Help! There are creepy faces in my photograph!

So you’ve taken a photograph with your phone and you notice something odd in the shadows. You zoom right in and there are creepy faces! Eyes and maybe a nose and a mouth or even a skull, staring out transparently from the shadows. Is it a ghost? Is it a demon? What have I captured? To answer this question we need to talk about three things; ISO, JPEG Compression and Pareidolia.

ISO

Animation showing the difference between a photo at ISO 100 and ISO 3200
The difference between a photo at ISO 100 and ISO 3200

Digital cameras like the one in your phone have ISO settings. The ISO is a measure of light sensitivity. The higher the ISO number, the more sensitive the camera is set to read the light coming into the camera. Most digital cameras have ISO settings ranging from 100 to 3200. By default, your phone or camera will usually automatically pick the best ISO setting for whatever you’re taking a photograph of. If it’s bright sunshine it will use a low ISO number. If it’s dark, or the lighting is poor and there a lots of shadows, it may automatically set the ISO to a very high setting so that it can pick up as much light information as possible.

The problem that occurs is when the ISO setting gets high, you get a lot of “digital noise”. High ISO settings result in randomised light and dark pixels or static, and leads to poorer image quality. Digital noise isn’t noticeable when you’re taking the photo, it gets introduced when the camera actually snaps its picture. When you zoom into the photo you’ll see parts of the photo are made up of a mottled mix of light and dark patches.

JPEG Compression

An example of JPEG compression. In addition to the individual pixels, you can see the 8x8 grid.
An example of JPEG compression. In addition to the individual pixels, you can see the 8×8 grid.

Most people don’t want to fill up their phone or camera with huge files. That is why we have JPEG compression. When you send an image off your phone or camera to your computer, most often it will be a JPEG or JPG file. When your phone or camera stores these files in its memory, it uses JPEG compression to make the file smaller.

How this compression works is a little bit complicated, however, part of what it does is divide the image up into a grid with spaces a few pixels wide and then says “Hey! All these little pixels in this one grid area are pretty much the same colour, so I’ll group them all together and make them one colour so instead of needing to store 10 colours, I only need to store 1 colour!” or “This little grid area seems to be transitioning from one colour to another, but it’s doing it a messy way. I’ll smooth it out a bit!” Since all these little groupings and smoothings are happening in a little square around 8 pixels wide, it’s not really noticeable until you zoom right in.

Now, think back to what we said about high ISO settings and digital noise before. If you’ve got a lot of digital noise in the photograph, when the JPEG compression does its work, it can get a little bit muddled about what it’s seeing in each of its little 8 pixel squares. So the noise gets grouped together or smoothed, turning it into something rather different to what you would see with your own eyes, which leads us on to…

Pareidolia

If you haven’t heard of this term before, do a Google search for “faces in strange places”. That pretty much sums up what pareidolia is all about. When you look at something that’s not actually a face, but you see a face. From the time we are born, human brains are wired to look for faces. It’s basically how babies learn to first recognise their parents. This process is so ingrained that we can end up seeing faces where none exist. In addition, human brains like trying to find patterns in chaos, since working out the pattern of things is how we learn. The human face is a strong pattern we learn to recognise; two eyes, a nose and a mouth.

The Face on Mars - A well-known example of pareidolia.
The Face on Mars – A well-known example of pareidolia.

Now combine all three and we see faces…

So you’ve taken a photo, it was dark and your camera or phone has introduced a lot of high ISO digital noise (or what we could call “chaos”), which has then been concentrated and smoothed into mottled patterns in the shadows of our photograph by the JPEG Compression. You see something that looks a little odd in our photo. You enlarge the photo and zoom in… Those blotches there, there are two dark ones near each other that could be eyes… That patch below could be a mouth… Is that a face?!

Congratulations! Your brain’s pattern and facial recognition are working wonderfully! Unfortunately, it’s not a real face. It’s just something your camera has added into the photograph as it was processing it. It’s not a ghost or a demon, it’s just your brain trying to make sense of some reorganised pixelated randomness.

An example of pareidolia in a high ISO JPEG image. How many faces do you see?
An example of pareidolia in a high ISO JPEG image. The shape of what appears to be a skull is partially visible in the image noise in this zoomed-in copy of the image.

Leave a Reply