How To Remove Noise And People From Your Shots. Image Averaging Technique: No Need of ND Filter.

ShadowHow To Remove Noise And People From Your Shots. Image Averaging Technique: No Need of ND Filter

Ever since content capturing devices have gone digital, artists can enjoy a new ultimate level of creative freedom. Post-processing of digital content opened up a whole new way of working with and manipulating content that previously was difficult to accomplish. There are literally limitless options of what you can now achieve. Unfortunately, not everyone is aware of these possibilities and so they might struggle taking their content to the next level. So, today we will help you master one extremely powerful technique that can be used for photo-processing. It is called image averaging.

Image averaging does exactly what its name suggests: analyzes multiple images and then combines them, keeping unchanging parts but removing things that have moved. But what can it be used for? Well, a whole number of things actually: removing noise, for example. So, let’s take a more detailed look.

Getting long exposure effect on water and sky with short shutter speeds

Normally, if you want to capture that nice smooth water surface for your photo you need to use an exposure time of at least 2 seconds, sometimes even longer. This is not a problem at dusk or night, but what if you’re shooting under the day’s bright sun? You can try to drag out shutter-speed by boosting ISO to a minimum and closing the aperture all the way. But this respectively will cause you to lose small dynamic range and sharpness due to light diffraction. And even this might not be enough to get desired shutter speed. That is why many people use Natural density filters. But those are quite expensive (about $80) when good and detrimental to image quality when cheap, causing color-casting, sharpness and contrast loss - and that doesn't even speak to all the additional fuss it will lead to, such as having to take the ND filter with you, screwing it on and off, cleaning it, etc. But this is the only way to get the desired effect, right? Well, maybe this was true for the film days but not in the era of digital mediums. Today you can create this effect in post-processing without it looking fake or unnatural. In fact, it will look almost exactly the same as long exposure but without all the above-mentioned quality degradation.
So, what do you need to do?

Water before averaging

Water after averaging

1. First, when shooting, you need to place your camera firmly in one place, switch it to fully manual mode, lock the white balance and focus, then switch the camera to burst mode. Now, sometimes you can even get away with shooting hand-held (something that’s completely impossible for typical long-exposure photography), but this adds more work in post-processing and significantly decreases your chances of getting a great result, so I don’t recommend doing it unless you really can’t find any firm surface to place your camera on.

2. Having made sure the burst mode is selected, you need to press your shutter and take several pictures. Be careful not to apply to much pressure on camera to cause it to move, ideally you want to use remote release. How many pictures you want to take depends on how strong you want the final effect to be. You can calculate the perceptual shutter speed you'll get in the end by multiplying number of images by the shutter speed of each.

Perceptual shutter speed = number of pictures * shutter speed of each picture

In the example above, I had 12 pictures with 6 sec. exposure each, so the water on final image looks like it would with 1.2 minute single exposure. I couldn't get the single exposure in that case for couple of reasons:

  • I didn't have my tripod, so the camera was placed on pavement of an overpass and every now and then huge truck or bus would drive by causing vibrations and ruining the shot
  • boats were swimming by every 15 seconds, so they would've left traces on water that I did not like
  • my camera has a minimum shutter speed of 30 seconds and I didn't have my remote trigger that evening

So, as you can see there are quite a few reasons why this method works where a single long exposure doesn't. Back to the photo, even with relatively short 6 second exposures I had to pick my time windows very carefully (no boat, no bus).

3. Open your files as layers in Photoshop, previously having made and synced basic adjustments in Lightroom or Camera Raw. Also you should check for and delete any blurred images, if there are any. For the picture above, even with carefully picked windows I still had to delete 2 pictures out of initial 14 because they were blurred.

4. If you shot your images on a steady tripod, proceed straight to action 5. If not, go to Edit => Auto-Align Layers. You can leave everything in dialog box at default settings. Now Photoshop will attempt to align each image, so it matches perfectly to all the others. This will take some time, depending on your computer's processor and the number of pictures.

5. When Photoshop is done, make sure all layers are selected once again, after go to Layers => Smart Objects => Convert to Smart Object. Wait until Photoshop finishes.

6. Lastly, go to Layers => Smart Objects => Stack Mode => Mean

This last step will take the most time for Photoshop to complete but when it’s done – voila! – you have an image that looks like it was captured with one long exposure. You might have to crop edges, if you had to do step 4, but other than that it’s done. As you can see, this method is very simple for you to follow, but places rather high loads on your processor, so it pays off to have good hardware.

Now you can make whatever final adjustments you need and share your photo!

 Also, to make you life simpler you can get a free Photoshop action from us that will do all the steps above, without the need for you to go through this process manually each time.

If you don’t have a Photoshop, you can do the same thing using command line on Windows, just download the free software pack from first.

The command will look as follows:
C:\Windows>convert -average “path to the folder with your files” “path to the destination of the final image

So, if your files are on a Desktop folder called Sea and you want the final image to be exported to the Desktop, the command will look like this:

C:\Windows>convert -average “C:\Users\Denis\Desktop\Sea\*.jpg” “C:\Users\Denis\Desktop\output.jpg”

Unfortunately, ImageMagick does not align images, so you will have to make sure you shoot on a tripod with minimum disturbance.

Reducing high ISO noise

The same technique can be also used for reducing random noise in your image - high ISO noise falls to this group. If you don’t know it yet, noise caused by increasing the ISO setting on your camera has to do with the fact that with the increased sensitivity of the sensor comes an increased number of artifacts captured. In other words, the signal to noise ratio (SNR) decreases. But since the noise is things that are not actually there, they change their position from image to image, and the signal remains constant from image to image. So, by averaging the result you can keep everything static in the frame (your subject) and get rid of everything that varies from image to image (noise).

For a more technical explanation, let’s assume that we have a set of five images with the exact same scene. The value of any given pixel may then have 5 different values among 5 images, for instance:
Red (115, 120, 130, 190, 191)
 Green (150, 166, 169, 200, 210)
 Blue (200, 205, 209, 211, 220)

By performing averaging, the software will analyze these values and in the output file a median value for each pixel will be recorded. So, for our example it will be 130, 169, 209 for Red, Green and Blue respectively.

This technique is used a lot in astrophotography but nothing stops you from making use of it in different scenarios as well. It will work perfectly when:

  • You’re limited with exposure time. For instance, if you don’t have a tripod or you have just a small window of opportunity to get the scene (such as boats floating by every 30 seconds)
  • You want to freeze motion in low details while still keeping low noise in high details. Example of this can be tree leaves in a foreground and starry night sky in the background
  • You want to reduce shadow noise to be able to push them up during post-processing, even if image is shot at low ISO

Keep in mind, however, that this method doesn’t work on noise caused by long exposure times (aka hot pixels) as those are caused by your sensor’s overheating and are not random in nature.

Noise comparison before/after averaging

Noise comparison before/after averaging, 200% crop

So, to decrease the amount of noise, you will need to do all the same steps as before. But this time when taking images you can estimate the effect of noise reduction as a square root of the number of exposures

SNR improvement = √number of pictures

So, if you take 4 pictures, you will get 2 times less noise, but with 10 pictures you will only reduce your noise 3.2 times.

As for perceptual ISO, it can be determined by taking the ISO of each picture divided by the number of pictures.

Perceptual ISO = ISO/number of pictures

In the example above we had 12 pictures, each with an ISO 400. So, by averaging them we got the same result as we would shooting with ISO 33. Or in other words, we decreased noise 3.46 times. If you wonder why our resulting ISO 33 is smaller than the starting 400 more times than 3.46, this is because noise doesn't increase in a direct linear proportion with ISO.

Also, this time when you get to step 6, instead of Mean choose Median stack mode. This will work better and produce less artifacts for noise reduction. Mean will work too, but the difference is that Mean leaves all the data in place resulting in images that look more like a long exposure image and Median eliminates varying factors all together. So, it's better for Noise reduction.

Image Averaging

Output image

You might also have noticed that in our output sample image the moon is gone. This happened because it actually was moving the whole time and the algorithms got rid of it (remember, they do so with every changing thing).

But don't worry, with a simple masking we can bring it right back. For this you need to load one of the initial images as a separate layer into Photoshop and then mask the whole layer out, and paint with a white brush on the areas to mask and reveal the subject you want to keep. Alternatively, you can just use the magic wand to select the moon and fill that area of mask with white, if you don't feel like painting by hand.

Image Averaging masking

Brining back the Moon

Image Averaging final

Final image

Here it is. Just be careful not to go over edges of the moon as this will bring back all that noise you spent precious time getting rid of. No worries about the moon itself though, because it's very bright none of the noise will be visible in that area.

Removing tourists and other moving objects

Now, if you’re a travel photographer or just a person who enjoys traveling and wants pictures of attractions – and not other tourists wandering all over the place – you will fall in love with this effect of image averaging. In fact, it is probably the most amazing one of all three! And just as simple. The process is the same as it was with noise reduction. But this time there’s no rule or formula as for how many pictures to take. Basically, you need to have enough of them, so that each spot isn’t covered by a person on at least one picture. In other words, you need people not to overlap in any given place on all of the pictures.

After that, bring images into Photoshop and perform all the actions described above and make sure the stacking mode is Median, not Mean.
And there you go, you wiped dozens of people out of existence. Makes you feel pretty powerful, doesn't it?


About the Author

Denis ProtopopovLandscape, lifestyle and product photographer for the past 3 yearsView all posts by Denis Protopopov →

Leave a Reply

Yours FREE

 450 Pro Lightroom Presets Bundle

Subscribe to receive MediaMaker.Tips updates directly into your mailbox