PDA

View Full Version : Still vs Video IQ - Canon is still not video IQ



philiplipetz
10-05-2009, 11:23 AM
As far as the image quality goes there are two schools of IQ. Canon has always been part of the school that goes for the maximum image impact straight out of the camera, lots of saturation and sharpening. The problem with these images is that they can not be easily adjusted in post production. That is why so many still photographers shoot RAW images.

Panasonic and Nikon go for a softer and less saturated image that can be more easily manipulated in post production. Since video images, except RED, can not be shot in RAW it becomes a question of whether you want to give up ultimate control for a quick out of the box solution. While I agree that GH1 images can be horrible, the good one I have seen are either shot in strong light or have been subjected to post production color grading, usually with Magic Bullet. Then they are wonderfully cinema like!

The Canon approach produces images that are so detailed that eye cannot be easily directed unless you use shallow DOF pull focus shots. Everyone complains about "video-like" images and tries to avoid them, but even the 24p stuff from Canon is so detailed that it seems to be HYPER video-like unless you use shallow DOF. Details like this makes for great still images, they make for great pixel-peeking, but does the Canon approach lend itself to great story telling with an audience used to film and its compromises?

There are substantial differences between desirable IQ in still and video cameras and I wonder why so many people that apply still standards to their video pixel-peeking. That works only for the experienced DPs like Philip Bloom who know how to use, and obscure, detail. When you take the great detailed IQ capacity of the Canons and focus it on only the subject, then you have the best of both worlds. And this is the domain of experienced DPs in studio and location shooting with substantial rigs. NO doubt this the ultimate use of the Canons.

However, I am making a documentary. Some of it involves stealth shooting. Except for the fixed interviews there is no possibility of using large rigs. Forget the Canons. Forget pull focus. Will the GH1 do it? Do not know. Only got to play with the 7d.

Then I am concerned that I will have to take footage shot in wildly different lighting and make it all conform to a single look. I have serious doubts that the Canon footage will let me have that color grading latitude. I hope I am wrong.

All of this will probably change with the DIGIC V video optimized processor rumored to be introduced on the 1dMkIV. Then we will hopefully have video images not a series of still images. They are very different things. Of course, by then we will have the GH2 and the Fuji m4/3. And there are rumors that Canon will create a large sensor video camera that will be self contained.

Before everyone starts to attack me because I do not own these rigs, because I am only a low level indie film maker wannabe, that is the point of my post. Just because these cameras can be manipulated with addons and TLC to produce unbelievably good results under controlled conditions that does not mean that they are optimal video solutions for most people. I am not Philip Bloom, and will never be that good.

If you put me in a F1 race car I could barely drive it. Only an expert can. It will get me from point a to point b but it will never perform as it does for Lewis Hamilton with his huge support team.

So far the Canon stuff I have seen posted by us mere mortals is detailed but fatiguing to watch. The stuff posted by the great videographers is beyond belief, with no viewing fatigue. So I must ask if great films can be made by story tellers who are not great DPs.

I am asking for a solution that is designed for video, not still images. I do not want have jump through hoops to overcome the still orientation of the IQ.

Say want you want about the many real shortcomings of the GH1, but it is still the only hybrid DSLR more oriented to video IQ than still. And there is a difference between the optimal still and video IQ.

Still images are designed to capture a moment in time and so you explore every detail of that moment. Video images are designed to capture the flow of time and so the details as not as important since they will pass too quickly for every detail to be absorbed. What matters is a sense of flow, and that requires obscuring detail with shallow focus, color grading, or softness. Time flow is a different quality than still.

Sacrufuce
10-05-2009, 11:34 AM
So you say you are doing documentary "stealh" shots? So i guess allot of handheld? on the move always ready type of action?

In that case i cant really see why you ever want to go for really any of the DSLRs, Can look nice in controlled situations, but in my opinion they are a really bad match for fast on the move action. Jello cam, no real Autofocus, no quick NDs, no preset color temperatures, to shallow DoF for most work, no motorised zoom. well just my 2 cents. :happy:

philiplipetz
10-05-2009, 11:47 AM
I need to film at places that allows still but not video cameras. Especially India. Also, I have found that big video rigs intimidate people while hybrid DSLRs do not.

samphibbs
10-05-2009, 12:06 PM
Hey Philip,

You can make a custom preset to flatten the picture by turning down contrast, saturation and enabling highlight tone priority, this will give you a very flat picture for grading a little like cinelike d gamma on the dvx or hvx. This will allow for alot of grading options. I have also found the codec to hold up very well to colour correction even secondary/ isolating of colours. It even chroma keys very well in my experience and I have alot of experience using the hvx for greenscreen work. By the way I dont have experience with the 7d yet, I only have experience with the 5d though they should be similar performing. The real problem you should think of with the 7d is overheating especially in India! Can other members chime in on this issue?

philiplipetz
10-05-2009, 12:25 PM
Thanks for reminding me to set everything to reduce saturation, NR and sharpening. However helpful that may be I still feel there is a substantial difference between stacked still images and video images.

Video codecs encode differences in time and this reflects the way the eye sees. Still images do not see difference. Film is about the flow of difference.

Yes, Canon runs the sill images from Live View through a video codec but AFTER they have been recorded as still images. This means that fundamentally they do not look at differences in time, but differences between still images.

Canon has acknowledged that is is a major problem. This is why one of their engineers said that felt the 5d would only be used for Internet videos. That is why the DIGIC V processor will have a video core added on. Video requires a different sort of processing than is done with the stacked still images of today's hybrid cameras. Wish I could afford a 1dMkIV when it comes with a video core. Aliasing and moire are more pronounced in stack still images than in properly encoded video.

I know that what I am talking about may seem subtle but my experience is that it is not, especially in a feature length film when viewing fatigue amy set in.

I am reminded of the audiophile world where it was found that adding high frequency distortion created sound that was captivating in the short term but caused listener fatigue in the long term. That is my personal impression of today's hybrid DSLRs.

I still do not understand why so many people here talk about abandoning the video look for the smeared look of 24p film but then love the video like details of the Canon look.

All of this will be academic when the new cameras appear. Canon engineers know the difference and so they can, and will, deliver a distinctly superior product. Panasonic has already shown it understands the difference, only the GH1 was a flawed implementation. However, the GH1 handles more like a video camera because they understand the difference.

Ian-T
10-05-2009, 12:50 PM
Sure, the Canon engineers didn’t think the DSLR would go beyond the Web but in terms of motion I don’t see a difference at all. The handling of a camera is one thing but (IMO) getting the best possible image from your gear is the more important goal (putting aside content for discussion’s sake). I’m not sure I’m following the issue here but isn’t how a (film) movie camera capture it's images the same as a DSLR capturing a still image? If anything was different wouldn’t it be traditional video cameras?

Yuor example of audio distortion was a good one…but that had more to do with audio compression and how loud music is mixed today as compared to 10 or even 20 years ago. That affected the dynamic range of audio in a big way and causes listening fatigue…but guess what….the industry never looked back. What I mean is just about every main stream recording is mixed that way today. The same will happen with “video.” I personally like that each frame looks like a DSLR picture. What we need is more control in tuning that down to get a more organic look. There are many ways to accomplish this. But ultimately RAW (like) recording is probably the answer.

visugeek
10-05-2009, 01:27 PM
You are also going to want something that can shoot longer takes for a documentary. The 7D is limited to 12 minute takes, that might not work well for an interview environment.

I think it would be a good candidate for stealth video, but you may not get the sound that you need with it alone, and might need a sound guy.

As far as detail goes, that comes with resolution. If the image looks too contrasty to you, adjust your settings to dial that down. The 7d has a lot of flexibility, it's brand new, and is going to take a while for people to get really dialed in on how to use it well. That's part of the fun.

You might want to rent a 5D for a weekend and play with it before making your decision. You can DL the manual from Canon's web site and really get to know how to adjust the settings and then be a better judge of what it or the 7D are capable of doing in your hands. I've seen mediocre images made with great cameras, and great images made with mediocre cameras.

Most cinematographers do test shoots before a feature with cameras, film stocks, get to know the labs, etc, so there are no questions about whether the camera will meet their needs once they start production. If they are using a Digital pipeline, they test it through, from camera to editorial, to grading making sure that it is going to do the job for them. The success of their project relys on their making the best possible choices, and they don't leave that up to chance. They investigate everything.

As far as footage matching from shot to shot, that is always a problem on any shoot. Any film that you see with exteriors will have lighting continuity errors. The sun is the great humbler you wait on it, it doesn't wait on you. You have to try to match as best you can when you shoot, and then fix where you can in post, and hope the story is compelling enough that the audience isn't too distracted by it when they see it.

Cameras are only tools for telling a story, find the one that works best for the story that you want to tell, and for the techniques that you want to use to capture that story. No camera is perfect for every situation. The 7D is a good tool, so is the 5D, the GH1, the HVX's, the Sony Ex's. Don't get too hung up on tech specs that are going to be ancient history in a year. Do the best work you can with the tools that you have available that you can afford.

Phillip Bloom is as good as he is because he practices...a lot! There was a point where he wasn't as good as he is now, but by spending a lot of time just shooting, learning his camera, lighting, and composition, he got better. If you want to be good, that is the only way, put in the time. His rig for his shoots was very stripped down, a Z-finder, and a mic I believe. Check out his site and see.

There is no secret to getting good, no magic box that makes pretty pictures, just time spent learning.

Cheers!

philiplipetz
10-05-2009, 01:28 PM
There is a fallacy that film on photographic stock is a series of stacked still images, and therefore a hybrid DSLR can emulated film by stacking images captured from live view. Film on photographic stock is captured at slower shutter speeds so that a moving image, usually where you want the eye to go, is always subtlety smeared, especially on the trailing edges. We have grown accustomed to this subtle visual clue. It is what gives film a sense of time, there is history in every frame - where the subject was and where it is going. This is lost in current hybrid DSLRs. This sense of time is incorporated into all video codecs. They only encode for the moving parts. This gives the subtle clues that come close to emulating the clues inherent in photographic stock films. Displaying stacked images, even when run through a video codec, has almost none of this sense of history that the viewer relies upon. These "time history" clues are the difference between video looks and film looks.

Yes, you can shoot video at slow shutter speeds but there is something about the video sensor that creates a difference. It is this. A CMOS video shutter, such as on the Canons, works by scanning quickly from the top of the sensor to the botton. That means that a CMOS image shot at "1/50" is really a series of lines shot at speeds many times faster. That is why CMOS video cameras loose the sense of time. In contrast, CCD cameras, such as high end video equipment, have global shutters where a speed of 1/50 means that each line was exposed for that long. There is no stacking of quickly exposed lines. That is why CCD cameras still reign at the high end of our video world.

Michael Olsen
10-05-2009, 02:26 PM
Yes, you can shoot video at slow shutter speeds but there is something about the video sensor that creates a difference. It is this. A CMOS video shutter, such as on the Canons, works by scanning quickly from the top of the sensor to the botton. That means that a CMOS image shot at "1/50" is really a series of lines shot at speeds many times faster. That is why CMOS video cameras loose the sense of time. In contrast, CCD cameras, such as high end video equipment, have global shutters where a speed of 1/50 means that each line was exposed for that long. There is no stacking of quickly exposed lines. That is why CCD cameras still reign at the high end of our video world.

This has been an interesting read, thanks for all the thoughts.

This quote in particular, makes me want to chime in. Film cameras use a rolling shutter, though horizontal, rather than vertical. I have always wondered why everyone comments on getting the look closest to that of film, which has a rolling shutter (and indeed, rolling shutter artifacts), but lamented the skew in the DSLRs. Then I saw what they were talking about and I wondered, how can a rolling shutter produce two decidedly different effects? It turns out, the shutter on film cameras rolls significantly faster than current CMOS readouts, around 4 microseconds (I think RED ONE is about 13, for comparison). So to get closest to the film look, I would want:

1. An S35 sized sensor.
2. A horizontally rolling shutter.
3. A shutter that rolls at approximately 4 microseconds.
4. A variable shutter speed, capable of real 180 degree shutter speeds (1/48).

The 7D gets us only #1.

Near future cameras (RED EPIC) promise us 1, 3, and 4.

I'd imagine in five years or less, we'll have all four for under $10K.

I just wanted to throw my 2 cents in: I think global shutter will be usurped by faster electronic shutters, at least for cinema cameras.

Spartacus
10-05-2009, 02:34 PM
No offence phillip, but this is just a philosophical point youīre malking, that wonīt matter in real life IMHO.
I get your point, but this "time history clues" should only matter to people strictly conditioned to them.
Itīs all a matter of perception, there is no "rule" you are describing, just a technical "is" that now is a "can be".
If you could show stuff shot on HDSRL and a Cinaalta to someone who has never seen a film before, he would never point to the HDSLR footage and say "Hey, where are those time history clues?".
Perception will simply adapt.

philiplipetz
10-05-2009, 02:53 PM
Yes, like the audio world, we will grow to have different expectations. For good or bad. But this is not a philosophical point to today's audiences. Their brains have adapted to a set of visual clues and feel comfortable only when those clues are present.

Spartacus
10-05-2009, 03:33 PM
Yes, like the audio world, we will grow to have different expectations. For good or bad. But this is not a philosophical point to today's audiences. Their brains have adapted to a set of visual clues and feel comfortable only when those clues are present.
Yeah, this is your thesis, but do you have any research on this to back it up?

We watch video on phones, mp3 players, Car-LCDs, Plasmas, Beamers, LED walls, Oleds.
Videos of all kind and technical origin, itīs not like the way HDSLRs record moving images is the only thing not following the "Lumiere Laws"...

commanderspike
10-05-2009, 03:52 PM
I've been flirting with the idea of swapping my GH1 for a 5D but I probably won't do it, for exactly the reasons Philip outlined in his posts. The GH1 is as true a video camera as you are going to get from the DSLR market until way into next year.

Why? Well, it shoots 24p in a video format - AVCHD. That's not to be underestimated. And most importantly it HANDLES like a video camera - and a very small, portable one at that. Okay the codec doesn't quite keep up with the sensor, which is the main reason I considered jumping ship, but having now actually used the 5D for 9 days I just HATE handling it as a video camera. It's not a video camera, it's an SLR and has been designed from the ground up as a stills camera.

The 7D is exactly the same in this regard, in my opinion. I hate the way it still has a mirror box. It offers absolutely nothing genuinely new over the 5D other than 24p.

I can understand why many of you guys are excited - for good reason. You will shoot some wicked footage with it.

Okay I admit the lure of the sexy new 7D is strong and the attraction of the 5D's full frame IQ (and it's ability to get the most out of 3 decades worth of amazing SLR lenses) is obvious, but I can't help being drawn back to using the GH1 because, for me, it feels the most fun to go shooting with the GH1 on so many levels.

And for a sensor which is half the size of the 5D's, it actually stacks up pretty well in terms of ISO and low light performance considering.

Isaac_Brody
10-05-2009, 04:04 PM
I've been flirting with the idea of swapping my GH1 for a 5D but I probably won't do it, for exactly the reasons Philip outlined in his posts. The GH1 is as true a video camera as you are going to get from the DSLR market until way into next year.

You love your GH1, great, go talk about it in the GH1 forum.


Say want you want about the many real shortcomings of the GH1, but it is still the only hybrid DSLR more oriented to video IQ than still. And there is a difference between the optimal still and video IQ.
And Phillip, this camera probably isn't right for you and won't meet your expectations. Your best bet is try a dedicated camcorder, or shoot your documentary on film, or wait for something better to come along. Or go buy a GH1 since it sounds like that will better suit your vision and needs for your project.