Noise In Photography: The Guide
The terrible high ISO noise performance of the Leica M9 camera is evident in the sky of this night time photo made at ISO 1000.
Quieten Down, There's Too Much Noise
I just wanted to share with you an issue that may prove problematic on occasions in your own photography. I'm talking about noise.
To illustrate some of the concerns associated with noise I've included some photos from a Leica M9 camera.
There were many wonderful things about this camera, such as it's color rendition and, of course, the quality of Leica-M lenses.
However, the Leica M9 camera, incorporated a CCD sensor which produced unacceptably high levels of noise when used at a moderate or high ISO.
While there was a lot to like about the camera, and despite the fact that I was a long term Leica user, these unacceptably high levels of noise caused me to sell the camera.
These days my camera of choice is a Sony A7rii. It's noise characteristics, for both high ISO and long exposure photography, are excellent.
Noise: A Simple Explanation
Noise is a consequence of your camera's sensor being heated up, as more current is passed through it, during long exposure times (e.g., 1 second or longer) or when set to a relatively high ISO.
Noise is generally more likely to become visible on images made with a smaller sensor camera rather than a larger sensor camera that's of a similar vintage from the same manufacturer.
That's simply because the larger sensor (e.g., full frame) camera will have larger individual photo sites than is the case with a smaller sized sensor of the same pixel count.
Noise: Here Today and Gone Tomorrow
Noise can occur anywhere in your image, but noise will be particularly evident in the following areas:
Underexposed images.
Shadows and darker tone areas of the image.
Areas of smooth tonality (e.g., blue sky, clear skin, largely texture-free wall).
Conversely, noise can remain largely hidden in highly textured areas which is why, within the same photo, you're more likely to see noise appearing in a blue sky, but not in grass.
Noise can appear in an image as white dots which is referred to as luminance noise. Colored dots are called chroma noise.
While loosely associated with grain, from the days of film-based photography, noise tends to have a less random and, therefore, a more even appearance than was the case with grain.
Film grain is roughly round or tabular in shape.
Conversely digital sensors produce noise that’s more of a square shape. That’s because camera sensors incorporate light-sensitive, square pixels that are arranged in a grid pattern known as an array.
Poor high ISO performance evident in the Leica M9 camera at ISO 640 of this night scene in Paris, France. Look in the top left corner of the frame.
Noise: JPEG vs RAW
As mentioned previously under exposed images are more likely to display noise. That’s particularly the case with JPEG images.
With less data available, compared to that of a RAW file, it's often harder to remove noise from a JPEG file.
If you like smooth, relatively clean photos but also like the idea of being able to apply a controlled grain-like look to your images, you could consider applying the look you seek on the desktop.
Applications like Adobe Lightroom or Photoshop allow you to start with a clean image and add the degree of digitally generated noise that meets your needs.
Signal To Noise Ratio
The more signal (i.e., data) you have in an image the less noise you'll experience.
The notion of Expose To The Right (ETTR) is, to my mind, the best way forward providing it doesn't cause a slow shutter speed that results in unwanted blur from camera shake or subject movement.
The ETTR technique simply requires that you expose your photo so that the highlight data has been pushed as far to the right hand edge of the histogram as possible.
Just make sure the data doesn’t actually touch that same right hand edge. Doing so would indicate that it's actually spilled over the edge resulting in a loss of potentially important highlight information in your image.
It’s important to understand that exposing to the right will usually result in an image that appears too light and unsharp.
Your job then is to process the image in a RAW Converter, such as Adobe Lightroom or Adobe Camera RAW, until you achieve the desired result.
HDR and Noise Reduction
As the effect of noise is cumulative I apply some slight noise reduction prior to processing a series of images into a HDR composite.
What's more I never add Clarity or Sharpening prior to producing the HDR composite image.
Be careful with how much noise reduction you apply as there's a trade off between noise reduction and sharpness.
You see the more noise reduction you apply the less sharp your image will become.
As sharpening increases noise it's helpful to apply sharpening locally, thereby limiting it to only the areas that need it.
There are various ways to do this one. The use of Masks in Photoshop is a relatively straightforward option.
To actually see the benefit of noise reduction, on screen, it's a good practice to click on the photo to view it at 100% magnification, referred to as actual pixels, in the Develop module in Lightroom prior to applying the noise reduction.
Then, via the use of a before and after preview feature, determine the effectiveness of the noise reduction you’re applying with some certainly.
While it's fine to work with the Luminance noise reduction slider in Lightroom, I'd stay away from the Color Noise reduction slider, other than what Lightroom already applies as a default value.
That is unless you can actually see disagreeable areas of color noise on your photos.
How much noise reduction you apply for general picture taking situations is dependent upon the following considerations:
The camera’s make and model
Exposure technique (e.g., ETTR)
The relative brightness of the scene or subject matter you’re photographing
The ISO at which the image is made
The duration of a long exposure
The image processing application you use
The way you employ that application
Here are the amounts of Luminance Noise Reduction I'd recommend, as starting points, for general picture making situations with cameras dating back to 2010.
ISO 100-400 Luminance Noise Reduction to 20
ISO 800 Luminance Noise Reduction to 30
ISO 1600 Luminance Noise Reduction to 30-50