Low-light Images often have noise issues, particularly in the dark areas.
So you’ve decided to take night time photographs. But the light is so low you’re worried about noise. You want the image sharp and the blacks to be black. And noise reduction reduces sharpness, so it’s going to be a problem. (Noise is always a problem with low-light images.)
In these situations you should always shoot at your camera’s lowest ISO setting and increase the duration of the exposure, right? Well, maybe not. The counterintuitive solution might be to increase the ISO and take multiple images of the same subject.
Single image at ISO 1600 and cropped showing lots of noise.
Increasing ISO Increases Noise
Hang on a minute. If increasing the ISO increases noise, how will reshooting the scene at a higher ISO improve low-light performance? Won’t it just increase the noise?
Conventional approaches to noise reduction reduce the sharpness of the image, making them soft or blurred. And blending multiple images won’t reduce the noise. Or will it?
Cropped image of stacked and blended images at ISO 1600.
Some Low Light Images Need Short Exposures
The other potential problem is that long exposures don’t always work. Some night photography involves taking images of objects that move, and shorter exposures can help control that movement.
Low-light image at ISO 1600 (single image).
Photoshop to the Rescue
Adobe Photoshop has powerful tools that let you blend multiple images, but most of these blend modes won’t help. However, there is a way to blend images in Photoshop to reduce noise. The key to shooting with a higher ISO to improve low light performance is to shoot multiple images of the same scene using the same settings (i.e. white balance, focus, aperture and shutter speed).
While the technique is fairly straightforward, it does take some discipline.
Stacked and blended image of six images at ISO 3200.
The key is understanding what causes noise. In general there are two types of image noise – chromatic and luminance. Chromatic noise is color aberrations where there are none, while luminance noise is variance in light levels where there is none. Both are instances where the sensor has registered some data that isn’t there. (It’s common in sensitive electronic equipment such as digital sensors.)
If you take a single image, the noise is part of that image. But if you take a second image in the same location, chances are the noise won’t be in the same spot (unless you have a bad sensor). The noise actually moves around.
If you think about exposure in simple terms, it’s the amount of light that hits the sensor or film. Changing the aperture from f/4 to f/2.8 doubles the amount light hitting the sensor. Similarly, if you decrease the ISO from ISO 400 to ISO 200 you need twice as much light for the same image.
But taking a properly exposed image and then blending a second properly exposed image doesn’t actually improve your exposure. Is there another way?
Cityscape at ISO 400 (single image).
The short answer is “Yes”. This technique relies on the fact that the noise moves around on the sensor. You can take one image at ISO 400, or you can take two images at ISO 800. As long as the total length of exposure (assuming the same aperture) is similar, while the noise will have gone up you’ll effectively have the same image. That’s because you’re simply doubling the amount of light on the sensor at ISO 800, and there’s a proportional increase in sensitivity. Similarly, if you take four images at ISO 1600 you should end up with the same exposure.
But what if I use ten images?
Ten images at 1600 blended.
You may be thinking, “So what? At ISO 1600 I now have ten noisier images than my image at ISO 400. How does it improve my camera’s performance and reduce noise?”
The answer is to stack them, and then blend them together using a particular method in Photoshop.
Multiple Images Can Overcome Noise
By importing the images as a stack of layers in Photoshop and blending the stack together, you can improve your image quality. Using my earlier example, if you use ten images at ISO 1600 you effectively have an image comparable to an ISO 400 image.
Single image at ISO 400 with a tight crop.
Ten stacked and blended images at ISO 1600.
As I said earlier, while this technique is pretty straightforward it’s not exactly obvious. Following the steps is critical.
You don’t get extra resolution. But you do get less noise, and the image seems sharper.
Pick a subject (not the night sky) that’s under low-light conditions and take multiple images of the same perspective at a higher ISO than you’d normally use – 1600, 3200 or even 6400. (Don’t use ISO values in the extended range because they’re not native to your camera’s sensor.)
Manually set your focus so it doesn’t change between shots. You should either shoot the images in RAW format or make sure all the White Balance settings are the same. Using RAW lets you edit the white balance later, but fixing it before you taking the shots will also address the issue. Take one image at a lower ISO value (probably with a long shutter) and many at the high ISO value. This will allow you to compare the results.
Step 1: Ensure all the images have the same White Balance. (You can correct RAW images together if you shoot in RAW.)
Use a RAW Processor to match the White Balance and Exposure
Step 2: Import the images as layers into Photoshop. (Bridge and Lightroom both can do this).
Step 3: Highlight all the layers in Photoshop.
Step 4: From the Edit menu, choose Auto-Align Layers.
Align the images.
Auto-Align works well.
Step 5: Crop the image to eliminate any missing parts of the image.
Sixth step: Highlight all the layers, and then from the Layers menu choose Convert to Smart Object.
Convert the Layers into a Smart Object
Step 7: Click on the Smart object, and from the Layers menu choose Smart Objects -> Stack Mode -> Median.
Use the median stack mode to blend the layers.
Step 8: Look at the result. (It’s pretty dramatic.)
Single Exposure at ISO 1600.
Stacked images at ISO 6400.
Cropped image at ISO 1600.
Cropped image at ISO 6400 (stacked).
What just happened?
Photoshop blended all the (now aligned) layers together, looked at where most of the images showed the same data and decided that data was correct. It then discarded any data that didn’t match. Because chromatic and luminance noise varies from image to image, blending multiple images like this eliminates the pixels showing incorrect color or luminance.
Stacked and Blended Evening Image
As you can see, it significantly reduces noise without losing sharpness or introducing unwanted artifacts. So the next time you’re shooting in low light, why not give this technique a try?