A characteristic of most digital camera images is something called digital noise. Noise is the broad term used to describe the occurence of dots or specks, some coloured,some not, where there shouldn’t be any, in a digital image. Noise only really becomes noticeable in an image when the iso is raised to higher levels. Some cameras show less noise than others, and newer cameras usually show less noise than older versions.
People react differently to noise. Some people don’t even notice it or bother about it. Others go to great lengths to minimise or remove noise from their images. If you shoot jpeg images with your camera, you can choose to have the camera apply noise reduction to the high iso images as you take them. However, noise reduction like this gets rid of the noise, which is detail, at the cost of other fine, detail in the image that you don’t want to lose. Strong global noise reduction applied in-camera can lead to images that lack detail and appear ‘waxy’. It is also common practice when shooting in raw, to apply noise reduction when processing the image afterwards. You can then control where and how much noise reduction is applied. This is an effective technique but it is time-consuming.
Noise usually becomes much more prevalent and noticeable, the higher the iso or sensitivity setting of your camera. As a wildlife and nature photographer, I make use of natural light. I am often forced into photographing when light levels are low, when my subjects may be active. In order to attain a fast enough shutter speed, I am compelled to put up the iso sensitivity to a level where image quality may be visibly impacted by noise.
After processing thousands of images over several years, there are certain things that I have noticed about where noise shows up the most, and where it is least visible. There are ways of composing that can help minimise how much the visible noise affects the image.
Canon 1Dmk4, Canon 300 f2.8 IS. Shutter speed 1/320s at f4.0. Iso 3200. No noise reduction, lightly sharpened. Click on the image directly see a larger version.
Noise shows up most prominently against smooth, even colours and tones in an image. This isn’t surprizing, as noise is a form of unwanted detail, and detail stands out best against backgrounds lacking detail. In images that have been heavily cropped, noise will become especially visible against low contrast or smooth areas. The leopard image has had minimal cropping, and only light sharpening. What noise there is shows up more against the background than in the strongly textured fur of the cats. This image was included for illustrative purposes only.
Verreaux’s eagle owl. Canon 1Dmk4 and Canon 300 L f2.8 IS. Shutter speed 1/100sec, f/4.0. Iso 1600. Plus 1/3 exp comp. Click on the image directly to see it at a larger size.
As a nature photographer, virtually all of my subjects have strongly textured, detailed coats, fur, skin or feathers. I find that noise is not nearly so visible against such strong textures. Once I am forced by very low light to go to a high iso setting, I can expect noise to visibly show up in the parts of my image where contrast is lower, usually the out-of-focus background, and for the noise to remain less visible in the higher contrast textures of the subject. I keep this in mind when I am composing, and in such situations I will try to get as close as I can to the subject, and to compose a little tighter than I would otherwise. In the photograph of the owl above, I wanted more of the owls feathers filling my frame rather than the green out-focus background behind, or the sky, both of which would show the noise more clearly. Of course, if you are very far away from your subject, and the light is low, then you cannot benefit from this approach. It is worth thinking about when you have options with your framing.
Similarly, your choice of aperture will influence how much of your subject is in focus, and at close range, or with long focal lengths, it is possible to have such shallow depth of field that only some part of the subject is in sharp focus. Again, be aware that those parts of the subject that are out of focus due to depth of field will show more noise when sharpened than the rest of the subject.
In a similar vein to this, it follows that sharply focused images, properly sharp, will show less noise than those images that are not quite in-focus. This is because the sharply focused image will have a larger part of the frame with higher contrast to the textures, and more definition, and the noise will be less visible, if at all, in those parts. If the same subject is captured even slightly out-of-focus, the contrast is lower and the noise will stand out more.
I have made use of two sequential images from a burst, taken, with my Canon 5Dmk3, to show this more clearly. The leopard images that follow are posted here not for their aesthetic value, if any, but rather for illustrative purposes.
Canon 5Dmk3, Canon 70-300L. Shutter speed 1/2000sec at f5.6, Iso 1600
This is the full-sized image, downsized for web use, with no cropping. In this frame, which is the earlier frame of the two shown, the leopards face is clearly in focus. This was shot at a focal length of 188mm, and the cat was moving relatively fast, directly toward the camera. I kept the iso raised because I wanted my shutter speed to remain high for the cats approach.
Below this is a very deep crop, (approx 1.2 megapixels) of the face and eyes, posted to show the sharp focus, and noise level, as it is.
Sharply focused, 1.2 megapixel crop, resized for web, lightly sharpened. No noise reduction. Click directly on image for bigger size.
Even at this large size, there is little noise visible in the leopards fur, face or eyes. For this frame, I managed to keep the focus point on the cats face and the image is acceptably sharp, which was a very good result given that shooting conditions were not that easy for autofocus, with no direct light, and a bright background.
The next frame in the sequence is posted below.
Canon 5Dmk3, Canon EF 70-300L. Shutter speed 1/2000sec at f5.6, Iso 1600.
This is the full-sized image, no cropping, downsized for web. At this small viewing size, it is not really possible to see any focus difference between this image and the previous one.
Poorly focused, cropped to 1.2 megapixel, resized for web. Lightly sharpened. Shutter speed 1/2000sec at f5.6. Iso 1600. Click directly on image for bigger size
In this frame, I failed to keep the focus point on the leopards face and instead have sharp focus on the top of her front leg. I kept this image to illustrate the differences in how visible the noise is in this image, compared to the sharply-focused image. I cropped the picture very heavily until just 1.2 megapixels remain.
To properly appreciate the differences in focus, click directly on each of the cropped images and examine the areas around the cats face especially. It is easiest to see the extra noise in the second, poorly-focused image in the smooth, yellow tones in the lower half of the cats eyes. Both of these images have only been very lightly sharpened. Adding more sharpening, even a moderate amount, would make the noise on the poorly focused image much more visible. Any noise in the background will remain the same between the two images.
Although learning a little more about where noise shows up doesn’t prevent it from being present in your images taken at high iso settings, it does still allow you to better understand why noise may be worse in some images than in others.
There are some things you can do to minimise just how much visible noise you have to deal with.
To sum up, noise shows up the most in parts of an image that have even, constant tones and very low contrast. When the light gets very low, those compositions where the subject fills a lot of the frame may allow you to shoot at even higher iso settings without the noise showing as much.
Good lenses that can focus accurately on moving subjects even when the light levels are low, like the Canon EF 70-300L IS and the Canon EF 300 f2.8 IS II, help the photographer to attain high-quality images. Sharp shots show less noise, at least in the subject if that is where the focus was.
There is a lot more that can be done during processing to minimize noise, from the selective application of contrast and sharpening to the subject only, and specialized noise reduction programs to take out what remains, but those may be topics for another post. By starting out with source images that don’t show excessive noise you will be able to achieve the best results in your finished images.