Luminous Landscape Forum
Equipment & Techniques => Motion & Video => Topic started by: spotmeter on June 30, 2011, 11:29:10 pm
-
Anyone know why the HD video resolution of the 5D2 is so pathetic?
When I play the footage on my Sharp 43" LCD screen, it's not any better than my Canon XF305, which has only 1/3rd" sensors, when I am shooting landscapes with lots of small detail.
I'm using a Zeiss ZF 50mm macro lens on the D2.
-
I think that the video resolution of the 5Dmk2 is limited by the ability of the high-resolution (still-image optimized) sensor to read out pixels at 24-60 frames per second. The way they do this is by reading every n-th line of sensels, then trying to estimate 1080p full-color pixels from that (similar to Demosaic).
-h
-
When I play the footage on my Sharp 43" LCD screen, it's not any better than my Canon XF305, which has only 1/3rd" sensors, when I am shooting landscapes with lots of small detail.
How are you connecting the two cameras to your LCD?
-
How are you connecting the two cameras to your LCD?
Via the HDMI ports
-
I'm a real newbie at DSLR video so.....FWIW, I did not notice a problem like this with the "House" episode that was shot with a 5DII. I also have a friend who shoots art videos with his and the finalized files looked great on a n HD tv setup. So there must be something in your setup that is off. I have shot a fair amount with mine but not played in on anything larger than my computer screen. I hope to learn something from this discussion.
-
You might find this post interesting:
http://www.dvxuser.com/V6/showthread.php?187165-GH1-meet-the-DSC-MegaTrumpets-res-chart
Net conclusions? The GH1 and the 7D (and the 5D) don't display true high-def images. The GH1 and 7D look like they probably have about the same amount of real detail in 1080p mode, which would be the amount of detail you could expect from a 720p camera. They do, however, render a much sharper-looking image because of all the false detail they let through. Even though their overall true detail might be comparable, the GH1 is clearly much better behaved on the charts, showing less aliasing and less bothersome color contamination.
-
The OP isn't doing anything wrong - that's just how the 5D2 works and measures. The video function doesn't read all the sensor and skips lines, and some rather serious shortcuts are taken in the production of the moving images. I've measured the 5D2 and it's resolution is vastly lower than 1920x1080, and quite corrupt with aliasing from the line skipping.
-
Does someone knows about the GH2 in that aspect?
-
I think the GH2 uses a different method than the 5D2, but I'm not sure of it's precise nature. We can tell it's a different method due to the lack of chroma moire which the 5D2 is rather prone to. The thing to do if you own one of these cameras is to lock the camera off, and shoot a still of a scene, then a quick video. Take the still and crop / scale in Photoshop to match the aspect ratio and framing of the video and compare. What you'll see is what the shortcuts are doing to the image and be able to evaluate what the video looks like, compared to what it would be like if the full sensor had been used then downsampled properly.
Graeme
-
I think the GH2 uses a different method than the 5D2, but I'm not sure of it's precise nature. We can tell it's a different method due to the lack of chroma moire which the 5D2 is rather prone to. The thing to do if you own one of these cameras is to lock the camera off, and shoot a still of a scene, then a quick video. Take the still and crop / scale in Photoshop to match the aspect ratio and framing of the video and compare. What you'll see is what the shortcuts are doing to the image and be able to evaluate what the video looks like, compared to what it would be like if the full sensor had been used then downsampled properly.
Graeme
Thanks Graeme.
-
I think the GH2 uses a different method than the 5D2, but I'm not sure of it's precise nature. We can tell it's a different method due to the lack of chroma moire which the 5D2 is rather prone to. The thing to do if you own one of these cameras is to lock the camera off, and shoot a still of a scene, then a quick video. Take the still and crop / scale in Photoshop to match the aspect ratio and framing of the video and compare. What you'll see is what the shortcuts are doing to the image and be able to evaluate what the video looks like, compared to what it would be like if the full sensor had been used then downsampled properly.
Graeme
Graeme,
I did this early on with the 5d2.
Both are full resolution, native crop.
http://ishotit.com/rundsmc.jpg
IMO
BC
-
I think the GH2 uses a different method than the 5D2, but I'm not sure of it's precise nature. We can tell it's a different method due to the lack of chroma moire which the 5D2 is rather prone to. The thing to do if you own one of these cameras is to lock the camera off, and shoot a still of a scene, then a quick video. Take the still and crop / scale in Photoshop to match the aspect ratio and framing of the video and compare. What you'll see is what the shortcuts are doing to the image and be able to evaluate what the video looks like, compared to what it would be like if the full sensor had been used then downsampled properly.
Graeme
Thanks, Graeme, for your thoughtful and informative post. I shot a still of the scene after viewing my original footage just to make sure there was nothing wrong with the camera or my connection to the screen. Sure enough, the still was stunning on a 43" screen, but the video was pathetic. Hopefully on the D3 they will use the full 16:9 center of the sensor and downsample properly as you suggest.
Sampling every second or third line of a sensor is like viewing the world through thin venetian blinds turned partially closed. You just have to guess what is behind each blind.
-
Sampling every second or third line of a sensor is like viewing the world through thin venetian blinds turned partially closed. You just have to guess what is behind each blind.
If you could do proper smoothing before, it might not be that bad. But the 5D AA filter is of course not optimized for that mode.
It seems that the reason they are doing this is that either the sensor is overheated, or it does not support the bandwidth needed for moving that large amounts of pixels per second.
-h
-
Although...
Yesterday I bought a Zeiss Sonnar, it's an optic that I like. In the 50's-60's, the russian industry did several copies of the Sonnars. I wasn't born.
They are extensive testings available on resolution and as expected the Zeiss have a better overall resolution, center and edges,
But despite everything on the paper, every testing, I still prefer how a Jupiter delivers the images than the Zeiss.
In fact, I will end to buy the russian copy because of the look and feel.
What I mean is that a lot of good stuff have been done with the 5D2 so far.
-
Did you know the HDMI in record mode on the 5D2 is not 1080 but only SD? It is full HD only in live view mode. Were you using live view or record mode? For all intents and purposes, the HDMI on the 5D2 is useless for HD video work, even when using a monitor.
Can you give us screen grabs of the 5D2 vs the XF305? In HDMI and via SDHC on a computer screen (separately?) Just wanted to know what you meant by 'crappy'.
Are you using a consumer LCD panel or a professional ultra-HD monitor? Have you seen the zacuto shootout series? There is no way you would see that big a difference in resolution on any LCD panel. How far away are you viewing the video from the screen?
-
Graeme,
I did this early on with the 5d2.
Both are full resolution, native crop.
http://ishotit.com/rundsmc.jpg
IMO
BC
That's amazing. Wow... I just got my 5Dii a few days ago. I can't wait to try it out!
-
Great test! If I take your full rez and downsample it so that it matches the video shot it exhibits the very same effect that I first noticed when investigating this. The video side doesn't look too bad until you compare it with what it should be like and you can see what is missing.
Graeme
-
Great test! If I take your full rez and downsample it so that it matches the video shot it exhibits the very same effect that I first noticed when investigating this. The video side doesn't look too bad until you compare it with what it should be like and you can see what is missing.
Graeme
None of us users, really knows what goes on inside the camera. I'm sure the Canon softens and lines skips and somehow makes a a 20 megapixel sensor a lot less for video, because sometimes we see moire or artifacts.
But heck it's a $2,500 camera that shoots good motion and great stills, so it's kind of hard to complain about it.
Then again I don't know why our RED One with the MX sensor is 4k and with the Epic 5k. I guess it's processing power, though something is getting tossed, but in reality all I care about is the final look.
We've used stills from our RED One in print and usually they're fine as long as the subject is waist up. Full length on a horizontal format becomes more challenged, but I'm not 1000% convinced that one camera can do everything anyway, for still or motion projects.
We primarily shoot the RED, but until we get an Epic, it's a beast and the 5d is a great camera for quick set ups, tight quarters like a quick window mount and low low light.
We're also adding that Sony FX 100 or 1000 or whatever it's called because it does some things the 5d/7d/ and the RED don't do. though once again, the A camera is usually the RED.
Every camera has limitations. The RED has good resolution and a thick file though it requires a more thoughtful professional approach to shooting, the Sony is fast and runs for 10 hours but it takes care in shooting not to blow highlights and some post work to make it un video looking, the 5d can be amazing or awful depending on the subject. Also the 5d2 can get hot and degrade the image so care has to be taken in how long you run a series of sequences.
To match our cameras the one thing that seems to help is to go to ziess lenses all around. I can visibly see more detail and sharpness from that change that anything else.
What I don't think the original poster realizes is the smaller cameras usually shoot 8 to 10 bits, and are 4:1:1 or 4:2:2 where the RED is 12 bits and 4:4:4. Also in any common NLE it all goes back to 10 bits.
I also don't think he realizes that even if you drop 50 grand that's not a lot in the motion imaging world. Not for a camera, heck not for anything.
I can't show it because our last few projects haven't been released, though the client selected 16 still images from the RED and after post production and some retouching, anyone would be hard pressed to know if they were shot with a still or motion camera.
IMO
BC
-
No, we, not working in Canon engineering, don't know exactly what is going on - we can only measure the results and see if they're suitable or unsuitable for our needs, and that's an individual choice.
I do know on the RED side what we do there, and it's basically three things: a sensor capable of a high native full-sensor fps, serious horse power in the chips to deal with that data, and a really nice raw compression engine. Each one is expensive enough on it's own, and massive data throughput also costs more in terms of heat dissipation too.
Every camera does indeed have it's pros and cons, which is why it's vital you test a camera for it's intended use - and as you point out, you really need to test a "camera system" and not ignore the contribution good glass makes to the image.
Graeme
-
No, we, not working in Canon engineering, don't know exactly what is going on - we can only measure the results and see if they're suitable or unsuitable for our needs, and that's an individual choice.
I do know on the RED side what we do there, and it's basically three things: a sensor capable of a high native full-sensor fps, serious horse power in the chips to deal with that data, and a really nice raw compression engine. Each one is expensive enough on it's own, and massive data throughput also costs more in terms of heat dissipation too.
Every camera does indeed have it's pros and cons, which is why it's vital you test a camera for it's intended use - and as you point out, you really need to test a "camera system" and not ignore the contribution good glass makes to the image.
Graeme
Graeme, You should be decorated of the merit cross ! for those words indeed.
In the photo discutions at least, it's sensor, sensor, and more sensor. But much more discrete are the lenses impacts. It's tremendous in digital. IMO, lenses in digital are 70% of the final result.
You can notice it on the "high" but even on the "low". For ex in the "crap" side, there is a fashionable imagery now where a lot of people apply vigneting within digital filters. It looks flat, redneck and horrible. Put a C mount good glass on a digital and you obtain an analogic way of natural vigneting that can be truly beautifull. Digital will not render that way. Even in the crappy really crap, used properlly, certain glasses like the CTVlenses (but not all) can give a sort of view camera look to a video footage. In the high-end, no need to say that the lens contribution to the image is key in one way or another and determines a lot the personality.
A good sensor with a regular lens is like having a regular sensor and vice-versa. Of course a good sensor and top lens is the grail.(but not necessary always)
But in the end it's more a question of personality of the combinations. One tool tends to give that look, the other this look...it's like a paintor's palette. Each camera is a palette and each lens a color. Does a painter use just one color and palette?
-
Lenses can make the world of a difference, but in the case we're talking about here, they can't make up for how the 5D2 reads the sensor. But in the case of the RED where we don't shortcut like that, the camera responds well to the character of the lens, and will let it show of it's qualities, good, bad or otherwise - same with filtration, which can sometimes really help achieve the image you want.
-
Lenses can make the world of a difference, but in the case we're talking about here, they can't make up for how the 5D2 reads the sensor. But in the case of the RED where we don't shortcut like that, the camera responds well to the character of the lens, and will let it show of it's qualities, good, bad or otherwise - same with filtration, which can sometimes really help achieve the image you want.
Yes, I can see perfectly what you mean. But I think Canon has a particular approach in their electronics. Even in stills something very similar happens.
The Canon's look (if there is such a thing but I think there really is) has always been acheived at the expense of a certain amount of resolution. But...they actually deliver very plaisant results right-out-the box.
Very specially (and I really don't know why but sees it) when humans are involved. But whatever you mount on them, you'll have the Canon's touch, not that much the lens touch.
In the congress here, I was amazed to see that among all press photographers, 100% shooted Canon. Not just one Nikon. All Canons. Even in sports, you see logically more Nikons but Canon is still very present.
Their approach is maybe not technically the most rigourous in terms of preserving the maximum sensor performance but for a lot of pros, the delivered look just works bloody fine and + or - hassle free. I think that it is maybe the real success reason of that brand and why despite its questionable resolution aproach you find those cameras almost everywhere in prod.
Red One just plays in another league and amazingly cheap for such products.
-
Lenses can make the world of a difference, but in the case we're talking about here, they can't make up for how the 5D2 reads the sensor. But in the case of the RED where we don't shortcut like that, the camera responds well to the character of the lens, and will let it show of it's qualities, good, bad or otherwise - same with filtration, which can sometimes really help achieve the image you want.
Graeme,
I agree, but for the price the Canon is pretty amazing.
The only issue with Canon is they backdoored the camera for video. It's only been like 100 million times that people have asked for a physically larger sensor in digital motion and you'd think they'd asked for the security code to the pentagon.
Speaking of that, I'd have an Epic sitting in my garage (uh I mean studio) if I could just buy one of the bloody things without going on list 12, stage 6, item 4, see sub paragraph 9.
Do yourself a favor. Put on one of those mission impossible masks and go into your hollywood store looking like a normal guy and ask a few questions. Even little stuff like EVF's and the guys at the counter, look at the floor, and say uh I can't answer that cause my sister is coming into town, I have to wash my hair, my car broke down on Sunset, I just work here . . .
You guys make amazing stuff, but getting information is tough. I think I'll send a note to that Wikileaks guy and see if he can get on it.
Really, though, at the end of the day,if the story is good, the cinematography interesting, nobody that views this stuff seems to notice.
We're starting a round the world gig and shot the test story on the RED mixed with some 5d2 footage. Nobody said anything other than they like it and what they wanted to change, but nothing bout' cameras.
So please, please, don't be like the still medium format guys and keep information so clubby and hidden. Get those cameras on the shelves ready to sell.
IMO
BC
P.s. back to the lenses. I have all kinds but when I use the Zeiss still lenses (not the PL mount) I can see a huge noticeale difference in the look of the image, especially on the RED.
-
Working on getting the cameras out there. It's been tricky to say the least.
The guys at the Hollywood store know me too well, and I do make a point of catching up with them regularly and going over their questions and making sure they're up to speed on that side of things.
For good info - my email is wildly known and I'm pretty easy to get hold of - I can answer a lot of the image processing side of things, and put people in touch or get other kinds of answers if it's not something that's been posted about on reduser.net.
Indeed, if the story is good, the dialogue clear and the cinematography great, not much else matters, other than if those aspects are really good, you don't want to spoil the result with a less than optimal image on the technical side of things. I believe that for every use and budget there's an optimum camera. To think otherwise is silly.
Different lenses really do alter things - a lot of it is due to the contrast of the lens from the different coatings used. Just the other day I was shooting on Epic with Canon mount, a Sigma 50mm F1.4, and then some old Olympus OM mount glass (via adapter) - a wide angle and a zoom, and they looked really nice too.
Graeme
-
Coating is key. Some lenses only work fine with a proper hood and can pass from average to stellar with the hood. I tend to avoid however the purple coatings because they generally have a strong cast and wb is weired to get it right. On the vintage Zeiss primes that work brillantly in motion, the yellowish coating is important. Included I learned that the Russian copies have been build with 2 different coating. The oldest the best, yellow tone or neutral-warm but a hood is obligatory with them.
I had many OM (olympus was my first camera ever) but if you could try the Pentaxes on the Epic, specially the Takumar and super Takumar you will be amazed. Now my dealers are in Germany, specially the former eastern where you trully find some impressive lenses made in USSR in the middle of a lot of impressive crapperies too, it's true. They feature stepless aperture and generally a generous number of blades.
Built quality is below the western standarts but the glasses are superb.
If you go 39mm mount be carrefull of one point: there has been versions done for certain reflex that used the 39mm mount like the rangefinders. The result is that with any 39mm adaptor you will never get infinity focus because those adaptors are made with the Leica calculation in mind. When you see 2 M39 released models with different shapes for the same model it is the case. So carefull on the Leica 39 mount offers because most of vendors don't even know it.
Some PL mount Angenieux are amazing.(but not cheap)
I love it. It's much better that going to the shopping center! Those vintageries on a today's camera are working really well if like me you're on a budget. They have now a second life!
-
Coating is key. Some lenses only work fine with a proper hood and can pass from average to stellar with the hood.
Which would suggest that the hood is key ;) . Actually, they both are key.
I tend to avoid however the purple coatings because they generally have a strong cast and wb is weired to get it right. On the vintage Zeiss primes that work brillantly in motion, the yellowish coating is important. Included I learned that the Russian copies have been build with 2 different coating. The oldest the best, yellow tone or neutral-warm but a hood is obligatory with them.
The coating color you can see is mostly contributed by the most outer coating. It requires the whole set of lenses/groups to be adequately coated, and the interior of the lens barrel, and the edges of the lens groups to be blackened. Modern lens coatings exceed the more simple 'wavelength suppression by internal cancellation' principle, which only works best when rays hit the air/glass surface perpendicularly.
Cheers,
Bart
-
HDMI out on the 5D Mark II plays at 480p.
This demo reel was shot on 5D mark II and is super sharp:
http://www.dananeibert.com/main/motion.php
-
Y'all might want to look at Part II of the single sensor camera tests (resolution, noise, sensitivity, etc) at Zacuto.com. The contrast between the 5D2 still shots and it's video resolution (vs higher end cameras) stirrs-up some commentary.
-
Y'all might want to look at Part II of the single sensor camera tests (resolution, noise, sensitivity, etc) at Zacuto.com. The contrast between the 5D2 still shots and it's video resolution (vs higher end cameras) stirrs-up some commentary.
Perhaps you would be good enough to summarize their findings for us?
Thanks.
-
Won't summarize their findings as they actually let the viewers come to their on conclusions. But suffice it to say, it jives with what your OP was about. That said, when you're comparing the video performance aspects of cameras from the Arri Alexa and it's snack-bracket competitors down thru the 5DII, 7D & D7K for video quality.... The two videos in the series (with a 3rd to come) make so a good watch.
-
That said, when you're comparing the video performance aspects of cameras from the Arri Alexa and it's snack-bracket competitors down thru the 5DII, 7D & D7K for video quality.... The two videos in the series (with a 3rd to come) make so a good watch.
Sorry, but I don't understand this sentence. 'Snack-bracket' means what exactly? And how does the 5DII compare to the Alexa?
-
'snack bracket' -> price range
As for the 5DII (or any SLR) vs. Alexa, Phantom, Sony F3, Red (can't recall model) etc., in their test - when it comes to video resolution, there was no comparison.
http://www.zacuto.com/
-
As for the 5DII (or any SLR) vs. Alexa, Phantom, Sony F3, Red (can't recall model) etc., in their test - when it comes to video resolution, there was no comparison.
http://www.zacuto.com/
When you say 'there was no comparison', does that mean they did not compare the 5DII to the Alexa, etc., or that the Alexa, etc. resolution was so much better than the 5DII?
-
The higher-end (dedicated motion) cameras held much more detail and color accuracy (in part, as mentioned earlier in thread, due to compression used) than the DSLRs.
-
The higher-end (dedicated motion) cameras held much more detail and color accuracy (in part, as mentioned earlier in thread, due to compression used) than the DSLRs.
How do you know that it is the compression?
-h
-
The higher-end (dedicated motion) cameras held much more detail and color accuracy (in part, as mentioned earlier in thread, due to compression used) than the DSLRs.
Have you guys heard of the Ptool hack for the Panasonic GH1 and, as of last month, the GH2? With the firmware hacked (simple 5 minute procedure) the GH1 becomes a GH13 shooting AVCHD @ 50mbps 24p/25p 1080; the GH2 settings still in development but AVCHD @ 40mbps and superb quality intraframe 1080p MJPEG @ 100mbps.
Easily out-performs image quality and low light performance of many "dedicated" high end video cameras (without a Nanoflash type device). The Canon 5D2 doesn't come close (sensor line-skipping = low 600 line resolution + moire+aliasing; relatively low bit rate) ...I wasted my time with one for a year before the GH13 arrived.
-
That hack sounds interesting..
"....(in part, as mentioned earlier in thread, due to compression used)....". The testers discuss what is likely contributing to the video performance of the DSLRS vs dedicated higher-end units. Beyond that www.zacuto.com.
-
Compression plays a part, but the compression can only compress what it receives as an image, which is severely compromised by the line-skipping sensor sampling methods used. In other words, capturing an uncompressed feed still shows you the majority of the camera imaging issues.
Graeme