→ May 2007 Contents → Column
Warning: Ethical Problems Ahead (Perhaps)
First of all can I say that this column is not, repeat not, meant as an attack on anyone but as a warning and, perhaps, something to think about.
There are limits to everything.
The limits of an HDV (1080i) frame grab (from a Sony Z1 or a Canon XL-H1) are 1920 pixels by 1080 pixels because that's the definition of a 1080i HDV frame. Camera makers design their chips that way. It's one of the HDV standards.
That's a 5.93Mb (uncompressed, RGB, 8-bit) TIFF file.
That's all the technology can manage at the moment.
If you take that 5.93Mb file (hereinafter known as "the file") and print it at 72 dots-per-inch (hereinafter known as DPI) you'll get a 26.667-inch by 15-inch print that will look good from far away and look like s**t close up.
If you take that file and print it at 300 DPI you'll get a 6.4-inch by 3.6-inch print that will look good close up but be difficult to see from a distance.
If you print it at the newspaper norm (200DPI) you'll get a 9.6-inch by 5.4-inch front-page photo.
That's the trade-off.
There is a certain defined, and limited, amount of information in an HDV video frame. It isn't "grainless." It isn't magic. You can spread that information over a big area or a small area, depending on your needs and requirements. It's kind of like a 35mm negative.
Remember the 35mm B&W negative?
It too had a limited amount of information stored within it. If you put it into an enlarger and projected the image on to an 8x10-inch piece of paper you got a print.
If you projected it on to a 16x20-inch piece of paper you got a bigger print, with more noticeable "grain."
If you projected it on to a 4x6-foot piece of paper (I've seen it done) you'd get a BIG print that would look great from 20 feet but look like crap from a foot and a half.
If you used a "slow" or "fine-grain" film you could make a bigger print before it looked like s**t. You could always make the deal and trade extra sensitivity ("higher speed") for bigger grain.
When you made an "enlargement" you didn't create more information than was available in the negative, you just spread it over a bigger or smaller piece of paper.
You might even think that that negative is quite digital as the information is contained in small pieces of silver suspended in a hardened gelatinous mass attached to a piece of plastic. Black bits and not-black bits. On and off. Almost like pixels.
The current, and very necessary, attempt by photographers and photo organizations to make the transition from still to video (I am one of them) has brought some confusion to the fore.
You can use Photoshop to increase the size of a digital image file. It's easy. It's the default in Photoshop's "Image Size" selection.
What you are doing when you ask Photoshop, or any of the other imaging programs or those Photoshop plug-ins, to do when you "re-size" an image is to make stuff up. You're asking the computer to create information "out of thin air" using algorithms and calculations determined by a programmer.
Some algorithms (mathematical formulas) work better than others and some work better for some things than for others. Bicubic works differently from Bicubic Sharper or Bicubic Smoother or Nearest Neighbor and Genuine Fractuals takes a different direction altogether.
But what they all do is make stuff up.
If you increase your HDV frame grab from 5.93Mb to 40Mb you have told your computer to make up 35Mb of information. Where did the 35Mb of information come from? The computer made it up. That's the fact of the matter and some day it's going to come back and bite someone in the ass.
In a world where an added-stubble-OJ cover causes concern and an added basketball or a couple of missing legs gets a photographer fired we should all be forewarned.
Some day. Somewhere. Someone is going to look at that picture splashed big across the front page of a newspaper and, knowing that it's a frame grab from an HDV camera, will say, "Where did the extra data come from?" "It's not what the camera recorded!" "How can these people (the newspaper) lie to us (the reader) like that?"
That's when the tempest will start and the tears will follow.
I was very interested in the release of the "Voodoo Tool" from the Dallas Morning News but it leaves a lot to be desired.
For one thing it's dependent of the size of your monitor so that the screen grab you can make form a 23-inch monitor is better than one from a 19-inch one. If you're working with a 15-inch Apple laptop you're screwed.
Secondly, it takes the screen shot and uses Photoshop's Bicubic algorithm to increase the file size to over 60mb, an increase of more than 300 percent. My reading of Photoshop books and expert Web sites tells me that the general thought is that you shouldn't generally use Bicubic to increase an image by more than 100 percent (some say 200 percent max.)
Thirdly, the Voodoo Tools droplet (a series of Photoshop actions) then compresses the expanded file and saves it as a JPEG file (JPEG being a "lossy" format in which information is discarded in order to compress the file).
So you're adding information using Photoshop Bicubic and then throwing away some of that information to save it as a JPEG file. Then, presumably, you open that JPEG file and re-size it for your publication and then save it, again, as another JPEG file, throwing away still more information.
There are better ways to do it.
Download a program called MPEG Streamclip from www.squared5.com. You can't beat the price (it's free) and it does some amazing things if you want to convert one video stream format to another.
But the best part (in this context) is that it will make a full-resolution (1920 x 1080 pixels) frame grab from your 1080i HDV Quicktime file (or any one of a number of different files) and save it as an uncompressed TIFF file to your computer.
Once you have that TIFF file you can open it in Photoshop and, if you want, enlarge it or reduce it by just the (small) amount you need. You can use the "step" method (making multiple 10 percent file enlargements) of Bicubic enlargement or a plug-in like Genuine Fractals for better results. If you're really strange you can use the "step" method with Genuine Fractals...
It's not as easy as Voodoo Tools but it isn't monitor-dependent. You make the choice.
© James Colburn
Back to May 2007 Contents