Understand Noise in Digital Image

There is lots of confusion about image noise. The noise that is produced by image sensors when used at high ISO (amplified output).

Bottomline is, that nature sets the limits of the perfomance of sensors that use the same basic technology. And all current sensors do. Meaning, that they will all produce noise. Small differences can be established by using amplifying technology that produces little noise. And again all cameras do, some a little better than others.

So what makes that there are larger differences in noise performance between camera models? Well, one comes from the sensor/pixel size, because larger pixels capture more photons/light and need less amplification. A 12 mpixel full frame sensor (like the D3) will produce less noise than a 12 mpixel cropped (smaller) sensor (like the D300).

But the biggest difference comes from ........ noise reduction and not from the sensor or some wonderful new technology. This is nothing new and noise reduction was first widely used in digital compact cameras which use very tiny sensors. These cameras and their small sensors produced so much noise that noise reduction was necessary to get decent output even at lower ISO. No wonder that strong noise reduction is used in all modern compacts. Fuji was the first to introduce clever noise reduction in compacts that didn't damage the image data too much.

For DSLR's very little to no noise reduction was used in the earlier models. But when the megapixel race started and the pixel sizes began to shrink, it became necessary to use more noise reduction at higher ISO, but still very moderate compared to compacts.

So what has Nikon actually done with the D300. Well. they didn't invent a noise free processor unfortunatelly. They just introduced massive noise reduction in the DSLR world like the compacts already used for some years. So, noise reduction has nothing to do with the performance of sensor or amplifiers. It's a computational process that uses a clever noise reduction algorithm. These calculations need computing power which is only limited available inside a camera body. It's not possible to house a quad core processor on 3 Gigaherz inside a camera. The camera would run out of power after 1 picture. With the limited computing power inside cameras only relatively simple noise calculations can be performed. Manufacturers have to make choices how to remove the noise, what part of the noise they want to remove and to what extent. It is clear that Nikon made some clever choises with the algorithms used in the D300. A good compromise.

But is this a blessing or a curse? Well, a bit of both. For reporter stile photography it is very easy now to use high ISO and get very decent results. A bit like the snapshooter that uses compacts. But for less action driven photography (landscape, artistic, architecture, ...) it is still much better to use noise reduction outside the camera with RAW converters and/or dedicated noise reduction programs designed for use on powerful computers, not aimed at compromises but at maximum results. These results will in many cases be better than from noise reduction in camera. But more importantly, YOU will be in total control of the noise reduction and the end result.

This I see as the biggest draw back of putting more and more post-procesing inside the camera. It could create laisy photographers that find it convenient that their camera makes all the decisions for them. You don't buy an expensive D300 to become a snapshooter.

Ron - Netherlands
Art Director