Thinking about buying a new digital camera? The best advice I can give you is to avoid maxing out megapixels and you’ll get sharper photographs with less grain (noise). This seems counter-intuitive, but is the effect of camera technology running into physical limitations of lenses and light itself.
David Goldstein has written a full-length paper that explains the physics, but here are the key take-aways on megapixels:
- For a full-frame (35mm) SLR camera, 25 megapixels is about as good as it gets.
- For a smaller “APS” digital SLR camera, about 20% smaller than 35mm, 10 megapixels is the point of diminishing returns, and 15 megapixels is on the edge (and has noticeably more noise)
- For a pocket camera where the sensor is about 1/5th the size of 35mm, 4-5 megapixels is the point of diminishing returns, 10 megapixels is pushing it even at very wide apertures, and 12 megapixel or greater pocket cameras will take disappointingly fuzzy pictures at aperture f/2.8 or greater.
Adding megapixels is good until it’s not
Why is this? Fundamentally, camera manufacturers have to balance increasing resolution from more megapixels with two competing forces: noise from the digital camera sensor and softness caused by lens diffraction. You’ve probably seen that photos taken at ISO 1600 on your camera are much more grainy than those taken at ISO 100 — that’s what noise looks like in your pictures. At the same time, the lens aperture (f-stop) puts an upper limit on how much detail can be resolved by the camera. Overall, more megapixels don’t automatically mean better pictures.
Noise is a result of the fundamental physics of light and the way that digital color cameras capture images. Light is made up of photons, with more photons meaning brighter light. The digital camera sensor counts the number of photons that arrive at each pixel to build the overall digital image. In a 21-24 megapixel 35mm digital camera such as the Canon 5D Mark II/III, a properly exposed image at ISO 3200 results in about 20 photons hitting each sensor pixel, assuming a perfect lens. That’s not very many, and it gets worse because of the way digital cameras deal with color. A red, green or blue lens covers each pixel so that only light of that color is detected at each pixel. Dividing 20 by 3, that’s just 6 or 7 photons arriving at each pixel.
Even with a super efficient sensor, this means there’s a lot of noise due to mis-counting of photons. If we add more pixels to the sensor, the number of photons arriving at each pixel goes down because each pixel has to be smaller, which means more noise. That’s why adding more megapixels makes the photos more grainy, especially in shadow areas.
When it comes to image resolution, adding more megapixels helps the camera resolve more detail – up to a point. Adding more pixels reduces the pixel size on the sensor, which also increases the effects of diffraction, caused by the lens. Diffraction is a fact of life: Isaac Newton discovered that light beams spread into a circle pattern as they pass through an aperture such as the iris of a lens. Diffraction effects make pictures look “soft” and lack sharpness. Diffraction softness becomes more noticeable as sensor pixels get smaller because the diffraction circles spread over more pixels.
Too many megapixels can soften your photos
Newton showed that diffraction increases as the aperture gets smaller. As the f-stop increases and the aperture of the lens gets smaller, diffraction increases. The point where diffraction starts to noticeably affect picture quality is called the diffraction limit. The diffraction limit for a 21-megapixel full-frame (35mm) sensor is f/10. This means that at f/11 or greater, the image will get softer and softer. The limit is f/8 for a 10 megapixel APS-sized sensor, and f/2.8 for a 12 megapixel pocket camera. Most pocket cameras don’t have a lens that offers that wide an aperture!
In other words, while adding megapixels initially increases camera resolution, there comes a point when it leads to noticeably softer pictures (as well as more noise). For example, in this New York Times review of a new, tiny Samsung ST80 14.2 megapixel pocket camera, the reviewer notices that “people and scenery in the background looked murky, and the photos lacked crispness”. Recall that to get a sharp picture on a pocket camera of 12 megapixels, you need f/2.8 or better aperture to avoid softness from diffraction. The ST80 has 14.2 megapixels and a maximum aperture of f/3.3, so pictures are guaranteed to be soft from diffraction – in addition to the extra noise from pixels that are so tiny you can fit 50 million of them in an area the size of your fingernail.
With today’s cameras, don’t obsess over megapixels, and stick to lower pixel counts where possible — especially with pocket cameras, where the limitation in picture quality is more likely to be the lens rather than the sensor.
To find a camera with low megapixel count, use DPreview.com’s Camera Statistics pages. For more than 1600 cameras, DPReview lists the number of megapixels per centimeter squared (cm2) — lower is better. For example, the Canon 1D mark IV has just 3.1 megapixels per cm2 as you might expect for high-end professional gear. The 5D Mark II has just 2.4 megapixels per cm2. The Canon SX210 pocket camera has 50 megapixels per cm2 as it’s a 14.1 megapixel pocket camera with a tiny sensor. Instead, consider the Canon S95 with a pixel density of 23 megapixels per cm2 — a 10 megapixel camera.