This article is directed primarily at my step-father; if you can take something useful away as well, fantastic!
Ken‘s a smart guy, and though it’s guaranteed he doesn’t know, he has taught me much over the years. His article on The Megapixel Myth is dated, but about 99.7% of the content is still true and relevant today in October of 2014. If you want something newer, check out this Digital Trends article or, heck, just do a search yourself. The bottom line is that after about 4 to 6 megapixels, megapixel count no longer matters; at 6 megapixels, you can print to 20 x 30 inches and get results that most people can’t differentiate from higher resolution images (the only problem with this is that many shops won’t actually let you print that large from an image of that size because it falls under certain dot-per-inch thresholds).
Here’s the ugly reality that flies in the face of the facts presented above: Most people – either by accident, by intention, or out of ignorance – don’t fill the frame when they shoot an image. When I shoot weddings and seniors, I have no idea if Mom is going to want any particular image printed at wallet, 4×6, 5×7, 8×10, 11×14, or even 20×20. So I have to shoot both horizontal and vertical on most images, and though it’s deeply ingrained in me from my college and DeCloud days (FILM IS EXPENSIVE, FILL IT ALL!!!), there’s absolutely no way that I can fill a frame and get it to crop correctly in post. I absolutely and intentionally cannot fill the frame when I’m shooting.
So when I’m done cropping an image, it’s not uncommon for me to wind up with a finished image that’s “only” 2524 x 3212 (a little smaller than 8MP) – and I’ve occasionally cropped down to the 4MP range. And that’s from a 20MP original RAW image. If you don’t see why that’s alarming, hang on – I’m getting there.
The point is this: all other things (noise, contrast, tonal gradation, color accuracy, etc) being equal, pixel count matters – and the more pixels you have, the more breathing room you’re affording yourself in post-processing. Especially if you’re shooting JPG (as opposed to RAW), here’s the truth: If you’re shooting in other than “Large”, you’re robbing yourself of precious (mega)pixels.
Yes, Yes, Yes: If your images are going online you don’t need a whole lot of pixel real estate. Unless you’re on the bleeding edge (and have too much money), you’re not running a 4K monitor – which is to say that as of this writing, you (and your audience) are probably running a “high def” monitor with a maximum resolution of 1920 x 1080 (or about 2MP). With that in mind, for online posting, your images shouldn’t be much more than 1,000 pixels on the wide edge and probably shouldn’t be taller than about 700 pixels. People get annoyed when they have to scroll and move windows to see an image in its entirety. (On a side note, posting in smaller than full-rez also helps to deter image theft and copyright infringement.)
However, if you’re going to take an image to print, things change drastically. For best results, your images should be printed at a minimum of 300 dots per inch (dpi). MPix, one of the best consumer processing and printing labs in the world, recommends 150dpi as the minimum, and in practice, they won’t actually allow you to print an image that falls into a lower dpi count.
Wait, what’s “dots per inch”? According to Wikipedia, it’s “a measure of spatial printing or video dot density, in particular the number of individual dots that can be placed in a line within the span of 1 inch”.
As digital photographers, we live and work in a word of pointillism. In this close-up excerpt from Seurat’s La Parade, you can see the paint daubs that comprise this man’s head:
When viewed from a distance, the dots become more-or-less indistinguishable and the guy looks somewhat normal – at least for his time and place.
In photography, we don’t work with dots; instead, we work with pixels. For example, take a look at this nice picture of Pi:
If you zoom in on her eye, you can actually see the pixels:
When you go to print a digital image, the terms “dot” and “pixel” effectively become interchangeable. Which means, all of a sudden, that a 1000 x 700 pixel image is too small, period. Here’s the simplified math:
1000px / 150dpi = 6.6 inches
700px / 150dpi = 4.6 inches
The actual math is more complicated than that because dots per inch doesn’t translate well into line spacing – but the complicated math doesn’t matter. The real-life end-result is that at 1000px by 700px, your image won’t print larger than 4″ x 6″, the end. At the recommended 300 dpi, you would wind up with a maximum print size of just 2″ x 3″!
If you want an 8″ x 10″, the reality-based requirement is 1200px by 1500px (~2MP) image – and, as it turns out, when people go to hang something over the fireplace, they actually want 16″ x 20″ (2400px by 3000px, aka ~7MP) or bigger. All of sudden, cropping a 20MP original image down to a 4MP final image is a big deal. See why I was concerned way up there towards the beginning of this article??
But wait, you say! I can just resize (or rescale, depending upon your software) the image! I can turn a 4MP image into an 8MP image! And yes, you probably can, and get acceptable results – but if you try doing that with a 1000px by 700px image, you’re going to see defects in the in the final image.
Why, you ask? Here’s why: If you have 4 inches of image at 150dpi (that’s 600 pixels), and you need to get to 16 inches of image at 150dpi (that is, you need 2400 pixels), then you need 1800 extra pixels. Where do those extra pixels come from? Especially bearing in mind that those pixels need to be in the right places, be in the right order, and be of the correct shade of the correct color, sourcing those extra pixels quickly becomes a daunting task, no?
So when you ask your software to resize something bigger, you’re literally asking your software to invent image data, to create it out of thin air. This process normally involves looking at the existing pixels and taking a best guess at what the “missing” pixels might look like – and believe me, some software does this job a lot better than other software. I’ve seen some of my resized images where a genuinely small and innocuous freckle in the original image became a large misproportioned birthmark in the new larger image (yes, this was in the late 90s using free software, but the point is still valid).
Which is all to say that once you have met the requirements of pixel quality, pixel count matters.