David Pogue's Resolution Test

[livin4lax09]

http://pogue.blogs.nytimes.com/2006/...ogues-posts-2/

very interesting read indeed...this is why I laugh when people scoff at my *measly* 4.2 MP 1D.

[walterick]

Great article! I hope it redirects some budding photographers away from the tech and back in towards the heart, where the real source of a good image lies.

btw, anyone trying to access the link, it took me 2 tries to get it to access the blog. I got an error from NYT the first time.

Rick

[masdog]

Excellent link, Brent. This makes me want to keep my current cameras for now.

[gahspidy]

His test is flawed, or misleading actually. He says he took one image and resized it to the various resolutions. If the image were not of such great quality to begin with, then the lower resolutions would not stand out from the rest anyway. Also, if the image were top notch quality, then it would most likely hold up at that size in the lower resolutions. A true test for the person who asked him how a small resolution camera could make such large prints as claimed, would be to use actual cameras of different MP and shoot the same image at same iso with tripod. Then try blowing each one up to the test size, with no interpolation, and see what happens. I think anyone would be able to see the difference. However, a low resolution camera with a good lens and very clean image could probably interpolate nicely up to a much larger image.

[Sebastian]

His test is flawed, or misleading actually. He says he took one image and resized it to the various resolutions. If the image were not of such great quality to begin with, then the lower resolutions would not stand out from the rest anyway. Also, if the image were top notch quality, then it would most likely hold up at that size in the lower resolutions. A true test for the person who asked him how a small resolution camera could make such large prints as claimed, would be to use actual cameras of different MP and shoot the same image at same iso with tripod. Then try blowing each one up to the test size, with no interpolation, and see what happens. I think anyone would be able to see the difference. However, a low resolution camera with a good lens and very clean image could probably interpolate nicely up to a much larger image.

Gary,

Actually I think your idea is flawed. There is simply no legitimate reason to EVER print an image without interpolation. Wether resizing up or down, interpolation needs to be used to keep the image from totally falling apart. That's why in Photoshop you don't even get an option to turn it off.

God it feels nice to actually have the time to post something...

[gahspidy]

Sebastian,
Correct me if I'm wrong, and I know you will
Interpolation is done by resampling. I have the option in CS2 to turn off resampling if I choose, thus not having my image interpolated when going up or down. My files are roughly 8x12" at 324dpi out of the camera. If I wanted to print out 4x6 prints, I would size the dimensions to 4X6 without interpolating, which would bring my resolution up to 648 dpi.
That high resolution is not going to look any better, but does not hurt. File not affected or broken down whatsoever. If I wanted to go higher print size while keeping the 328 resolution, then I would have to interpolate by keeping Resampling on.

As for the test that was done, I'm saying that by taking a given image at 16x24 and interpolating it's resolution to match lower ones from lower MP cameras does not seem to me to be the same thing as taking an image from an actual lower MP camera and interpolating it's image up to 16x24 at a given resolution. Alot depends on the quality of the lens, image sensor, and also the interpolating software. In my initial comment i did say blow them up without interpolation and that would be flawed because it surely must be interpolated to achieve that size with any chance of looking good.

And after all is said and done, I too feel that more emphasis need be on technique and vision than gear and software.

[Mr Yuck]

my 3mp A510 printed a perfect 11x14 and my brother's 5mp S2IS printed a perfect 16x20

(By printed, I mean the cameras took the pictures uploaded to shutterfly)

I dont know if I would trust my 3mp camera beyond 11x14 and higher resolution cameras certainly give more cropping ability, but the average consumer wont be able to tell the difference between a 3 megapixel and an 8 megapixel. (I didnt notice the flaws in my 2mp A40 until I upgraded to a 19" Wide LCD over my 17" CRT.)

[Ronnoco]

I agree with Gary on this one and so does microsoft.com in the article on their web site: Printing Quality Photos.

Ronnoco

[Sebastian]

Gary,

Not trying to pick on you, honest.

BUT, you are incorrect in thinking that there is no interpolation done. Printer drivers arbitrarily interpolate at the time of printing, your 6XX dpi images become whatever the printer needs them to be at the time you print, depending on the printer and its settings. The Olympus P-400 for example prints out very poor images unless you specifically set the correct pixel dimension for the desired output at exactly 314 DPI, the native resolution of the print head. Your lack of using interpolation may in fact be hurting your output, depending on way too many factors than just an option in the Image Size panel of Photoshop.

Interpolation is a good thing, never a bad thing. It helps keep images looking good when downsizing and upsizing. Poor downsizing algorithms look terrible, one only has to look at the fast averaging Photoshop itself does when images are shown on screen at non-even percentage values. Compare an image shown at 31% to an image shown at 25% to see the obvious differences in artifacting. A fast averaging method is used to keep the workflow interactive, and much more computationally intensive methods are used to resize the image when resampling is on, providing much higher quality at a speed penalty.

It's always best to resize for the desired print size at the desired DPI and sharpen for that particular output method. That's why people using software like Nik Sharpener get such excellent results, you tell the software the desired output method and size and it does the footwork for you. The excellent Qimage on windows does that along with being a great printing workflow tool. It also utilizes several different methods of interpolation, including Lanczos and vector, but its real beauty lies in how it totally removes you form the grind of worrying about this stuff yourself. I used it all the time for on-site printing.

Great articles:

Scroll down to "The three pillars of printing." http://www.ddisoftware.com/qimage/quality/

Kind of off topic, but a cool visual example of the differences between some methods: http://www.fredmiranda.com/SI/index.html

So in the end the theory about giving the printer data that is as close to native res as possible is that it keeps the printer's own software form potentially degrading the image even further without any control on your part, wether it's reducing or increasing the size. So in the example you mention of making the 648 DPI 4x6, if you resampled the file down to 4x6, and sharpened for the intended output size and output method, you would have a noticeably nicer looking image than if you only resized without resampling and printed. So in effect, it may indeed hurt the final output.

And for what it's worth, I notice the driver interpolation on my Olympus pretty readily, the reduction in quality is really pretty extreme. But even on my old Epson I noticed a difference in apparent detail resolution. The closer the images were to native res, the crisper they became. The difference was subtle, unlike the Olympus, but it was there. So some will notice it with their setup, some might not, but it is definitely happening behind the scenes.

Anyway, I hope any of this rambling makes sense.

[another view]

So it's basically a control thing? Either you control the resolution or the printer does it for you, and at that point you're not sure how it's doing it? Maybe this is one reason I'm so happy with my print quality; I've almost always done this but not specifically for this reason (sharpening, or a different crop like 8x10).

If I have it right, it sounds like an analogy could be manual vs. auto exposure mode on a camera (although one isn't clearly better 100% of the time). Just trying to understand.

[Sebastian]

Yeah, I guess so. The printer driver doesn't account for what you want to do. If you step in and make it clear, leaving no wiggle room for the software to mess with anything, it will look better.

[gahspidy]

[QUOTE=Sebastian]Gary,

Not trying to pick on you, honest. QUOTE]

Never thought of it that way, Sebastian. I'm glad this discussion came about because there is alot you brought out here that I had not been aware of before. In particuliar, the idea that there is an optimum resolution for every printer and that a printers driver will interpolate to achieve it's desired resolution. So , when I size my image to 4x6 having a resolution of 648dpi, I have not interpolated, but when I send it to my printer (Epson 2400) it's driver will interpolate the file to 720 ppi at 4x6.
So, according to the article put out by Qimage, it would be best for me to set my resolution to 720 dpi (which will give me roughly a 3.6 x5.4 image size) and then interpolate the file to 4x6 or 8x12 etc. at 720 dpi. Bicubic Smoother going up, or better yet Qimage.
I had heard good things about Qimage before.
This is something that I will surely experiment with.

[QUOTE=Sebastian]Anyway, I hope any of this rambling makes sense.QUOTE]

Absolutely, I appreciate the time you took to lay it out and point out the articles.