Too many megapixels?

When the 10D came out with a whopping 6MP, I couldn't figure why canon would come to market with a new dSLR with no improvement in resolution. Now that I've got several weeks under my belt on the D100, an interesting concept comes to mind. The converted raw files from the D100 weigh in at a whopping 34 MB. Even my blazing 1GHZ powerbook is challenged when it comes time to crunch these files. I printed seahorses yesterday to 13X19 prints that were amazingly good. Why would I want higher resolution? My digital workflow would be further stifled crunching 60, 80, 100MB files. I think that buffers and processors may be the new bottleneck, not the CCD/CMOS sensor. I haven't had the opportunity to work on files from the new full framers, but I suspect it must get awfully tedious. In fact, I wonder if shooters of these cameras actually utilize the highest quality settings on these cameras.

Just a rant meant to spur discussion.

Ain't no such thing as too many megapixels

Just out of interest, what do you mean by "Even my blazing 1GHZ powerbook is challenged when it comes time to crunch these files." Which particular crunches are you referring too. I just timed mine (1GHz, 1 gig RAM) doing an unsharp mask on a 27 Mb file, it took 2.5 seconds, which seems pretty snappy . Printing, though, is another issue.

Anyway, we mac users are hamstrung by a poor, slow speed system bus. Your powerbook bus speed is about 133Mhz, which is quite a bit slower than the current P IV systems (they are about 300-500Mhz I believe, but may well be wrong). So your processor is spending way too much time starved for data to work on. Hopefully this will be addressed in the near future if Apple starts using a new PowerPC chip that IBM has recently announced.

When the 10D came out with a whopping 6MP, I couldn't figure why canon would come to market with a new dSLR with no improvement in resolution.  Now that I've got several weeks under my belt on the D100, an interesting concept comes to mind.  The converted raw files from the D100 weigh in at a whopping 34 MB.  Even my blazing 1GHZ powerbook is challenged when it comes time to crunch these files.  I printed seahorses yesterday to 13X19 prints that were amazingly good.  Why would I want higher resolution?  My digital workflow would be further stifled crunching 60, 80, 100MB files.  I think that buffers and processors may be the new bottleneck, not the CCD/CMOS sensor.  I haven't had the opportunity to work on files from the new full framers, but I suspect it must get awfully tedious.  In fact, I wonder if shooters of these cameras actually utilize the highest quality settings on these cameras.  

 

Just a rant meant to spur discussion.

Sales, sales, sales and more sales. For publication it is very important.

Megapixels aside (14mp is probably enough, maybe 22mp to 'match film' or so they say), dynamic range is the next thing the digital cameras have to conquer. I still like my film camera for shooting sunballs on wrecks and pelagics. Most of us I think would be happy if they manage to do this.

If you don't get the higher resolution, I can't make the image larger (i.e. above 16x20) and maintain the quality. I got a request for a 10'x6' image just recently. You'll need every pixel you an get for that image.

In addition, most shots ARE cropped. Some significantly. That means that you are looking at less megapixels again for a final image. So this helps - if they want to change the vertical to a horizontal, I lose an awful lot of quality when I then blow it up to a reasonable size.

Right now, I would be happy to see a camera with another 3-4 stops of dynamic range, than to see one with a higher megapixel range.

I absolutely use the highest settings on my camera (a RAW S2Pro file is 12mp, the tiff is over 70mp). I spend some time making them suitable for viewing on my site/catalog. For those that I am printing I spend a lot of time on.

Lastly, I process all images on a 2.4ghz machine dedicated to image processing. It cranks away all day long most days. It handles the images at a very reasonable speed. Certainly a lot faster than scanning my slides and then having to correct for the scanner (and still with all the processes to go).

So I guess it comes down more to what are you planning to do with your images?

M

Kasey, you bring up two good discussion issues here. One, being megapixels and the other, processing speed.

First, megapixels...

Criag is absolutely right. You can never have too many pixels. If you want to crop an image, you'll need higher resolution. If you want to really enlarge an image, the bigger, the better. I recently sold an image to a guy in Minnesota that would be printed at 12' x 6' and used as a wallpaper for his jacuzzi room. Even from a 35mm slide scan, I had a hard time getting enought resolution to accommodate this. You can always size down, but there's no real good way to size up.

Second, processing speed...

Make sure you have as much memory allocated to Photoshop as you possibly can. Also make sure you have tons of space on your scratch discs and that you have the Photoshop preferences set up so they are using that scratch disc space. I just picked up an 80 Gb hard disk for my G4 for $100 that I use only for scratch disc space. It makes all the difference in the world.

For Nikon users the Nikon Capture software seems to process images more quicky in RAW format than when using Photoshop. I have a friend doing extensive testing on this if anyone would like to know more about this.

Hope this helps at least a little!

Bonnie

You may have too many Megapixels if you are only going to e-mail them or post them on the web. But when I finally get that Nat. Geo cover shot I want it to be as many pixels as possible. I know that next shot is the one so I always try to be ready for it. And as mentioned above, you can never have too much RAM in your computer.

Woohoo, this is a great thread Kasey!

I recently wrote an article on Digital Noise, so this topic is near and dear to my heart.

You CAN have too many pixels, if they are packed into a small sensor. For example, many people who bought the Canon S__ series cameras (S30, S40, S45) have realized that the 3 megapixel camera takes better pictures than the 4 megapixel camera! That's due to the fact that the size of the sensor didn't change, so the pixels got smaller in order to pack more of them onto the sensor. This results in noisier photos... Not only that, but diffraction starts to come into play as the photosites get really small.

A 3 megapixel camera with an APS sized sensor (like the Canon D30 or a Fuji S1) will kick the stuffing out of a 5 megapixel camera with a consumer sized sensor.

Now on to the other stuff.

RAM is cheap. Get as much as you can. RAM for the PC is especially cheap right now. I'm talking "everyone should have over a gig" cheap. Same goes for hard drives. By the time I fill mine, one twice as large costs 1/2 as much...

On to Photoshop. Bonnie is absolutely correct. You need a big scratch disk and one that is not your "C Drive" Since you have firewire, you can attach a firewire hard drive to your MAC if you can't fit another hard drive inside. That will help a LOT.

Cheers

James

Several good comments already:

1) Per James, you can have too many pixels, all things being equal.

2) Limiting pixels based on file size/processing speed is a self defeating. Such limits can be defeated now and later. One shouldn't limit oneself based on hardware limitations. One of my earlier job assignments was being locked in a vault in a corner of some unknown space between the 1st and 2nd floors of corporate headquaters. The room housed a mini system strictly for the use of the President and tied to the board of directors meeting room. This pre-dated PCs. On this mini system was a 3 dimensional spreadsheet program that I was working on. The system was so slow that if I hit the recalc button I was done for the day. I could only save the program at the end of the day as this was a 6 hour process.

3) How many pixels is enough? Less than most people think. Your computer screen doesn't care that much. Approximately 95% of all prints are 8x10 or smaller. You don't need 20 megapixels to print an 8x10. People have talked for years about how many megapixels it takes to equal film. The proof is in the final output and not a theoretical maximum resolution. More pixels on the same size sensor just costs more.

4) The 10D corrected shortcomings of the D60 which had nothing to do with resolution. 6mp is proving to be plenty. Many of the images from Iraq are from 3mp D1H cameras.

5) Newer models will continue to be evolutionary, rather than revolutionary. Improvements will continue to address areas and weaknesses that have nothing to do with resolution. Paul mentioned dynamic range, although my D100 has much more dynamic range than shooting with Velvia. Other areas of improvement will include moire, color fringing, chromatic aberation, noise, blah, blah, blah, yada, yada, yada.

6) Printing large with existing megapixels has gotten much better. There's plenty of software out there to help the "upsize" process. Professional printers RIP software will continue to get better, reducing the need for mega megapixels.

This is a really interesting thread. Comments on RAM and scratch discs are useful. How about write times and buffer size in the 14N and 1Ds? Can they keep up with a D100 at its maximum resolution? I would think that the final files are a bear to post-process.

I agree with scorpio as far as the post-processing improvements. I've never understood why you'd want to interpolate within the camera like the S2 - but I understand that it does a better job than most software. I've never used GF, is it as effective as others claim?

The slowest processes seem to be opening files and converting in batch to web size. I didn't mean literally that more MP wouldn't be an improvement, just whether it would bring with it new and frustrating limits.

You may have too many Megapixels if you are only going to e-mail them or post them on the web. But when I finally get that Nat. Geo cover shot I want it to be as many pixels as possible. I know that next shot is the one so I always try to be ready for it. And as mentioned above, you can never have too much RAM in your computer.

I don't think getting more MP matters for your nat geo cover shot unless its going to be a heavy crop - take a real close look at the printing quality in terms of resolution of fine features and edges of any magazine or even an expensive coffee table photo book and note how crappy it is compared to what you can output on a decent ink jet photo printer.

amen on the ram - for processing photos get more ram before you get the computer upgrade - for photoshop imho any reasonable clockspeed is mostly ok as long as you've got

768-1gb of ram, but even a dual-cpu 3ghz system with only 256mb of ram would suck bad. the computer I do most of my photoshop on is a dual 400 (!!) mhz system with 1gb of ram and it is usually fine. also a big high quality moniotr helps a lot - if I didn't already have a nice 20" monitor on this system, I'd upgrade the monitor before the cpu.

I don't think getting more MP matters for your nat geo cover shot unless its going to be a heavy crop - take a real close look at the printing quality in terms of resolution of fine features and edges of any magazine or even an expensive coffee table photo book and note how crappy it is compared to what you can output on a decent ink jet photo printer.

Actually you need even higher resolution for anything that is going to print on a printing press. This is due to the screening process necessary to color separate and print color photos. Even in spite of this, the final printed piece will not look as crisp as a continuous tone print or photograph that has no "dots", also in part because of the color gamut of the CMYK process. For the photographer, this stinks because its very difficult to reproduce an image in its full glory, but for now its all we've got.

Desktop and digital photo printers use an entirely different technology to reproduce the image and can get much higher quality because they don't have these limitations.

In the large format and print reproduction world, a 14-18 Mb file coming off a 5-6 MP camera is tiny. So until I've got a 20 MP camera, I won't be happy. Most professional stock agencies won't even accept anything less than 50 Mb. When you start working on that file in Photoshop and add a couple of layers, factor in what's cached for the undo's and try to write alternate versions, you quickly find yourself with a file that requires a couple hundred Mb just to process.

If all you're going to do is display your image on a screen or email it, then you don't need the larger file size, but I don't know why anyone would want to limit themself. When you've got an opportunity to get published someplace, anyplace, its a real disappointment to be turned down because all you muster up is a couple of Mb. Programs like Genuine Fractals do a pretty good job of upsizing, but the results are far from perfect and are not acceptable at all with most stock agencies.

Sorry for getting "techy" but I used to teach this boring stuff when I worked for Kodak in another life.

A Raw file from the S2 generates a 4,500 x 3,000 16bit TIFF in whatever color space you choose. That's pretty darned big Bonnie. 42 Megs in fact.

That's within "significant error" of your stock agency's magical "50 megabyte" number.

Cheers

James

That's within "significant error" of your stock agency's magical "50 megabyte" number.

I wish my stock agency would agree to that. The standards they set are about more than just Mb. They won't accept anything that's been compressed, upsized, interpolated or clipped (dynamic range) in any way. It took me three frustrating attempts to get them to accept my scans from slides even when following their specs.

I also produce a lot of brochures for dive travel & manufacturers and often get discs from photographers with images that cannot be used because they are too small. Not everyone is shooting with an S2. Its too bad because there's a lot of good photography out there, but if the pixels aren't there, the uses are limited.

I don't think getting more MP matters for your nat geo cover shot unless its going to be a heavy crop - take a real close look at the printing quality in terms of resolution of fine features and edges of any magazine or even an expensive coffee table photo book and note how crappy it is compared to what you can output on a decent ink jet photo printer.

 

Actually you need even higher resolution for anything that is going to print on a printing press. This is due to the screening process necessary to color separate and print color photos. Even in spite of this, the final printed piece will not look as crisp as a continuous tone print or photograph that has no "dots", also in part because of the color gamut of the CMYK process. For the photographer, this stinks because its very difficult to reproduce an image in its full glory, but for now its all we've got.

 

Desktop and digital photo printers use an entirely different technology to reproduce the image and can get much higher quality because they don't have these limitations.

 

In the large format and print reproduction world, a 14-18 Mb file coming off a 5-6 MP camera is tiny. So until I've got a 20 MP camera, I won't be happy. Most professional stock agencies won't even accept anything less than 50 Mb. When you start working on that file in Photoshop and add a couple of layers, factor in what's cached for the undo's and try to write alternate versions, you quickly find yourself with a file that requires a couple hundred Mb just to process.

 

If all you're going to do is display your image on a screen or email it, then you don't need the larger file size, but I don't know why anyone would want to limit themself. When you've got an opportunity to get published someplace, anyplace, its a real disappointment to be turned down because all you muster up is a couple of Mb. Programs like Genuine Fractals do a pretty good job of upsizing, but the results are far from perfect and are not acceptable at all with most stock agencies.

 

Sorry for getting "techy" but I used to teach this boring stuff when I worked for Kodak in another life.

Can you explain in more detail? Maybe its an artifact of the process, but usually the lower the quality of the sampling in the final image, the poorer source material you can get away with. Maybe its an artifact of the process used to make the plates? I'm a techy, but don't know much about how images are prepared for printing presses. I can see from all conventionally printed materials handy that the sampling "foot print" is pretty large.

Hi Bonnie,

I spoke recently with Joyce and Frank Burek about this. They were recently back from a big meeting/conference in New Mexico with a bunch of stock agencies.

They are working on "going all digital" and have had many discussions w/ agents about this. I'll see what their standards are.

They were recently surprised to receive ALL of their stock slides in the mail. It turns out their agency had had them all scanned and sent the slides back! It would be funny if they didn't accept digital files after that...lol. I'm not sure how they scanned them, but I'm guessing the output was probably around that "magic 50 megabyte" size.

Cheers

James

Can you explain in more detail? Maybe its an artifact of the process, but usually the lower the quality of the sampling in the final image, the poorer source material you can get away with. Maybe its an artifact of the process used to make the plates? I'm a techy, but don't know much about how images are prepared for printing presses. I can see from all conventionally printed materials handy that the sampling "foot print" is pretty large.

Hi chrisg,

I'll dig aroung to see if I can find something online that explains this in better detail and will get back to you.

OK, here are a few good links:

http://marvin.mrtoads.com/index.html

http://entertainment.howstuffworks.com/off...t-printing3.htm

To summarize, a photograph must be broken up into different sized dots (screened) so it can be output to film and plates and get printed on a press. You'll see an example of this screening if you use a magnifying glass on any printed photograph. To make these dots, an image must be twice the resolution of the line screen. This is true with both B&W and color. To make things even worse, every color you see printed off a press is done with various combinations of cyan, magenta, yellow and black ink. No red, no blue, no green, just CMYK*. So getting a nice turquoise background that fades to deep azure is a read trick when using only CMYK.

*With the exception of Hi-Fi color, which I know someone would call me on if I didn't mention.

What it boils down to, is that the printing process hasn't changed significantly in 100 years. Today there are digital presses, but they still use dots and CMYK and many of the traditional theories. Until this changes, we are going to suffer from some loss of quality in our beautiful underwater images.

Hope this helps.

To make sure I wasn't putting wrong info out there, I did a little follow-up on this too and talked to my editor at Corbis Images to see if they had anything new on this. They have set new standards, but unfortunately a D100 or S2 barely meets the criteria. Rules differ for editorial vs commercial, with commercial being more rigid. I suspect if I were a really hot photographer they might make some exceptions to these rules, but...

Although I can cough up enough Mb if I keep the file in RAW format at 12 bits per pixel, the minute I convert it to a standard 8 bit TIFF, I'm back to 17.2 Mb. For editorial submissions this might be OK (but would not allow any cropping), but for commercial submissions would not be accepted.

Here are the requirements (edited some for simplicity)

COMMERCIAL: Technical File Specifications

Scanned File Size Requirements

o The Minimum Commercial Image file size is 50MB, RGB Tiff in Adobe RGB (1998)-Color Space (16.7MB gray scale Color Space "Dot gain 20%".) Maximum file size is 80MB

Digital Camera File Size Requirements

o The Minimum Commercial Digital Image file size is 33MB, or higher, RGB Tiff in Adobe RGB (1998) -Color Space (11MB gray scale Color Space "Dot gain 20%") Maximum file size is 80MB

File format requirements:

o Tiff in Adobe RGB (1998) Color Space

o Raw digital camera files in Adobe 1998 RGB Color Space

EDITORIAL: Technical File Specifications

Scanned File Size Requirements

o The Minimum Editorial Image file size is 33MB, RGB Tiff in Adobe RGB (1998)-Color Space (11MB gray scale Color Space "Dot gain 20%".) Maximum file size is 80MB

Digital Camera File Size Requirements

o The Minimum Editorial Digital Image file size is 17MB, or higher, RGB Tiff in Adobe RGB (1998) -Color Space (5.6MB gray scale Color Space "Dot gain 20%") Maximum file size is 80MB

File format requirements:

o Tiff in Adobe RGB (1998) Color Space

o Raw digital camera files in Adobe 1998 RGB Color Space

They also have a whole list of image criteria but I'm not sure if I should be publicly distributing this much info so I'll leave it at this.

I have found a work-around to the restrictions of my stock agencies, that is to start one of my own. There's more than one way to filet a fish.

If you read Jim Watt's webpage, he claims that the images captured by his D60 is a higher quality image than that captured by a drum scanner despite the smaller file size. A higher fidelity file should interpolate better for enlargements. Time will tell, but I suspect it won't be long now until the agencies get a clue.

Is the difference between the 17MB 8 bit file and the 33MB 12 bit file actually visible or significant? It seems like using file size as a judge of image resolution and depth is a little pointless.

If you read Jim Watt's webpage, he claims that the images captured by his D60 is a higher quality image than that captured by a drum scanner despite the smaller file size. A higher fidelity file should interpolate better for enlargements. Time will tell, but I suspect it won't be long now until the agencies get a clue. 

That depends on a lot of things. I've seen some pretty bad drum scans. Comparing apples to apples, this is probably right.

Is the difference between the 17MB 8 bit file and the 33MB 12 bit file actually visible or significant? It seems like using file size as a judge of image resolution and depth is a little pointless.

The difference is in bits per pixel, not in actual resolution. If you were to compare two images side by side and look carefully at things like shadow detail and tone transition, you might see the difference. I honestly don't belive most people would see any difference at all unless it were pointed out to them. You are absolutely right about the file size being pointless. There are many other factors that determine image quality.

Bonnie,

Thanks for posting those submission guidelines.

I was pleased to see that the digital camera file requirements are reasonable, and more importantly, they are less than those for a scanned image. That makes sense to me.

I was also pleased to see that a Tiff converted from the Fuji S2 RAW will meet their requirements. (they are 42 megs)

Cheers

James

It seems like using file size as a judge of image resolution and depth is a little pointless.

Somewhat arbitrary, but not pointless. It prevents woefully inadequate submissions. The actual file size parameter goes back to another "old school" rule of thumb about file sizes and printing.

As a matter of fact, some well known photogs have printed their paltry digital files, then scanned the prints at a fairly high resolution in order to meet file size requirements.

If you read Jim Watt's webpage, he claims that the images captured by his D60 is a higher quality image than that captured by a drum scanner despite the smaller file size. A higher fidelity file should interpolate better for enlargements.

This is by no means a consensus opinion. A properly executed drum scan will yield results that will outperform the interpolated digital file at larger sizes. Defining "larger sizes" has been a matter of much debate, but usually goes beyond normal printing sizes of up to 11x17.