The Great Megapixel Swindle: An Example @ Petavoxel

Original Post Here

---

So, you’ve probably gotten the idea: I’m a bit outragedabout ridiculous megapixel inflation in point & shoot cameras. But Is this just some theoretical problem? Or does it really make for bad pictures?
My apologies that I haven’t yet given an illustration of what I’m talking about. So let’s take a look at the image below:

100% crop from typical point & shoot

Whoa, dreamy! What on earth is going on here? Have I mistakenly substituted a Matisse painting by accident? Is this an image from Photoshop’s “watercolor” filter?
Not at all. This is a 100% crop from a typical compact-camera snapshot. We’re seeing about 4% of the total frame. The original image is here.
(I should be clear that I don’t know the photographer; and I am only singling out this image as being typical for its camera type. This source conveniently includes both the fullsize image and the EXIF data, below the image.)
What’s horrifying is that this photo (as we learn from the EXIF) was taken in what ought to be the ideal situation for a digital compact: Bright sunlight, with the sensitivity setting at a moderate ISO 100.
There is no issue of camera shake, as the shutter speed is 1/500 second. The f/4.6 aperture is within one f/stop of the widest possible (at that focal length), to reduce diffraction.
Yet the image looks JUST TERRIBLE.
Okay, let’s name names here. This photo was taken with aninexpensive camera from Olympus: the FE-26.

Olympus FE-26, a typical P&S

So, is Olympus just the crappiest manufacturer ever? Well, I will concede that the lens on this camera seems to be particularly poor. Look at that crazy color fringing!
But when it comes to the smudgy lack of detail, the problem is the same as with every other compact camera today—too many pixels.
The FE-26 is a “12 megapixel” model (actually it’s more like 11.8 Mp) using a 1/2.33″ sensor. This means each pixel is about 1.5 microns wide. When pixels are that small, the random difference in photon counts between adjacent pixels can add quite a bit of noise to the image. To solve this, the camera’s processor chip applies a noise-suppressing algorithm, which unfortunately smears out all the fine detail and texture in the scene.
Admittedly, different camera companies can be more or less clever about their noise-reduction processing. This one looks especially bad, but it nicely illustrates the kinds of artifacts that can result.
But what’s clear is that the surplus megapixels of this camera are certainly not delivering additional image detail. And as you increase the ISO or stop down the lens, quality will only get worse.
As I discussed last time, 1.5 micron pixels are always going to struggle with diffraction blur. The theoretical minimum size for a light spot focused by an f/3.7 lens is 5 microns. Stopping down the lens makes the diffraction blur larger.
You can be certain that somewhere within Olympus, engineers are quite aware of the noise and diffraction problems caused by tiny pixels. But the marketing department steamrollers on, demanding that every year the megapixel spec keeps going up. Olympus’s new FE-47 has 1.4 micron pixels—50 Mp per square centimeter.
This is madness. Higher megapixel numbers are a swindle. They make pictures worse. Stop.