visible resolution that is allowed to make its way onto a Bluray or HD-DVD disk from older movies.

Another possible reason is that a softer video source is easier to encode without macroblocking. We may be prepared to put up with occasional macrobocking when waching an AFL match on TV, but we expect our Blu-ray and HDTV movies to look smooth at all times.

ps as an aside what I find by the way absolutely hilarious, is we have here the king of visual aquity, who pushes down everyones throat that we should sit with our noses to the screens to be able to resolve 1920x1080, infact now saying it is not possible at all to ever even get 1920x1080 onscreen ! . which is plain wrong as per all the links above, but anyways would explain why your misguided friend MLXXX, was infact doubt sitting at the wrong distance to the 57" telly he had so would have no hope of ever fully resolving 1920x1080, all the while preaching to every one the distance they should view their screens at

Just for the record, I've never had a 57" telly.

Viewing distance charts are designed for real world video (not using the display as a computer monitor).

So they will be designed for 960x540 visible resolution (or a little less) for the 1920x1080 format.

In practice, the visible resolution is usually a bit less than 960x540.

And now to respond to a calmer section of the post. I've interspersed comments in blue font:

quote name='alebonau' date='Oct 1 2007, 08:50 AM' post='794084'

MLXX if you read my following links you will infact see that

- film more than capable of 6K using the prime lenses they do and 4K very easily utilising the lenses they use these days that are very good.

Lenses may need to introduce additional widening for very wide screens, which places strain on the horizontal resolution. With film, 35mm is a little small. In the sixties, with widescreen productions, 70mm was sometimes used. IMAX uses large film stock and large lenses. I think the fixed lenses in cameras used today for making movies are good enough to do the job. However, it is very time consuming to edit at very high resolutions and if it is known the movie will end up being multi-generation copied to a distribution print, or merely converted to 1920x1080 Blu-ray or HD-DVD there may be no point in preserving the full resolution as the movie is processed. And it may be important that the whole of a movie have the same '"look" so everything may be cut back to a lower common denominator of visible resolution.

- Infact jsut check what the resolution capabilities of even a 35mm frame of a digital camera

Yes a negative by itself does have excellent resolution capacity. However I think the discussion topic is more about MTF for the 1980x1080 sampled format, than whether a negative has detectable resolution at very high spatial frequencies.

- Additionally film scans are done at 4K lowry process for hi-def which is then down converted to 1920x1080

I believe it is best to err on the side of scanning scan at a higher rez if possible, provided there is some detail present to be captured.

- The genesis camera is a 12mp one and with the lenses usesd is capable of a final output of 1920x1080, there are many films used to produce this for the hi-def eg apocalypto which received world wide aclaim for its PQ

Yes, the specificatons for Genesis cameras are very good, though they cannot handle the intensity range of film negative, in a single frame. I saw

Superman Returns

, which was filmed using Genesis cameras. It looked like slightly grainy film to me (a grain effect had been artificially introduced), and slightly soft compared with some conventionally created movies I'd seen in the cinema years in the past. There were no signs of digital artefacts. The softness may have been due in paret to limitations in the cinema equipment. I have not seen a high definition disk version of

Superman Returns

.

-then there is CG material which is 1920x1080 there are no film or cameras involved here.

Yes, but there has to be care that the CG material blends in with any real world footage present in the movie. If not it could be made quite sharp - so sharp that a 1024x768 TV could make the animation look decidely blurry compared with a 1920x1080 display. However there is still the issue of the encoding codec and the data rate. These may make it desirable to introduce a little bit of smoothing before encoding, for good results. I don't have any experience with encoding crisp animation sequences for MPEG2, VC-1, or h.264.

At the risk of oversimplying, I think that a lot of this discussion is at cross purposes. Owen has referred to the limitations of a 1920x1080 sampling format; initially in relation to information theory (Nyquist), and then by way of comment in relation to certain practical limitations that often lead to further -- often relatively slight -- degradation in visible resolution for the 1920x1080 format, compared with a theoretical ideal.

I don't quite understand why there has been such heated opposition.

- MLXXX

Edited October 1, 2007 by MLXXX

[Owen]

We are on the same page mate, but I don’t consider a 50% loss of pixel to pixel contrast at high resolutions a “relatively slight – degradation”.

PC text and graphics look bad with that sort of MTF.

I don’t have data on how much further reduction in MTF is caused by video compression, but in my experience losses are quite significant.

[MLXXX]

We are on the same page mate, but I don't consider a 50% loss of pixel to pixel contrast at high resolutions a "relatively slight – degradation".

PC text and graphics look bad with that sort of MTF.

I don't have data on how much further reduction in MTF is caused by video compression, but in my experience losses are quite significant.

What I don't know is whether the reductions are mostly tied up with Nyquist. I suspect they are.

If Nyquist is to be respected there needs to be a roll-off leading up to the Nyquist limit. The filtering is not a steep cliff although with an optical filter in front of an optical sensor it can be fairly rapid. If I read from the MTF curves at figure 4 (the 1080 camera) and figure 5 (the 720p camera) on page 6 in the pdf in my post two posts above, I get the following approximate figures:

Percentage of .........720p camera ....1080 camera

Nyquist frequency ........MTF ............... MTF

110% ......................... 20 ................. 20

100% ......................... 25 ................. 22

90% .......................... 35 ................. 30

80% .......................... 42 ...................40

74% (ref) .................... 50 ................. 45

60% .......................... 70 .................. 70

By inspection of the graphs, at Nyquist, the lenses are performing at roughly 90% and 78% respectively.

The performance of the overall camera at Nyquist is way lower than that, at roughly 25% and 22% resepctively.

So in the design of the camera although the lens is contributing to the reduction in MTF, it is not the predominant reason.

I suspect it may be largely an inescapable result from the physics of the setup, and the need to minimize aliasing.

Even with much better lenses, the MTFs would not be that much better. And perhaps better lenses would exacerbate aliasing problems anyway.

Compression

Loss of resolution due to the encoding into MPEG2, VC1 or H.264 would be evident in moving sections on close inspection. Because of the higher data rate than for DVDs or free to air I would expect the static parts of the image to be relatively unaffected.

With HD-DVD and the VC1 codec I have not noticed macroblocking. The relatively static parts of the image do scrub up well. But they are not as crisp as some FTA 1920x1080 material. There seems to be more softening and filtering than wth the crisper 1920x1080 FTA.

Conclusion

Visible resolution of the 1920x1080 format is much less than a lot of people would assume. Not only must the visible resolution be limited to half the sampling grid to avoid aliasing, but even within that limitation, the "contrast" or MTF is a lot less than 100%. So the resolution although "visible" is not strongly visible! Black and white patterns become patterns of dark grey and dark white.

Some people imagine that a video camera works like a flysceen, with a light detector at each square in the flyscreen so that each square is transmitted independently of adjacent squares. This is not how practical video cameras work. If you were to 'pixelate' an original image of very high resolution into 1920x1080 hard pixels it would look dreadful close up. You would see strange variations during a pan or zoom, or if the image moved independently of the camera. The pixels have to be softened to make the digital sampling process work.

The same applies when scanning a sequence of film frames at a high resolution. The result of the scan [the raw signal of a precision scanning head] cannot be used in hard form. The scan is softened using algorithms that average out the light levels by reference to adjacent sample points. By doing this the sampling grid disappears and we see the overall image [and avoid false patterns or 'aliasing'].

A compromise needs to be struck between sampling detail and overall picture smoothness.

This concept also applies with algorithms that resize pictures to a smaller resolution format. Adjacent pixels in the target image will not be full white and full black, but a lower contrast transition.

- MLXXX

Edited October 1, 2007 by MLXXX

[betty boop]

I said I could not find any MTF data for a Genesis, but I did find data for a Canon SLR with similar pixel count.

I also said that most digital cameras seem quite similar in their MTF characteristics at the Nyquist limit, be they still or video.

The 16 Mega pixel (16k) Canon is good but it still cannot “fully resolve” 1920x1080, the MTF is no where near 100% at that resolution, no camera can do that.

Unless you can show at least one example of a cinema or video camera that can provide 100% MTF at 1920x1080 or more, I don’t see how you have a case.

More pixels dont count if they dont have usefull data.

I am not saying the Genesis is not good, it should have higher MTF and hence sharpness then anything else at the moment, but its far from perfect.

owen get with the program, we are not talking digital slrs here and I have not seen anything to suggest the genesis ccd is same as the canon 12mp cmos, either way I am sure they are a completely different implentation.

yet to see anything from you inregards the film cameras, film production systems to dispute they cannot easily capture & output more than 2 and 4k more than sufficent for the lower image res required for hd dvd & blur ray. and I have provided a list of cameras used including the genesis. really yet to see anything from you inrelation to film, i've provided you info on its capabilities would like to see some evidence from you to say otherwise.

and normankoren ? give it a break the guy is a amateur photographer who got into digital slrs and bought a canon 10d slr around the time I did and posts pics on his website and discusses digital slrs, what the hell are we discussing material here on his website when what this thread is about is how holywood blockbusters are filmed, produced and transferred for hd dvd & blu-ray.

[betty boop]

That is really not what he is saying. It is much more like - for a high spatial resolution image of a real world scene (i.e. not CG) eg. 1920x1080 you need to capture at much higher resolution.

i.e. it isn't just nice to have 4k - for scenes with lots of fine detail eg. fur, hair etc - you need 4k capture in order to have pixel in a 1920x1080 image that are independent of each other.

So far you guys are in agreement - where you differ is in estimation of camera capabilites.

Owen acknowledges that the camera captures 4k frames, he just questions if each pixel is independent of each other at that or lower resolutions.

The only way to test this is objectively with a test pattern - Owen claims to have seen reports backing up his view. Others say no.

Simple 2k resolution by the laws of math/physics/information theory would not be enough - at 4k it starts to become possible, but would be even better at 6k.

Somewhere between 2k and 6k owen stops being right and starts being wrong - if the optics/ccd (capture) is up to it.

I can make my little cybershot capture a 4k image - but it would be crap and not have any real detail. The genesis camera may have great optics and a fantastic sensor - but in the absence of evidence proving otherwise, owen is entitle to be a little sceptical. If for no other reason than information theory stating that what you propose is towards the ideal end of what is possible.

That is not to say that the image from a genesis camera is crap - far from it. Objectively and subjectively they are great.... miles better than what we are used to. And HD is a great medium in which to experience it... but you are talking about a stickler with owen. someone who took apart a perfectly reasonable screen to give it a little extra oomph... of course he is going to care about a slight and possibly perceptible softening of a HD image over what will soon be possible.

Most people wouldn't pick it / wouldn't care if they did - but it is precisely folks that do care that drive innovation.

bris, I have put forward the capabilties of movie cameras with the genesis as one example, showing its 12mp capabilty plus that it can output 4k. and the capabilities of film resolution which as I qouted is 20mp which with the lenses today can easily achieve 6k, typically 4k, and hosko backed that up. and have posted inregards the lowry transfer process to and how that is easily capable of transferring at 4k which is 4 times ! the level of info needed for blu-ray & hd dvd.

I have not seen any info from owen to dispute the capabilties of the film capture, prodution & transfer process to say any otherwise.

[betty boop]

This is all I have time for in this thread tonight:-

Josh Zyber: A long-time movie buff and collector of discs from laserdisc to DVD. Trained at film school.

The comments are not expressed in a technical way but include the observation that DVDs derived from electronic sources look sharper than film derived ones. I believe this is because the MTF is maintained relatively high as the Nyquist limit is approached in a TV camera.

I note this:

Obviously a digital video source will retain its full resolution, but

although it would appear to be dependant on the very analogue lens system,

the MTF of the lens is more often than not, not the limiting factor (unless

you have a really cheap and nasty lens).

The MTF curves of a studio high definition camera lens separately, and when combined with the camera, can be seen at page 6 of this Canon pdf if anyone is interested.

The video camera sensor uses optical filters that specifically limit the resolution as Nyquist is approached. This allows the TV camera to offer a sharper picture. It will be noted that at the Nyquist limit, the MTF of the lens is about 78%, and of the camera overall around 20%. Twenty percent will cause a little bit of aliasing but viewers may prefer to put up with that if the picture looks a bit sharper.

The tradition of moviemaking is different. There was no concept of 1920x1080 sampling, so there was no optical filtering to limit the camera performance. The negative contains higher spatial frequencies that could generate moiré patterns.

Perhaps this is why filmmakers appear to have decided to be conservative in the visible resolution that is allowed to make its way onto a Bluray or HD-DVD disk from older movies.

Another possible reason is that a softer video source is easier to encode without macroblocking. We may be prepared to put up with occasional macrobocking when waching an AFL match on TV, but we expect our Blu-ray and HDTV movies to look smooth at all times.

Just for the record, I've never had a 57" telly.

Viewing distance charts are designed for real world video (not using the display as a computer monitor).

So they will be designed for 960x540 visible resolution (or a little less) for the 1920x1080 format.

In practice, the visible resolution is usually a bit less than 960x540.

And now to respond to a calmer section of the post. I've interspersed comments in blue font:

quote name='alebonau' date='Oct 1 2007, 08:50 AM' post='794084'

MLXX if you read my following links you will infact see that

- film more than capable of 6K using the prime lenses they do and 4K very easily utilising the lenses they use these days that are very good.

Lenses may need to introduce additional widening for very wide screens, which places strain on the horizontal resolution. With film, 35mm is a little small. In the sixties, with widescreen productions, 70mm was sometimes used. IMAX uses large film stock and large lenses. I think the fixed lenses in cameras used today for making movies are good enough to do the job. However, it is very time consuming to edit at very high resolutions and if it is known the movie will end up being multi-generation copied to a distribution print, or merely converted to 1920x1080 Blu-ray or HD-DVD there may be no point in preserving the full resolution as the movie is processed. And it may be important that the whole of a movie have the same '"look" so everything may be cut back to a lower common denominator of visible resolution.

- Infact jsut check what the resolution capabilities of even a 35mm frame of a digital camera

Yes a negative by itself does have excellent resolution capacity. However I think the discussion topic is more about MTF for the 1980x1080 sampled format, than whether a negative has detectable resolution at very high spatial frequencies.

- Additionally film scans are done at 4K lowry process for hi-def which is then down converted to 1920x1080

I believe it is best to err on the side of scanning scan at a higher rez if possible, provided there is some detail present to be captured.

- The genesis camera is a 12mp one and with the lenses usesd is capable of a final output of 1920x1080, there are many films used to produce this for the hi-def eg apocalypto which received world wide aclaim for its PQ

Yes, the specificatons for Genesis cameras are very good, though they cannot handle the intensity range of film negative, in a single frame. I saw

Superman Returns

, which was filmed using Genesis cameras. It looked like slightly grainy film to me (a grain effect had been artificially introduced), and slightly soft compared with some conventionally created movies I'd seen in the cinema years in the past. There were no signs of digital artefacts. The softness may have been due in paret to limitations in the cinema equipment. I have not seen a high definition disk version of

Superman Returns

.

-then there is CG material which is 1920x1080 there are no film or cameras involved here.

Yes, but there has to be care that the CG material blends in with any real world footage present in the movie. If not it could be made quite sharp - so sharp that a 1024x768 TV could make the animation look decidely blurry compared with a 1920x1080 display. However there is still the issue of the encoding codec and the data rate. These may make it desirable to introduce a little bit of smoothing before encoding, for good results. I don't have any experience with encoding crisp animation sequences for MPEG2, VC-1, or h.264.

At the risk of oversimplying, I think that a lot of this discussion is at cross purposes. Owen has referred to the limitations of a 1920x1080 sampling format; initially in relation to information theory (Nyquist), and then by way of comment in relation to certain practical limitations that often lead to further -- often relatively slight -- degradation in visible resolution for the 1920x1080 format, compared with a theoretical ideal.

I don't quite understand why there has been such heated opposition.

- MLXXX

firstly good on you to atleast respond and provide some rebuttal to the points I made and info provided.

the 57'' telly was for your mate owen and he knows what i'm talking about, the irony of which is very hard to pass up !

mlx we are not talking tv cameras here. additionally hosko has quite clearly outlined with his technical background the lens capabilties of film cameras, infact if I remember he said words to the effect ''show me one lens used in the last 10 used that can't do 4k''

inregards your other comments they are purely your subjective opinion, have not seen anything technical referenced with any evidence to say any otherwise.

[betty boop]

We are on the same page mate, but I don’t consider a 50% loss of pixel to pixel contrast at high resolutions a “relatively slight – degradation”.

PC text and graphics look bad with that sort of MTF.

I don’t have data on how much further reduction in MTF is caused by video compression, but in my experience losses are quite significant.

ofcourse you & mlx are on the same page, because all your talking about is digital slrs and lenses and now tv cameras and your subjective opinion based on what you can see on your screens. maybe that's as far as your knowledge & experiece extends which is fair enough.

to be quite frank in this thread really only seen hosko actually post with some authority given his technical knowledge on the matter, and there was some good technical dialogue with bris who also appears to have some technical knowledge on the matter.

and really we have seen nothing technical to dispute any of the info or links I posted. perhaps we can return to discussing the actuall topic of the thread, if you want to discuss tv camera or digital slr capabilty or what your digital display capabilty & limitations are please create a thread for that as its a separate discussion I have no interest in discussing, certainly not here as theres no relavance to the topic at hand.

not sure i'll get the chanxe to post much anyways next day or so, so on the other hand if all want to discuss here are these topics ..have fun !

[momaw]

That's it Al. I'd really like to see some more from Hosko.

We aren't talking home movies here or "happy snaps", we are talking Major Motion Pictures which are filmed to be screened on screens as large as IMAX. The resolutions are way beyond 1920x1080 (imagine that put on imax ). By the time the film is downscaled to that resolution any deficiencies have been long since filtered out and you are getting a pure 1920x1080 print.