Hi

I sold my JVC HD 100a and I`m now interested in 7D, mainly for narrative work.

Searching for info, I`ve read some users saying that the 1080 24p is "soft", and watching some downloaded 1280X720 7D stuff from Vimeu, I have the same "soft" impression, even aware of the diferent camera set up and losses from compression each user applied in editing process.

So my question for a Canon 7D owner that have seen its recorded original stuff played in a real 1080p display: is the 7D 1080 24p camera original resolution correct (like Sony EX1)?

Thanks
Ron

is the 7D 1080 24p camera original resolution correct (like Sony EX1)?

1. Are you asking if the original is soft or not? A lot of people are turning the detail setting all the way down, but I don't feel that my footage is overly "soft" or not sharp on things in focus.

2. Are you asking if the 1080p is full-frame and not some horizontal stretch? Yes, it's full 1920x1080 like the EX1.

I sold my HD100ua also. Of course, you can't shoot whip pans or go handheld without some rig (so the pivot point isn't so sharp in your hands).

I shot an 2min interview a year ago with my HD100. The client wasn't pleased with their reading so she wanted to shoot again. It was my first shoot on the 7D. She actually wore the same clothes and it was in the same location with the same lights. The difference is night and day. The 7D (with a fast lens) lets you create some beautiful shots.

Thank you James
I`m also going to turn the detail setting down, because I`m looking for a more filmic image. And I know 7D is full 1920x1080 like the EX1.
Maybe my question could be if it mesures like 1920x1080 in a resolution chart, or...if in similar set up (sharpness, contrast, saturation, etc), the "impression" of resolution is similar to a wel known 1920X1080 camera.
Best
Ron

In terms of "real resolution", the 7D (and all DSLRs, such as the GH1, 5D, and K-7) aren't very high. They're better than 480p, but not as high as 720p. They look much sharper than that, but in terms of actual solid resolved detail as viewed on a resolution chart, it's not encouraging. They gain a ton of additional "sharpness" through aliasing, which can look good but can also backfire (moire, stairstepping lines, shimmering, etc). As compared to a modern known video camera, none of the DSLRs can deliver an image as sharp as that from an EX1, HPX300, or even a little HMC40.

Thank you Barry.
Do you think this lack in DSLR actual solid resolved detail (actual resolution) could damage a 35mm blow up or a theater digital projection, despite the great increase in depth / volume the image gain because the large sensor?
Best
Ron

I am not Barry, but my guess to your question is Yes, it would probably look worse as a 35mm theatrical blow up than a more robust HD video camera.

These cams make lovely images though and many of the shortcomings are only noticed by pixel-peepers and not a normal "audience".

Thank you Benjamin.
Ron

I shot a short narrative film two weeks ago that's roughly 50/50 7D and EX1.

I transcoded the H.264 7D files to XDCAM and cut everything in Final Cut.

Chris Hurd loaned me his 7D for the project and in exchange I've promised him an in-depth article about the shoot and the results.

But the short answer, as it relates to this thread, is that the 7D footage looks better than the EX 1 footage even at 800 ASA. I've seen my rough cut projected on a 2K projector on a 16 foot screen and it looks amazing.

I'm prepping to shoot pick-ups next week, but I should have footage, stills and an account of the process available within a few weeks.

I'm not sure I agree with Barry. I will say that I have an EX1 and just sold my 5dmk2 for a 7d, and the ex1 is the sharpest and least prone to artifacting for sure. It is probably also the best equipped for a sharp blowup for 35mm projection. On the other hand, my old 5dmk2 was sharper than my new 7d, and both are quite sharp when downscaled (with a good scaler) to 720p from their native 1080p. I'd say if you want sharp 1080p, EX1 will just barely give you what you need and a camera that oversamples 1080p is probably the way to go (red, f23,f35,genesis,35mm film, viper, etc). If sharp 720p is what you need, then ex1 or 5dmk2 or 7d downscaled in post from 1080p source to 720p are all quite solid options.

I do not, however have charts to back this up, this is largely based on my experience. Maybe ill put a chart in front of my ex1 and 7d to see for sure.

I don't agree with Barry either. I have the Canon HV20 and it's very good looking and it's sharpness seems to match the 7d pretty well (in cinemode). My older Sony HC1 clearly has less resolution than either.

Thank you Barry.
Do you think this lack in DSLR actual solid resolved detail (actual resolution) could damage a 35mm blow up or a theater digital projection, despite the great increase in depth / volume the image gain because the large sensor?
Depends on what you've shot. If you're dealing with largely natural subjects, there's not much in the way of sharp straight lines and repeating patterns of detail to cause moire, so the aliased information would make the image look very sharp and may not show any clearly obvious drawbacks.

But if you're shooting a row of houses with shingled roofs and chain-link fences, with thin power lines running overhead, then the resulting aliasing and potential moire might look pretty nasty on a blowup.

I do not, however have charts to back this up, this is largely based on my experience. Maybe ill put a chart in front of my ex1 and 7d to see for sure.
Please do. And prepare to be shocked when you do. Then you might come back to this thread and think differently. In 720p mode a 7D can't even resolve 400 lines, and in 1080 mode it barely ekes out 500. The rest is all aliasing. And the 5D and GH1 aren't any better, they all make their "sharpness" through aliasing.

I used a few charts, mainly my DSC Labs' MegaTrumpets 12 (which is designed to measure resolution up to 4K) and my Chroma Du Monde Billups VF/X+, which shows res lines out to 1000. When you see how the DSLR cameras perform on a chart, you'll see that there are very significant compromises in the way they gain their sharpness, and you'll also see that the actual measurable resolution (sans aliasing) is not very high at all.

I don't agree with Barry either. I have the Canon HV20 and it's very good looking and it's sharpness seems to match the 7d pretty well (in cinemode). My older Sony HC1 clearly has less resolution than either.
A Canon HV20 will have quite a bit higher resolution than a 7D.

There is a difference between resolution and sharpness. Aliasing adds a ton of perceived "sharpness" but it is not resolved detail, it's spurious image contamination that -- in many conditions -- looks "good".

I know that I'm responding to the legendary Barry Green, whose encyclopedic knowledge of video production issues has been very useful to me in the past, but, I do have to say that in the real world experience of these cameras (which I know is highly subjective) the image quality (resolution, color depth, sharpness, etc.) is incomparable. The HV20 looks like consumer HD, nice enough if that's your thing, and the 5D Mark II (sorry, no 7d yet for me) is a world better. Breathtakingly better.

Everyone I've shown footage from the 5d too, even (especially) people with no experience in video production are blown away by the images the 5d captures, even with a cheapo 1.8 50mm on there. I shot something with the HV20 the same week and it was ho hum.

Also, can you explain how resolution (both of which are 1080p) is better one on than the other? You obviously aren't talking about raw lines of resolution, right?

Because in your above post it kind of sounds like Canon is practicing false advertising claiming that the 5d and 7d shoot full 1080p video, if you are saying it resolves only 500 lines.

I mean, when I have stuff from the 5d on my 1080p 46inch lcd it looks incredible. Not so much, particularly color reproduction, with the HV20.

All that to say, if we charted these cameras up, one might have statistically better video resolution, but I'll tell you right now which one "looks" better. And I'd be shocked if 100% of people you showed side by side footage to of each camera didn't agree with me.

I'd be thrilled to show something shot really well on a 5d to a client, not so much on the HV20, even if it were the same project, shot the same way.

I'm with Barry.

...I do have to say that in the real world experience of these cameras (which I know is highly subjective) the image quality (resolution, color depth, sharpness, etc.) is incomparable. The HV20 looks like consumer HD, nice enough if that's your thing, and the 5D Mark II (sorry, no 7d yet for me) is a world better. Breathtakingly better.


Have you tried shooting them both at the same DOF and angle of view? I think thin DOF is the one thing people like most about the 5D2. If you take that away, I think the HV20 trounces it soundly:

http://www.dvinfo.net/forum/1132912-post31.html

The 5D2 has other advantages, though, including higher contrast (with the right lenses), far greater color depth, and more dynamic range.


All that to say, if we charted these cameras up, one might have statistically better video resolution, but I'll tell you right now which one "looks" better. And I'd be shocked if 100% of people you showed side by side footage to of each camera didn't agree with me.


Take a nice, natural, unprocessed raw clip from the RED ONE. Apply normal sharpening to one and ridiculously over-the-top sharpening to the other. Guess which one 100% of people will pick "looks" better? The oversharpened one.

Now apply normal saturation to one version of the raw file, but oversaturate the other raw conversion until the grass is bright neon. People will pick the neon grass every time.

Leave a nice, small amount of noise in one version, providing subtle texture throughout. Apply heavy-handed plastic-like noise reduction to the other. 99% of people pick the plastic.

Convert one to use a full 9 stops of dynamic range with detail in the blacks and whites. In the other version clip the whites and crush the blacks for extremely high contrast. People pick the high contrast image every time.

Convert one to show the maximum possible detail at 1920x1080 with no aliasing artifacts. Make another version that has barely 720x480 worth of resolution, but is riddled with aliasing artifacts. Most viewers will pick the junky aliased image.

Render the audio to use a 40 dB of dynamic range and compare it with one that is compressed to within an inch of its life. Listeners pick the "louder" compressed version.

Camera manufacturers know this well, so they tune their cameras for maximum sharpening, saturation, contrast, etc. The results that come out of most cameras is terrible to me, but it sells cameras. The ultra-strong aliasing of the 5D2 is in the same category as the extreme saturation, NR, sharpening, contrast, etc. adjustments applied to digicams to make them appealing to consumers, except in this case there was nothing Canon could do to prevent it.

Agreed, Daniel. Every time I picked up the 5D Mk2 to shoot video, I had to be paranoid about avoiding shingled roofs, close vertical-lined fences, vehicle front grills, people wearing striped shirts etc. Often the moire and other artifacts showed up on the computer screens even without down-converting or scaling. I sometimes think these large-sensor DSLRs are made to shoot specifically background blurring scenes.

in the real world experience of these cameras (which I know is highly subjective) the image quality (resolution, color depth, sharpness, etc.) is incomparable. The HV20 looks like consumer HD, nice enough if that's your thing, and the 5D Mark II (sorry, no 7d yet for me) is a world better. Breathtakingly better.
I know exactly what you mean, and that is what people respond to. But the question was regarding actual resolved resolution. And I stand by my answer.

Also, can you explain how resolution (both of which are 1080p) is better one on than the other?
1080p specifies the recording format. It doesn't specify how much detail is in that shot.

You obviously aren't talking about raw lines of resolution, right?
That's exactly what I am talking about, yes. What I am not talking about, is the recording format. 1080/24p (and 720/60p, etc) are recording formats. They are not "resolutions". They do not specify how much detail is perceivable, they only reference that it's a video recording format with 1080 lines, etc.

You could dub VHS over to 1080p, and have a 1080p recording with VHS resolution. Saying that something records 1080p doesn't imply that you will actually see 1080 discernible lines of detail!

Because in your above post it kind of sounds like Canon is practicing false advertising claiming that the 5d and 7d shoot full 1080p video, if you are saying it resolves only 500 lines.
I'm not saying anything of the sort about false advertising or anything like that!

What I'm saying is that it resolves about 500 lines, maybe 550, which is what it does. Put it on a chart and take a look, you'll see. There is no actual detail beyond about 500, maybe 550 lines, when in 1080 mode. Which, by the way, is about the same as an HVX200, which has been an extremely popular HD camera for nearly four years. The amount of resolved detail is only one aspect of picture quality, and is, in fact, arguably one of the least-important factors.

The HDSLRs let an absolutely massive amount of aliasing through, which goes to help their perception of "sharper" images, but it is not actually resolvable detail. It's false detail, and it's an inaccurate representation of what was actually imaged. In many cases it looks great (because, when shooting a face, who cares whether the hairs being rendered are exactly in the right position or not? Who cares if a freckle isn't rendered in the exact spot it is in reality?) In some cases, it backfires and causes objectionable artifacting. But aliasing is not resolved detail.

Back to your question about "500 lines not being 1080" -- you can't find a camera out there, short of maybe an F950 or HPX3700, that can resolve a full 1080 lines. The HV20, etc., that class can handle about 700 or so. An HMC40 can handle over 800. An EX1 or HPX300 can easily resolve 800, and I don't know how much more because the chart I tested them on only went up to 800! And those cameras do so without aliasing.

All that to say, if we charted these cameras up, one might have statistically better video resolution
By a huge margin, yes.

but I'll tell you right now which one "looks" better. And I'd be shocked if 100% of people you showed side by side footage to of each camera didn't agree with me.
Agreed. But charts don't lie, and the problems that you'll see on a chart may very well show up in real world footage. Aliasing causes artifacts like moire, jaggy lines, and -- on the Canons, at least -- color pollution. Look at the chart for a 5D or 7D some time, you'll see some things that are simply not typical of a conventional video camera. Those type of things manifest themselves in rainbow-color moire, weird moving patterns in tiled roofs, chain-link fences, certain patterned fabrics, window blinds, fences, brick walls, anything with a repeating pattern of detail is a potential failure point.

Where the DSLRs excel is in natural shooting -- faces, natural landscapes, etc. Nature doesn't have patterns of repeating detail, like vinyl siding on a house or a pattern of windows on a building or perfectly straight lines. But manmade, that's a different story. Look at this photo and see what happens and the unpredictable nature of it -- same camera, same shot, but when you're at just the wrong distance, moire and colored jaggies spring out:
http://dvxuser.com/barry/7D-Elton-Alias.jpg

I'd be thrilled to show something shot really well on a 5d to a client, not so much on the HV20, even if it were the same project, shot the same way.
Depending on what you shot, I'd agree. But there are some things that you could be disappointed to have to show to a client, such as... oh, a bridal veil at just the right distance that turns into a massive purple/orange blob... or an architectural shot where the windows are a shimmering purple and the roof has a wavy bizarre pattern on it (a few roof examples in this video, shot on my 7D):
video (http://www.motionbox.com/videos/0096d6b31c18e3c08f)
SOMETIMES those things will look fine. But when you're at the right combination of focus and magnification, these false patterns emerge and can ruin the shots.

Point is, aliasing is a two-edged sword. It can make images look far sharper than they have any right to, and it can pollute the images with false detail and false patterns that can be very distracting.

But back to the original question of resolved detail -- shoot a newspaper at various distances with the HV20 and with the 5D or 7D or GH1 or whatever. You'll find that you can easily read and discern the text on the HV20 at significantly smaller type sizes than the DSLRs can resolve. That's because the raw resolving power just isn't there, and aliasing can't fix that.

"sharpness" doesn't equal resolution. The HDSLRs don't have all that much actual resolution. They do offer the perception of sharpness (and artifacts can come with that) but as for pure resolution, they just don't have that much.

Where the DSLRs excel is in natural shooting -- faces, natural landscapes, etc. Nature doesn't have patterns of repeating detail, like vinyl siding on a house or a pattern of windows on a building or perfectly straight lines.


Barry knows this already, but for what it's worth, some people are strongly allergic to aliasing even in natural shooting.

For example, the following natural image has strong aliasing:

http://thebrownings.name/photo/2009-10-aliasing/2009-01-30-3481-rt-400-point.png

And here it is the same image with much less aliasing:

http://thebrownings.name/photo/2009-10-aliasing/2009-01-30-3481-rt-400-lanczos.png

To me, the aliased image looks fake, disjointed, and "digital", whereas the anti-aliased image looks realistic, natural, and "filmic". For me, natural images with aliasing is like looking at the natural world through a screen door. A lot of other people, though, prefer the look of the aliased image.

To me the 7d looks great in low light and night scenes because of all the surrounding illumination it picks up on the subject matter. However, during the daytime the 7d looks nothing special compared to other cameras and its low resovling power starts to show.

Good info and comments.
If we could summarize a too large topic, the main point should be that this DSLR technology for narrative work must be used with a lot of care concerning it`s technical limitations.
Real world experiences like Steve Mims should also be considered.
Ron

Take a nice, natural, unprocessed raw clip from the RED ONE. Apply normal sharpening to one and ridiculously over-the-top sharpening to the other. Guess which one 100% of people will pick "looks" better? The oversharpened one.

Now apply normal saturation to one version of the raw file, but oversaturate the other raw conversion until the grass is bright neon. People will pick the neon grass every time.

Leave a nice, small amount of noise in one version, providing subtle texture throughout. Apply heavy-handed plastic-like noise reduction to the other. 99% of people pick the plastic.

Convert one to use a full 9 stops of dynamic range with detail in the blacks and whites. In the other version clip the whites and crush the blacks for extremely high contrast. People pick the high contrast image every time.

Convert one to show the maximum possible detail at 1920x1080 with no aliasing artifacts. Make another version that has barely 720x480 worth of resolution, but is riddled with aliasing artifacts. Most viewers will pick the junky aliased image.

Render the audio to use a 40 dB of dynamic range and compare it with one that is compressed to within an inch of its life. Listeners pick the "louder" compressed version.



If this truly is the case, then what's the argument here?
Who cares if the 7D doesn't have the resolution of the EX1 if there isn't a demand for it?

It would be a case of picking the camera for the job or perhaps even the shot in hand. If the DSLR has the look required and meets the needs of the story (subject to satisfying any commissioning broadcasters technical requirements etc) that could be the camera to go for. Or, perhaps use it as the B camera for the shots that its strengths are a requirement.

Why is everyone getting their knickers in a knot, the 5D and 7D really are fantastic stills cameras :-) If I was honest though I am using a 7D as a B-Camera on a documentary shoot in a couple of weeks not for the look as much as it's small form and low light capabilities.

If this truly is the case, then what's the argument here?
Who cares if the 7D doesn't have the resolution of the EX1 if there isn't a demand for it?

That's only one possible response: if people want junk, give them junk. If they want to eat a bag-full of high fructose corn syrup for dinner, give it to 'em.

Another possibility is that they only *think* they want junk. On a different level, maybe even subconsciously, they actually like and appreciate the quality image more. They might *say* they want to eat a bag full of candy, but if you give them a 5-star dinner by a world-class chef instead, at the end of it they will actually have liked the dinner better than if they had eaten candy.

"sharpness" doesn't equal resolution. The HDSLRs don't have all that much actual resolution. They do offer the perception of sharpness (and artifacts can come with that) but as for pure resolution, they just don't have that much.

Isn't the reason for the lack of resolution and the aliasing artifacts with HDSLRs due to how the information is read off of the chip and "binned" and compressed? It isn't due to a specific limitation of the CMOS chip and the optics, which should theoretically be capable of spectacular resolution. Ultimately isn't the cure to this problem new software (and I don't mean a firmware update) for reading and "binning" the pixels off of the CMOS chip and compressing the data stream?

If that is so, then the current generation of HDSLRs can be looked at as an early stage of the evolution of such devices for video use. They can produce spectacular results in some circumstances, but they have serious limitations. It seems that these limitations could be addressed in future products.

Whether the companies will do this is up to debate. Perhaps at the price point of current HDSLRs, manufacturers will saddle the cameras with these limitations, and will come out with higher end (more $$$) camcorders using the current chips and lenses of HDSLRs, but improved software, to differentiate the market.

David,

The problem is that at the current technology level, sensors that are the best for stills are not capable of the fast frame rates required by motion.

To get around that, less than the full info. is used from the sensor when a camera like the 5D is in video mode, which allows for faster reads.

But it's not just software, there are real design challenges.

Isn't the reason for the lack of resolution and the aliasing artifacts with HDSLRs due to how the information is read off of the chip and "binned" and compressed?
Yes, exactly.

It isn't due to a specific limitation of the CMOS chip and the optics, which should theoretically be capable of spectacular resolution.
They are capable of spectacular resolution. I shot a side-by-side of the 7D in movie mode, vs. the 7D shooting the same scene in motor-drive stills mode (6fps) and then resized the stills in PhotoShop to 1920x1080. It was shockingly better.

Ultimately isn't the cure to this problem new software (and I don't mean a firmware update) for reading and "binning" the pixels off of the CMOS chip and compressing the data stream?
Yes but no. The whole reason they have to do the pixel binning/line skipping thing is because the sensors cannot be read fast enough to deliver motion at full resolution. These are still-camera chips, with maximum read speeds of maybe 8fps. To get them to 24fps (or even 60fps!) they can't read the whole chip, hence they have to bin pixels and/or skip lines, and it is those processes that cause the aliasing and color mushing to occur.

Plus there's the whole other topic of the anti-alias filter, which you simply cannot do for both modes. Either it's tuned for stills, or for video, but not both. And in the HDSLRs, it's tuned for stills, designed to let a 14-to-18 megapixel image through. And that is what lets all this aliasing through in the video modes.

If that is so, then the current generation of HDSLRs can be looked at as an early stage of the evolution of such devices for video use. They can produce spectacular results in some circumstances, but they have serious limitations. It seems that these limitations could be addressed in future products.
Agreed 1,000%.

Whether the companies will do this is up to debate. Perhaps at the price point of current HDSLRs, manufacturers will saddle the cameras with these limitations, and will come out with higher end (more $$$) camcorders using the current chips and lenses of HDSLRs, but improved software, to differentiate the market.
Not with the current chips -- that's a hardware limitation. One company is addressing the chip read speed and will introduce an HDSLR that will have no such compromises; they should announce that the day after tomorrow.

But let's be clear on something: the companies aren't "saddling" anything with limitations! They're offering low-cost products. It is no more fair to say that they're "saddling" the current cameras with limitations, than it would be to say that a laptop manufacturer is "saddling" their netbooks with Atom chips. It's all about price point. Pay $6,000 and you can have a Core i7 chip. Pay $299, and you get the Atom. That's how it works. Better chips cost more. And these cameras' chips are still-camera chips, not designed for video at all. In terms of video performance, these cameras have given us the Atom chip right now (limited to 8fps read speed). If/when they produce the 60fps or even 120fps barnstormer we all want, I guarantee you it won't cost $1699, any more than a screaming Mac Pro Octo-Core is going to cost $599. You got $599, you get a Mac Mini, that's how it works. You want an Octo-Core Mac Pro, you find $5999 in your wallet for that. The cameras will be the same way.

Daniel,

The aliased image is also slightly warmer, which will cause more people to pick it.

I brought both grabs into Avid and CC'd just the non-aliased one to make it warmer.

I think it looks more appealing.


"AliasingNoCC4" is the aliased file, untouched, just imported into and exported from Avid.

"NoAliasingCCd4" is the non-aliased color corrected in Avid to make it warmer.

Very interesting discussion. I attempted to shoot a close up of a computer screen with the 5D for a narrative film, and the moire was horrible. I got the HV20 out, and it avoided the problem for the most part. Kind of proved Barry's and Daniel's points to me.

That being said, I would rather shoot the 5D (especially when it gets 24p added) over the HVX200 which had been the standard bearer for "Rebel" digital film makers. That camera was poor in lower light situations, and grained up pretty quickly, even in open daylight shade, when using a 35mm adapter. And from Barry Greens assessment, it sounds like the 5D resolves as good as HVX200, and I am guessing that he is referring to that camera before you drop an adapter on it.

And let's face it. Barry and Daniel can tell us we are getting better resolution in the HV20 and its progeny (and I agree as I have one) and in other HD Cameras, but when you slap an adapter on those cameras to get to the depth of field characteristics we are looking for ( and that is, afterall the only reason we shoot this camera over regular video cameras), does the Canon come out on top, even resolution wise ? I would be curious about comments and anyone's testing there.

I for one am thankful, as an enthusiast who shoots no budget films, that I have access to full frame sensor cameras to work. I am also inspired by the professionals who are doing some amazing work out there and selling it, using this camera. And, afterall, the goal we all have is a watchable film.

In that respect, I have to challenge the comparison Dan raises between a bag of candy and a five star meal at a pretentious restaurant..... that may be his preference, but I would wager a lot more out there would opt for a pizza or burger....because of the waste of money associated with the overkill at that five star restaurant. The same would apply to choice of cameras, depending on your need.

That's only one possible response: if people want junk, give them junk. If they want to eat a bag-full of high fructose corn syrup for dinner, give it to 'em.

Another possibility is that they only *think* they want junk. On a different level, maybe even subconsciously, they actually like and appreciate the quality image more. They might *say* they want to eat a bag full of candy, but if you give them a 5-star dinner by a world-class chef instead, at the end of it they will actually have liked the dinner better than if they had eaten candy.

Not meaning to inflame any situation, but I think everyone here has a pretty good idea of who their client base are or at least the scope of their projects overall. I know for my purposes the footage I can achieve with the 7D is going to work just fine for my clients as well as for the viewing public. Having achieved this at a third of the cost of the EX1, my bosses are also pleased.

I think we must go down to earth - including my thelf.

For the price of an Top Photo Cam we became the 5D II / 7D including the zero bucks option to make films with a classical 35mm DoF.

Can we really expect to became for this price a cam that can we compare to a Red One for 20,000 Bucks?

Or can we really expect that this zero extra bucks option will have no limitations compare to an > 5,000 bucks videocam?


But if i read the Statement of Steve Mims - "But the short answer, as it relates to this thread, is that the 7D footage looks better than the EX 1 footage even at 800 ASA. I've seen my rough cut projected on a 2K projector on a 16 foot screen and it looks amazing"

Thats all i need to know and its far more then i have expected one year before.

So one of this cams is good enough to me and if i need more quality i think that in 1 or 2 year canon will present a professional videocam body with large chip for presumable more than 6 - 8 thousand bucks and the opportunity to go on use the same lens.


Sorry for my english - hope its clear what i want tu say.


Daniel

Thanks, Peter. I really appreciate the time you took the analyze and post process the images.

Daniel,

The aliased image is also slightly warmer, which will cause more people to pick it.


I do agree that people will tend to prefer a warmer image. I took another long, hard look at the images, and I do see a lot of differences, but for me, color isn't one of them. To my eye, the overall color warmth appears to the same (i.e. same average RGB values), even though they different in many other ways.


I brought both grabs into Avid and CC'd just the non-aliased one to make it warmer.

I think it looks more appealing.


I agree that the CC'd image is more appealing, but I think it makes the color very different between the two.

For what it's worth, both images were generated from the same file:

20 MB original PNG conversion (http://thebrownings.name/photo/2009-10-aliasing/2009-01-30-3481-rt.png)

In turn, the PNG was generated from this raw file:

24 MB Original raw file (http://thebrownings.name/photo/2009-10-aliasing/2009-01-30-3481.CR2)

To generate the aliased file, I downsampled the original PNG with a point-sampling algorithm in ImageMagick (a command-line image processing program):

convert -filter Point -resize 400x 2009-01-30-3481-rt.png 2009-01-30-3481-rt-400-point.png

To generate the anti-aliased file, I downsampled the same PNG file with the lanczos algorithm in ImageMagick:

convert -resize 400x 2009-01-30-3481.png 2009-01-30-3481-rt-400-lanczos.png

Neither of those operations introduce a color shift aside from what is caused by aliasing itself. In the case of this image, I don't really see an overall color shift due to the aliasing. (To my eye, anyway.) That's part of the reason why I picked this image for a demonstration. On other images, especially ones with fine, repeating detail (e.g. man-made), the aliasing definitely causes a significant color shift.

Daniel,

I made the CC'd non-aliased image warmer than the aliased image, just to make people's preference for warmth a little more obvious.

That said, looking at the original images in a vector scope confirms that the aliased image is slightly warmer.

A Canon HV20 will have quite a bit higher resolution than a 7D.

There is a difference between resolution and sharpness. Aliasing adds a ton of perceived "sharpness" but it is not resolved detail, it's spurious image contamination that -- in many conditions -- looks "good".

Aliasing never looks good and it doesn't add any "perceived sharpness". Atleast not to my eyes. I have them both and I have shot material with them both. Your awfully tempting me to pull out a couple of test images.

...the HVX200 which had been the standard bearer for "Rebel" digital film makers.


Maybe the new standard bearer for "Rebel" digital film makers is the "Rebel" digital T1i by Canon. ;)


[...] And, afterall, the goal we all have is a watchable film. In that respect, I have to challenge the comparison Dan raises between a bag of candy and a five star meal at a pretentious restaurant..... that may be his preference, but I would wager a lot more out there would opt for a pizza or burger....because of the waste of money associated with the overkill at that five star restaurant. The same would apply to choice of cameras, depending on your need.

The candy metaphor does not apply to the 5D2 as a whole (or any current video DSLR): just individual aspects of image quality in isolation.

Some will shoot the 5D2 with lifelike color, saturation, normal sharpening, and high dynamic range: gourmet meal. Others will configure it for over-the-top neon green grass, extreme sharpening, and ultra contrast: candy. The difference is that the 5D2 gives you a choice when it comes to saturation, sharpening, dynamic range, etc. You can pick anything from candy to pizza, burger, or gourmet if you want.

But when it comes to aliasing and resolution, the DSLR leaves you no choice. It is in full-on candy mode all the time: strong aliasing, low resolution. If you hate candy, then that is a negative. If you like candy, it's a positive. My opinion is that most audiences *think* they like candy (high aliasing, low resolution), but would actually enjoy a gourmet meal more. But in the case of the 5D2 it doesn't matter, since candy is the only option. You can't pick a pizza, burger, or anything else.

I know for my purposes the footage I can achieve with the 7D is going to work just fine for my clients as well as for the viewing public.

Here's my point with the candy metaphor: some people like aliasing (candy), so the aliasing of a DSLR is just another benefit to all types of shots. Others dislike candy, but they still use a video DSLR in spite of it, because the other benefits outweigh the aliasing. The category you belong to might change depending on what type of shot you're doing (e.g. deep DOF vs thin DOF).

I think we must go down to earth

Agreed. We acknowledge that the DSLRs have lower resolution than camcorders that cost an order of magnitude less, but we still use them because they have other benefits. We know that the aliasing artifacts are what make some viewers consider the image "sharp", even though it has very poor resolution, but that's OK too, because sometimes it's an acceptable trade off.

Not with the current chips -- that's a hardware limitation. One company is addressing the chip read speed and will introduce an HDSLR that will have no such compromises; they should announce that the day after tomorrow.

But let's be clear on something: the companies aren't "saddling" anything with limitations! They're offering low-cost products. It is no more fair to say that they're "saddling" the current cameras with limitations, than it would be to say that a laptop manufacturer is "saddling" their netbooks with Atom chips. It's all about price point. Pay $6,000 and you can have a Core i7 chip. Pay $299, and you get the Atom. That's how it works. Better chips cost more. And these cameras' chips are still-camera chips, not designed for video at all. In terms of video performance, these cameras have given us the Atom chip right now (limited to 8fps read speed). If/when they produce the 60fps or even 120fps barnstormer we all want, I guarantee you it won't cost $1699, any more than a screaming Mac Pro Octo-Core is going to cost $599. You got $599, you get a Mac Mini, that's how it works. You want an Octo-Core Mac Pro, you find $5999 in your wallet for that. The cameras will be the same way.

Barry, thanks for the well thought out feedback. When I wrote my other post, I really wasn't sure if there was a hardware limitation with the current CMOS chip itself, or just the software. Read out of CMOS chips will need to become a lot faster to cure this problem.

I am an advanced amateur that has been dabbling in video since the Hi8 days, and photography long before that. Most of my video work is for my own and my family's enjoyment, including vacations to places all over the globe. Some of the video I do I use as part of my job as a college professor, and the 7D/5D MkII, even with their limitations, are close to ideal for what I need in that capacity. I teach in a planetarium, and the 7D or 5D MkII, fitted with a Sigma true 180 degree fisheye lens, can take both stills or video that can be projected right onto our 180 degree dome through our digital planetarium projection system - the distortion introduced by the fisheye automatically corrects when projected onto the dome with our projection system fitted with a fisheye lens. Stills can be taken at intervals at very high resolutions, then combined into a movie, which can be projected on the dome. I have seen examples of others doing this for beautiful timelapse sequences of clouds and sunsets, etc. In that use, the excellent still capability of the 5D and 7D is a big plus, and the video footage is certainly usable.

Here is a real world look at 35mm film, F350 and Canon 5d. The D o P is holding a contest. To experts it should be easy to tell difference.


YouTube - Where's the Canon 5D? (http://www.youtube.com/watch?v=3eBTBnsq88E)

is the 7D 1080 24p camera original resolution correct (like Sony EX1)?

I shot an 2min interview a year ago with my HD100. The client wasn't pleased with their reading so she wanted to shoot again. It was my first shoot on the 7D. She actually wore the same clothes and it was in the same location with the same lights. The difference is night and day. The 7D (with a fast lens) lets you create some beautiful shots.

Hello, Any possibility of some frame comparisons? I have the same two rigs and so far I'm still able to get a better image on the JVC.

My opinion is that most audiences *think* they like candy (high aliasing, low resolution), but would actually enjoy a gourmet meal more.

You have to be really careful about this kind of assumption. While it's certainly true that over time a person may train themselves (or be trained) to appreciate the nuances of a 'gourmet meal' over 'candy' that doesn't mean that they don't actually, truly, prefer the candy right now - and may always prefer it because it's not an important enough part of their life to spend the time necessary to understand and appreciate the 'gourmet'.

I think it's important to keep sight of the fact that those of us on these forums who care about this stuff make up only a tiny fraction of the audience for our work - and to the rest of the audience things like dynamic range, resolution, aliasing, DOF and even frame rate are meaningless. Either the overall picture looks good to them or not - and it's never judged in isolation from the story, acting, sound, mise en scene, etc. If you've got a camera that lets you make images that your audience likes then there are more important things to spend your time on than worrying about whether it's producing real resolution or just aliasing - 99% of your audience just doesn't care as long as the whole production works for them.

Here is a real world look at 35mm film, F350 and Canon 5d. The D o P is holding a contest. To experts it should be easy to tell difference.


YouTube - Where's the Canon 5D? (http://www.youtube.com/watch?v=3eBTBnsq88E)


I think the point may be that it might not be so easy to tell the difference.

His original post:
Where’s the 5D? | Hurlbut Visuals (http://hurlbutvisuals.com/blog/2009/10/25/wheres-the-5d/)

Regarding the candy vs. gourmet choice... the viewer never has a choice. The viewer watches the film/TV show/short/ad as presented. Nobody outside of an edit bay - or web forum - does an A/B comparison.

The options are 1) I liked the way it looked, 2) I didn't like the way it looked, and 3) I was paying attention to the story, so I didn't notice how it looked. As long as they don't go with number 2 (pun intended), it's good enough for most real world situations.

Of course, the client might be more picky. Gotta satisfy whoever is paying the bills.

I think the point may be that it might not be so easy to tell the difference.

His original post:
Where’s the 5D? | Hurlbut Visuals (http://hurlbutvisuals.com/blog/2009/10/25/wheres-the-5d/)

That was my point in posting it. But since there are those that can tell, this is an opportunity to show how they can detect it.

That was my point in posting it. But since there are those that can tell, this is an opportunity to show how they can detect it.

I agree, if anyone knows I would like to see how they know. It looks from the comments on his site that most people are guessing based on jello or aliasing.

If you've got a camera that lets you make images that your audience likes then there are more important things to spend your time on than worrying about whether it's producing real resolution or just aliasing


Yes. It's always good to have a reminder that resolution is just one small part of image quality, and image quality is just one small part of the production, and any given production is just one small part of a lifetime of productions.

You have to be really careful about this kind of assumption. While it's certainly true that over time a person may train themselves (or be trained) to appreciate the nuances of a 'gourmet meal' over 'candy' that doesn't mean that they don't actually, truly, prefer the candy right now - and may always prefer it because it's not an important enough part of their life to spend the time necessary to understand and appreciate the 'gourmet'.


I'm not sure if you understood my position. I don't make any assumptions about training, learning, or eventual appreciation over time. I'm talking about untrained, unlearned, and immediate appreciation. My idea is that there is a disconnect between the specific sensory inputs that they think they like on the surface level and the truly important, deep-down appreciation that they feel of a moving image as a whole.

Say you have two screenings. The candy screening is low res, bad aliasing, oversaturated, oversharpened, overcontrasty (blown whites & clipped blacks), 24 converted to 120hz, plastic noise reduction, and heavily compressed (low DR) audio. The gourmet screening has high res, no aliasing, normal saturation, normal sharpening, normal contrast (detail in whites and blacks), 24p left alone, normal amount of noise/grain, and high DR audio.

My opinion is that most people will say that they like the second (gourmet) screening a lot more. If they put it into words they might call it more realistic, immersive, beautiful, film-like, etc. The first screening they'll just not enjoy as much. If they try to say why, they might say it looked fake, didn't draw them in, unspectacular, and looks like their own home videos.

If you try to get specific with them, and do A/B comparisons asking "which one is sharper?" they'll still pick the aliased and oversharpened image. For color, they'll still say that oversaturated and overcontrasty one has better color. For sound, they'll pick the loudest low-DR one. And so on. They may rationalize the difference by saying that the gourmet movie was better in *spite* of having less aliasing, or despite normal sharpening; they may even suggest that the gourmet would have been better if you made it more like the first (candy) screening; not realizing those were the very factors that made them love the gourmet so much.

In other words, I think they prefer the gourmet screening but they don't know why. They think they like aliasing, and they say that they like aliasing, but if you actually give it to them they will like it less on levels that are more important and lasting.

My position is that there is a big disconnect between the important feelings about a motion image (immersive, beautiful, quality) and what the viewer *thinks* is the way to get that feeling (aliasing, sharpening, etc.). I think it's our job to ignore what they are asking for in order to give them what they really want. Aliasing tickles the most obvious outer layers of sensory input, but it doesn't go down deep to really satisfy.

What they really want is the best dining experience possible. They think the right way to get it is a bag of corn syrup. But if you actually give them what they say they want, they will dislike it very much. They wont say that the meal wasn't sweet enough, but they'll find other words for it, like "low production values", "lackluster", "like my own homemade meals". If you give them a gourmet meal, they will love it, but they wont know why.

The same thing applies to other areas of life. Say your car is leaking water into the carpet. You think it's the heater core. You drop it off at the mechanic and say it needs a new heater core ($1,000), without saying why. Before he starts, the mechanic discovers the water leak and finds that it's caused by a plugged hose ($1). He can probably guess that the real reason you dropped it off was because of the leak, not the heater core. You only think you want the heater core, when in reality you want the leak fixed. Now, he can give you what you're asking for, which is a new heater core, but that wont fix your real need. Or, he can give you what you really want, which is to stop the leak. Viewers want a good image. They ask for aliasing, oversharpening, 24p to 120 hz, etc., because on the surface of it they think that's what gives them a good image. I think there is a disconnect. The difference, of course, is that a leak is an obvious, objective fact that can be proven; whereas the true appreciation of an image is subjective, and my opinion that such a disconnect exists in most viewers is just that: an opinion.

Yes, exactly.
They are capable of spectacular resolution. I shot a side-by-side of the 7D in movie mode, vs. the 7D shooting the same scene in motor-drive stills mode (6fps) and then resized the stills in PhotoShop to 1920x1080. It was shockingly better...


On the subject of using the still image option of DSLR cameras rather than the video option, it is worth taking a look at Stop-Motion techniques to make videos rather than using the video mode. Here Andrew Kornylak uses the Nikon D3 (NOT the D3s) for making videos by shooting still images on the camera's fastest motordrive setting. See D3s thread:

http://www.dvinfo.net/forum/photo-hd-video-d-slr-others/465696-say-hello-new-nikon-d3s-720p.html

Andrew mentions about his work using the Nikon D3 on his blog "...It is pretty close to a video look, but a little different. I can shoot with strobes (like in the Nikki Blue commercial), and the color depth of even a cropped jpeg is better than the best HD video frame. I can make 11×14 prints from a frame grab, which sweetens the pot for a client looking to do a still and motion shoot without an outrageous production..."

Some more Stop-Motion video using the D2x and D3 can be found here:

Stillmotion - Motion projects shot with a Nikon D3 (and D2X) on Vimeo (http://vimeo.com/album/118964)

I'm not sure if you understood my position. I don't make any assumptions about training, learning, or eventual appreciation over time. I'm talking about untrained, unlearned, and immediate appreciation. My idea is that there is a disconnect between the specific sensory inputs that they think they like on the surface level and the truly important, deep-down appreciation that they feel of a moving image as a whole.

snipped



You may be over thinking this. Sometimes people just like what they like. I like Candy and burgers. Chicken Kiev doesn't rock my world.

Aliasing never looks good and it doesn't add any "perceived sharpness". Atleast not to my eyes.
So then the simple question I put to you is: does the 7D look sharper than a 500-line video camera?

Or, let me put it another way -- how sharp does this wedge of a resolution chart look to you?
http://dvxuser.com/barry/7D-1080-Mid-Trumpets-sideways-lines.jpg

I have them both and I have shot material with them both. Your awfully tempting me to pull out a couple of test images.
Sure, whatever you'd like. Just pull out images that show actual resolved detail, since that's what this thread is about. Shoot charts, or shoot a newspaper at various text sizes, and let's see which one makes the text easier to read. I'd be willing to bet that if the test is conducted accurately, you'll find that the text is far more legible on an EX1 or HPX300 or even an HV20, than it is off of *any* HDSLR. Because those video cameras resolve more detail.

Don't just go shooting some random shot of trees or whatever, because that's where the aliasing gets to work its "magic". Shoot something where you actually want to be able to resolve detail, and you'll find the 1080p video cameras are better at it than the DSLRs are, hands down, no questions asked.

the 7d has footage that looks awesome and i can afford it.


case closed.

So then the simple question I put to you is: does the 7D look sharper than a 500-line video camera?

Or, let me put it another way -- how sharp does this wedge of a resolution chart look to you?
http://dvxuser.com/barry/7D-1080-Mid-Trumpets-sideways-lines.jpg


Sure, whatever you'd like. Just pull out images that show actual resolved detail, since that's what this thread is about. Shoot charts, or shoot a newspaper at various text sizes, and let's see which one makes the text easier to read. I'd be willing to bet that if the test is conducted accurately, you'll find that the text is far more legible on an EX1 or HPX300 or even an HV20, than it is off of *any* HDSLR. Because those video cameras resolve more detail.

Don't just go shooting some random shot of trees or whatever, because that's where the aliasing gets to work its "magic". Shoot something where you actually want to be able to resolve detail, and you'll find the 1080p video cameras are better at it than the DSLRs are, hands down, no questions asked.

Barry:

But what if you need the depth of field control we get with the 7D and 5D. What is better then-- Those same video cameras with adapters, or the 7D/5D choice? That is the real question. I am interessted to hear what you have found regarding resolving power in that situation.

I'm not sure if you understood my position. I don't make any assumptions about training, learning, or eventual appreciation over time. I'm talking about untrained, unlearned, and immediate appreciation. My idea is that there is a disconnect between the specific sensory inputs that they think they like on the surface level and the truly important, deep-down appreciation that they feel of a moving image as a whole.

No, I do understand your position, and I still think that your assumption is incorrect.... based primarily on the fact that I used to assume the same thing. I used to 'fix' peoples TVs for them by turning down sharpness, saturation, contrast, etc to get a more natural, organic looking image - more often than not they would complain that the image looked worse or I'd find it right back where it was a week later. Or I'd 'fix' the aspect ratio and they'd complain about the image not being full-screen.

I finally realized that they really did prefer it 'wrong' and that showing them something 'better' didn't make them change their minds - I was assuming that they would feel the same way I did if I just showed them the difference. This is a projection bias that I think is common to most people who do spend the time to learn and appreciate the finer nuances of any particular subject.