Claims about megapixels
David Pogue, in his Nov 21 article The Truth About Digital Cameras, makes the claim that there is no difference between a 13-megapixel, 8-megapixel, and 5-megapixel image when printed at 16 X 24 inches (the 8 and 5MP images were downsized from the 13MP image). He claims:
I’m telling you, there was NO DIFFERENCE.
Thousands of people read his columns, so such broad-brush claims are worrisome without at least some description of the methodology, camera used, etc. Here are just a few of the issues:
- With which camera was the image taken? I don’t know of any 13-megapixel camera. The Canon EOS 5D is 12.7 megapixels. Was the image one that claims 13MP with a “high-res upsize mode”, but really only a 6MP sensor? Was it a digicam or a DSLR?
- What kind of print was made, and how was it printed?
- Was the image sharp to begin with? Was it shot on a tripod? Was it focused accurately? Was the subject perfectly still? Was there adequate depth of field? Was the lens stopped down so far as to be optically limited? Or was the image just 13MP of substandard blurry pixels?
- What was the viewing distance at which the prints were viewed?
- Was there lighting on the images so that they could be seen clearly?
- Did the observers have 20/20 vision?
Pogue’s sample print appears to be a baby picture. Certainly, portraits are more amenable to lower megapixel images. But I’m skeptical that the picture is technically good enough to be used as the basis for a comparison, knowing that a 13-megapixel camera requires extremely accurate focus (or it’s instantly a 6 or 3 or 1-megapixel camera). We can’t know, since Mr. Pogue offers no actual-pixels crop, nor does it appear that he is aware of such issues.
My own test: I don’t like to theorize or guess about such things. So I performed my own test. My results are somewhat different. Since my prints were made slightly smaller (22.6" vs 24"), any differences will be less obvious.
At a distance of about 36 inches, the print from the full-size original looks superb when printed 22.6" wide. The impression of smoothness with fine detail is strong, and I consider it a fine-art quality print. The overall impression is of very high quality. It looks like it might have come from a 4X5 camera.
The print from the 8-megapixel image looks quite similar at first glance, but it has a slight harshness and loss of tonality to it, even when viewed at 5 feet away, where the differences in fine detail cannot be resolved. Fine detail is a bit smudged when observed at closer distances. This print would still be acceptable as a fine-art print. The average “person on the street” would still be impressed, and likely would not be discerning enough to note the differences (maybe this is why fine-art quality photos make for a tough living).
The print from the 5-megapixel image looks noticeably harsh and “brittle”—even from 5 feet away. I would not willingly sell such a print, as no discerning person would consider it a first rate effort. The “person on the street” might see the difference with side-by-side prints under good lighting at a distance of less than 4 feet, but many would not. I showed the prints to my wife, under less than ideal lighting, and she noted that the 5-megapixel image looked different, but she was hard-pressed to distinguish the prints. Then again, she was not wearing her reading glasses! Visual acuity, lighting, and a discerning eye all matter.
Conclusion: Does this mean that you should just buy a 6-megapixel camera instead of a 10 or 12-megapixel camera? It all depends on what you’re after—snapshot quality or fine-art quality or something in-between. I’ve never been satisfied with 6-megapixel landscape images (Nikon D1X, Nikon D70), but I’ve also been very satisfied with 4-megapixel portraits (Nikon D2H)! The subject matter counts, pixel quality matters (eg a DSLR and not a digicam), and good technique is critical to make those 10 or 12 megapixels perform.
Details: I used my Norway Birches image taken at f11 on a tripod using mirror lockup with the Canon EOS 5D and the Nikon 85mm f/2.8D PC Micro-Nikkor at f11, converted from the RAW (CR2) file into a 16-bit TIF (4368 X 2912 pixels or 12.72-megapixels). The image was critically focused, and is about as sharp an image as the 5D can produce.
Actual pixels crop, Unsharp Mask {150, 0.3, 0}
- the uncropped images were printed 22.6" wide, allowing the printer driver to scale the image.
- sharpening for all prints was {500, 0.3, 0} using Photoshop Unsharp Mask on the Lightness channel.
- printed on the Epson Style Pro 7600 with the ultrachrome ink set.
- the 8-megapixel (3464 X 2309) and 5-megapixel (2738 X 1825) images were downsized from the original using Photoshop
Bicubic Sharper.
