Destabilized Sensor Reflections

[Mick West]

The thin blue squiggle in the image above is a Destabilized Sensor Reflection. A relatively new phenomenon in photography.

A sensor reflection in a phone camera occurs when light reflects off the sensor (the digital "film" of the camera), then back through the lens, off the flat glass covering the lens assembly, and then back down to the sensor to get recorded. These reflections appear diagonally opposite the light source in the image. It's essentially a type of lens flare, and frequently mistaken for a UFO.

Modern phones have amazing image stabilization.So if you take a long exposure, say 1-2 seconds, the phone will generally do a great job of taking a stable photo which normally would have come out a blurry mess. The above image is 1.1 seconds, but still looks pretty sharp.

Destabilized Sensor Reflections (DSRs) can also show up in video, particularly when zoomed in. The dancing dot here is a reflection of the green light in the pool, but it moves so erratically while the pool light is fairly steady and smooth, so it's hard to recognize.


Stabilization works by shifting the image. When you move the camera a little then everything in the image moves together in the opposite direction. So all the stabilization has to do is move it back the right amount. As you can see it does it really well. However, the sensor reflection moves in the opposite direction, so the stabilization actually makes things worse.

Here's an example with multiple lights where I deliberately rotated the camera to the right, but the image of the socks is perfectly clear.



Example of social media:

Source: https://twitter.com/Saw_a_UFO/status/1602057571417874434

 

[deirdre]

so that twitter squiggle isa reflection of...? the windshield of that boat? (ps that boat looks like a crossbreed of a litebrite and a lego set. they should make legos that glow like that, kid would love em)

 

[AmberRobot]

So, is the stabilization basically 2-d charge shuffling on the sensor itself between pixels? What’s the input signal? Accelerometers?

 

[Mick West]

so that twitter squiggle isa reflection of...? the windshield of that boat? (ps that boat looks like a crossbreed of a litebrite and a lego set. they should make legos that glow like that, kid would love em)

It's the blue light on the front starboard edge.

 

[Mick West]

So, is the stabilization basically 2-d charge shuffling on the sensor itself between pixels? What’s the input signal? Accelerometers?

Stabilization is done in hardware by moving the lens or the sensor and in software by simply offsetting the image. The input would be gyroscopes (changes in direction), but I suspect it's also doing some real-time image analysis, based on how it handles the "Venus behind a tree" case:
Source: https://www.youtube.com/watch?v=DB4WFkvjaBg

(Notice that also has the sensor reflections of the lights unstabilized in the first two seconds)

 

[Z.W. Wolf]

This video caught my eye awhile ago. It took some time for me to figure out most of what was going on here.


The stabilization is the factor I didn't know about, which explains the odd movements of the flares, which baffled me. It was that odd movement that distracted me from the more straightforward problem of which ground light is responsible for which flare.

The odd movements of the flares didn't seem to correlate with the movement of the camera... so these flares "couldn't" be caused by the ground lights. That was my thinking.

I thought there might be some object with lights on it that was moving around in the neighborhood of the camera . But what could it be?

Note how the flare from the bright engine exhaust is sometimes below the rocket and sometimes above but always moving in sync with the other flares. This finally convinced me to just treat these flares as simple "filter flares" - (or in this case sensor flares) - and work out which ground light was responsible for which flare. A pretty straight forward process. I just had to accept that the movement was unexplainable to me. In fact I had intended to start a new thread on that mystery, (but never got around to it).

If you want to figure out which ground light is the source of which flare, the key is that the flares from the ground lights are "flipped" both left to right and bottom to top. With that in mind you can work out which ground light is the source of which flare pretty easily.


I imagine that if the sky had been clear and there were no trees this video could have been named: Giant Supersonic UFO's Follow Artemis I

We talked about filter flares here: https://www.metabunk.org/threads/ex...rida-offset-lens-reflection.6932/#post-170311

 

[Z.W. Wolf]

I'm wondering whether the source of the flare can sometimes be a reflection off the external surface of the lens onto the the internal surface of the flat glass cover, rather than from off the sensor.

I don't know how, or how well a phone camera lens is typically coated on that surface (or if all).

 

[Mick West]

I'm wondering whether the source of the flare can sometimes be a reflection off the external surface of the lens onto the the internal surface of the flat glass cover, rather than from off the sensor.

That would be a regular lens flare, as it's reflecting from a curved surface.

The reflection off the sensor goes through the lens assembly three times (in, out, in again), which it what results in it being a mirror image.

 

[Mick West]

A fun video example that got misidentified as ball lighting.

Source: https://www.youtube.com/watch?v=jwCDxCmW5pg

Here's an echo version, the reflection comes from the bright light at the bottom, which has been stabilized into a smooth curve. The destabilized reflection performs a typical random walk.

 

[Mick West]

Thinking about the name, "unstabilized" suggests a lack of stabilization, so might confuse people into thinking the camera is unstabilized. It's actually the stabilization that's a key part of the final image, stabilizing most of it, and doubling the motion of the reflection.

So I think I'll use "Destabilized Sensor Reflection"

I also considered "Antistabilized" to incorporate the idea of stabilization feedback.

 

[Ravi]

I'm wondering whether the source of the flare can sometimes be a reflection off the external surface of the lens onto the the internal surface of the flat glass cover, rather than from off the sensor.

I don't know how, or how well a phone camera lens is typically coated on that surface (or if all).

These are called "ghosts" in optics. It was indeed more an issue in the days before AR coatings.
The main back reflection from the sensor is tricky to get rid of, even in modern designs, because of the nature of the material (Silicon).

 

[Mick West]

Just adding another example. Not super exciting, but again a big disconnect between the camera motion and the reflection motion.

With this "orb" on the far left: