The Megapixel Race Continues

Background

In July 2004 I wrote my first Tech Corner article titled The Megapixel Race. The article discussed the steadily increasing pixel count of both consumer and professional cameras and the tradeoffs associated with stuffing more pixels into the same area on the imaging sensor. At the time (in 2004), manufacturers were increasing resolution by about one megapixel per year to keep the pot boiling and keep consumers coming back to the store to buy the latest models with higher resolution. Is the megapixel race still on? What has changed in the last four years, and are people still counting pixels when making decisions about which camera to buy? Let's take a look at the current state of the market with regard to the ever changing technology.

A steady race

Back in 2004 when I wrote my original article on the subject of camera resolution, the top of the line prosumer dSLR had about 8 megapixels and I stated in that article that the resolution was growing by about one megapixel per year. Well, here we are in 2008 and while the "top" is a little less easily defined these days, the high end dSLR's are now generally in the 12 - 13 megapixel range, so it's a steady race still running at the pace of about a megapixel per year increase in resolution. Back in 2004, manufacturers were content with just cramming more pixels into the same area, however, creating higher resolution images that consistently degraded in quality year by year. While the pixel count was going up, noise went up in proportion, bringing overall image quality down. Cramming more pixels into the same area reduced the pixel size thereby reducing sensitivity of the pixels and increasing noise. That's where things have changed a bit. Manufacturers finally realized that they couldn't keep stuffing more pixels into the same capture area while letting image quality suffer, so a more balanced approach is being used today with better hardware and better software (in the camera) to compensate for the increasing resolution.

Traffic control for crowded sensors

If you keep trying to see how many (more) people you can fit into a compact car, eventually you'll reach a point where you realize the car needs some upgrades to be able to carry the load. That's exactly what happened with the megapixel race. While resolution is still steadily increasing, so is the technology behind the pixels. Today's sensors have better on-chip noise control and better dynamic range. In addition, cameras are relying more and more on adaptive noise reduction. This adaptive noise reduction is basically noise reduction software that resides in the firmware of your camera. Just a few years ago, shooting at ISO 1600 meant getting a very noisy image from your camera and then applying post-processing noise reduction using one of the more clever noise reduction software programs on the market. Now manufacturers are building in this noise reduction right in the camera so that the JPEG images you get from your camera (or create from raw using the included raw processing software) are already filtered based on the ISO speed used for the shot. Using more complex noise reduction in camera can give the "illusion" of lower noise levels at higher ISO speeds while in fact, the noise levels are still quite high but have been reduced by in-camera noise reduction techniques.

Paying the piper

The balance of resolution and signal-to-noise ratio is a bit like conservation of energy. Noise reduction is a tradeoff. You can lower noise, but it is inevitable that lowering noise will lower resolving power of fine detail as well. Clever adaptive noise reduction algorithms can make this fact less noticeable by using things like edge detection to drive the strength of the filter, but the tradeoffs will still be visible if you look for them. People seem amazed that the very latest digital cameras like the Nikon D300 or Canon 40D have higher pixel counts but substantially less noise than prior models yet when you really look at photos from these cameras, the tradeoffs are clear. Sure, there is less noise, but the noise reduction on these cameras is so heavy-handed by ISO 1600 that there is a noticeable decrease in resolving power in order to achieve those low noise levels. Overall, the image looks "cleaner" as people tend to notice noise grain before lost detail. So perhaps this increasingly heavy noise reduction (more NR as you increase ISO speed) makes a better balance, but you're still paying the piper in the end!

Take a look at the two shots above. The one on the left was shot at ISO 200 with a Nikon D300. The one on the right was from the same camera with identical focus, shot at ISO 1600. Without comparing to the image on the left, you might be amazed at the low noise levels in the ISO 1600 shot, however, placed next to the ISO 200 shot on the left, we can see how much detail is lost in the ISO 1600 shot due to the increased noise reduction used by the camera at the higher speed. Note that these are both JPEG photos directly from the camera.

Putting it all together

While manufacturers have made great strides in hardware, increasing dynamic range while decreasing noise, you have to be careful when evaluating performance of newer cameras (particularly dSLR's). We all have a tendency to open ISO 1600 or ISO 3200 shots and sit at our monitors looking for noise grain in the shadows. Doing this may lead you to miss an important side effect of those smooth images: the fact that much of the detail has been smoothed over along with the noise grain that was removed! Keep in mind that the very latest model cameras have gotten quite heavy handed at noise removal so it is important to compare both noise and available detail in the photograph to what you'd get at a much lower ISO speed. This fact should be evident when online reviewers take the same shot with varying ISO and place them side by side in the review. Keep in mind that I'm not saying "heavy" noise reduction is a bad thing. Overall I think the high ISO photos from the latest dSLR's look very good. Just keep in mind that no miracles are being worked with the latest and greatest cameras that claim incredibly noise free images from very high ISO speeds. The answer is in the noise reduction... and with the right post-processing noise reduction software, you could probably come close to the same quality with an older camera. The bottom line is that the increase in hardware performance is certainly there, but what sets these latest cameras out ahead of the pack is the more complex and stronger noise reduction being used in the processing of the data.

Summary

While the megapixel race continues at a steady pace of about one megapixel per year or slightly more, advances are now being made to increase hardware and firmware performance so that cramming more pixels onto the same size sensor will not equate to noisy photographs. Keep in mind that the latest cameras are using some heavy handed noise reduction algorithms to achieve their much touted high ISO performance, however, so when reviewing high ISO performance from the latest cameras, keep an eye on more than just the amount of noise (grain) in photos: also take a look at how much detail gets "smudged over" by the noise reduction algorithms. I personally find some of the noise reduction algorithms a bit too heavy handed in that when shooting JPEG's in-camera, there is so much noise reduction at (say) ISO 1600 that a lot of fine detail gets lost along with the noise. For this reason, it is even more important to shoot raw photos with the latest cameras so that you can make the decision as to the proper balance of noise versus detail. A JPEG from the camera that has been overly softened due to the ISO 1600 or ISO 3200 noise reduction cannot be rescued whereas you may be able to reduce the strength of the noise reduction on the raw file in order to bring back some of the lost detail if needed. Suffice it to say that evaluating high ISO performance on the latest dSLR cameras is becoming a lot trickier. You can no longer simply open an ISO 1600 shot and look for noise grain. You must compare the shot against a lower ISO shot of the same scene to see just how much detail was lost to noise reduction. I hope this article will help people better evaluate the performance of the latest cameras, particularly dSLR's, and will enable people to better see and understand the whole picture... pun intended! :-)

-- Mike Chaney