A Few Megapixel Mega-Myths
While we're on the subject of megapixels, it's time to address a few common megapixel myths.
Are All Pixels Created Equal?
My new Samsung Galaxy Note 3 mobile phone features a camera with thirteen megapixels. That certainly does sound like a lot given that many digital SLR cameras weigh in at less than that. But the bottom line isn't really just a function of counting photosites or pixels.
While it is true that each one results in single set of red, green and blue values (ignoring the complexities of raw data interpolation and such), some values may be more accurate than others. As we discussed last week, light is both a wave and a particle. Each photosite essentially functions as a simple counter, tallying up the number of photons (particles) of light that fall on it during the exposure. But a number of factors cause noise in the signal, meaning that some of those counted pixels didn't really come through the lens at all, or did and somehow ended up in the wrong place when they struck the sensor. Especially under low lighting conditions, there can be scant few photons to be counted, so even small amounts of noise can add up to big problems.
Another important factor, and the one that is important to our topic at hand here, is that the surface area of each photosite is also a major factor on how many photons it can collect while the shutter is open. Think of it as being analogous to the change in light that makes it through your lens increasing as the size of the aperture opening increases. A bigger aperture hole lets through more light, and bigger photosite collects more photons. Both are proportional to the square of the diameter of the collector or opening in question. So if we cut the size of the lens aperture in half, only one quarter of the light can make it through in a given amount of time. Likewise, if we shrink the size of a photosite in half, it will see one quarter of the number of pixels to be collected.
Shrinking sensor sizes and increasing pixel density can make the job of each photosite that much harder. All other things being equal, bigger photosites create better images. And if we place the same number of photosites on a smaller sensor they will naturally be smaller than on a larger one.
Does Doubling the Pixel Count Double the Resolution?
It's easy to assume that if you go from shooting a six megapixel camera to shooting with one that has twelve megapixels, the resolution will double. After all, the megapixel count did, so naturally the resolution will as well, won't it? Nope.
To understand why it is first necessary to decouple the concept of image resolution from the technology used to capture that image. It shouldn't really matter whether one image was captured digitally while another one was shot on film. If we set the two of them side by side, the one that we can see more detail in must have a greater resolution than the other. If we think of resolution in an objective fashion as being the number of lines (or line pairs) per millimeter we can resolve, then resolution is a function not of area, but of linear measures.
If you're with me so far, it's time to consider what happens when we double the number of pixels on a sensor. Assuming it retains the same aspect ratio, its dimensions along each side will have only gone up by a factor of the square root of two. Many things in the world of photography boil down multiples of the square root of two, and this is another. In order to double linear resolution we'd need to actually quadruple the number of pixels. While every increase in resolution can indeed make for a better image, you'd need a twenty-four megapixel camera to actually double the resolution of your old six megapixel model.
Does the Megapixel Count Measure Resolution At All?
So given everything above, just what does knowing the number of megapixels for a camera mean? In a sense, not much. To illustrate, imagine I open an image — any image at all — in Photoshop and resample it to make it bigger. I'll naturally end up with more pixels, but the actual resolution won't change at all. It couldn't, since there's nowhere for any added detail to come from. Contrary to what reruns of crime scene dramas on cable television would lead one to believe, it's simply not possible to extract more information from an image than is actually present in the image to begin with. You can sharpen the existing detail to make it more evident, but if it's not there to begin with, no added resolution is possible.
Assuming everything else is equal, increasing the pixel count of a sensor should give us greater resolution, but rarely is everything else equal when people make such comparisons. Indeed, such comparisons are typically made between two explicitly different cameras. You might compare your old camera to your new one, or your cell phone to your DSLR. Regardless, many things are thereby different making any simple numerical comparison of pixel counts meaningless, or nearly so. Even comparing one generation of DSLR of the same brand to the next, the technology underlying the sensor used in each may have changed along with the megapixel count. Comparing them may not be as bad as comparing apples to oranges, but it's at least like comparing an old, over-ripe apple to a new, ripe one.
Any meaningful way to compare resolution would have nothing to do with counting pixels. Instead, it should be based on actual tests of image resolving power. I'm not here to promote any given such test, but a great place to start would be dxomark.com run by the same folks who produce the DxO Optics Pro raw converter.
Do Megapixels Magnify Your Mistakes?
Here's one last megapixel myth to consider. People often attempt to gauge the quality of their new camera by zooming in on the images it produces to a full 1:1 actual pixel view. Sometimes, they don't like what they see. Edges of objects that appear somewhat blurry at full scale can lead one to believe something is wrong. It may be, but exactly what is in fact just as unclear as those edges. Sometimes people blame the camera, assuming it must be defective or shoddy. If they paid a lot for that camera though, and it came from a reputable manufacturer, it can also be tempting for photographers to blame themselves for the problem. I hear this a lot, that an increase in megapixels magnifies not only resolution but also the visibility of any mistakes made by the photographer.
This can seem true, but it really isn't. Not if we are comparing comparable things. If we go back to our earlier example of a six megapixel camera compared with a twelve megapixel, or perhaps one with twenty-four megapixels, they clearly don't produce images that are the same size. If we blow them all up to the size of the biggest, or shrink them all down to the smallest, they should look pretty much the same side by side. That is, if we resize them all to be comparable, any appearance of blur caused by hand holding, cheap tripod use or other factors will also be resized to be comparable. But if we look at each at maximum native resolution, everything will be bigger in the bigger image. The trees, rocks, mountains, flowers and other subjects will appear larger just as will any softness from poor focus or other bad technique on the part of the photographer.
Often, what the true cause of apparent image softness from upgrading to a camera with more megapixels is the quality of the lenses (and filters if you use any) used in front of that higher megapixel camera sensor. Lens resolving power isn't something to ignore. Good glass isn't cheap, and while camera megapixel counts continue to go up as camera prices go down, the same can't be said for lens prices. High quality lens prices haven't changed much at all in recent years and likely won't. The difficulties in making good sharp lenses don't benefit from advances in computer chip and digital electronic technology. Yes, computers have entered the world of lens design, but the ultimate factor is the glass and the machining and polishing of that glass. And as sensor technology does improve, the consequences of saving money on lenses can become more evident, at least when evaluating the results at full size.