Optical versus digital reflections

[DaveC426913]

This is allegedly (part of) a bona fide photo of a critter, staring at its own reflection in the water.

I am skeptical. I am almost certain that the highlight in the reflected image is provably digitally reflected. To-wit:

If the virtual image in the water were a true reflection off the water (as opposed to Photoshopped), the highlight in the virtual image should be on top of the spherical eyeball.

I can't seem to find any images or diagrams online to confirm it.

 

[jedishrfu]

It's really hard to say with an image so small. I tried to blow it up but it gets too grainy to see any oddities in the pixel layout.

 

[DaveC426913]

Never mind that. I can't really get into the other evidence without possibly violating copyright.

Just look a the position of the highlight in the eye. It seems to me a reflection of a sphere should also have the highlight on the top, not the bottom.

 

[jedishrfu]

Maybe you could do an experiment with a mirror and a shiny ball where you could mark a spot and see what happened.

 

[DaveC426913]

Maybe you could do an experiment with a mirror and a shiny ball where you could mark a spot and see what happened.

Not 'mark a spot' - got to be a reflection of the sky.

I will check it, but it's bedtime.

What frustrates me is that I cannot find any diagrams online showing the geometry of a landscape versus its virtual image:

 

[jedishrfu]

Yeah, I had that problem too.

 

[A.T.]

, the highlight in the virtual image should be on top of the spherical eyeball.

I don't think so. Draw a diagram of the virtual sphere and of the virtual light source. The highlight is were the spectacular reflection occurs. It's not on the exact corresponding spot of the virtual sphere, as on the real sphere. But it can be on the the mirror-far-side for both.

 

[sophiecentaur]

It seems to me a reflection of a sphere should also have the highlight on the top, not the bottom.

I think I agree with that. If you were laying on the water, where the reflected rays hit it, the highlight would appear lower down (nearer the ground) with respect to the solid eye. Lateral inversion when observing the mirror image would, I think, put the image of the highlight on the image of the eyeball (i.e. nearer the creature's nose).
A reflected image, even in a plane mirror, moves 'with the observer' and the displacement would be higher with a convex surface.

 

[DaveC426913]

OK, after even more detailed analysis, I've reversed my verdict.

I found the artist's gallery. Here is the picture in question:
http://radwildlife.com/en/red-squirrel/#jp-carousel-5346

My original analysis of the primary and reflected image led me to conclude that the foreground (squirrel and vine) reflection was a digital duplicate of the primary. (Literally overlaying them in Photoshop.)

My rationale: An image reflected in a surface like this is not identical to the primary - there is a slight parallax, due to the fact that the reflected angle is slightly more oblique than the primary angle - leading to foreground and midground objects not aligning the same. The reflected image "sees" the subject slightly more "from the front "than the primary".


Further analysis: At first, it appeared that there was no such rotation of the image, but I was analyzing individual pieces of the foreground. Upon more careful analysis of the entire image as a whole, I've concluded that that slight rotation is indeed there.

This is more compelling evidence than the highlight on the eyeball.

Thus, I've reversed my verdict to conclude that this is indeed a bona fide optical reflection.

Steps to replicate the test:

1. Copy the full-size into Photoshop.
2. Duplicate the image to a new layer.
3. Flip layer2 vertically.
4. Align the subject on both layers as best as possible.
5. Rapidly turn layer 2 on and off.

By doing this, you get a flicker effect whereby the original and flipped layers alternate.
It is suddenly very apparent that there is an induced rotation of the reflected subject (a "pitch" rotation) by just a degree or less. This is exactly what one should expect from a reflected image. It should be rotated - pitched - a little because of the steeper angle of the reflected path. It is effectively impossible to simulate this reflection digitally.


I wish I could post my analysis without violating copyright - and I wish I could make animated gifs to prove my point.

 

[A.T.]

It seems to me a reflection of a sphere should also have the highlight on the top, not the bottom.

I think I agree with that.

You should both take a shiny sphere, a mirror, and just try it out.

 

[DaveC426913]

You should both take a shiny sphere, a mirror, and just try it out.

I've decided that my whole subject analysis, above, is more compelling than the highlight analysis.
I wish I could post it. It's quite obvious.

 

[DaveC426913]

Ah, I'll take the risk. I'll pull it down if you make me.

Here's the animated gif. It is now quite easy to see that the reflected image is rotated a fraction of a degree from the primary image - exactly as one would expect from a bona fide photo. You can't fake this in Photoshop.

 

[sophiecentaur]

You should both take a shiny sphere, a mirror, and just try it out.

Ah yes (with the help of a spoon) the highlight moves less and not more as you move. Highlight position wouldn’t change enough to offset by a noticeable amount so reality is believable - clinched by the parallax argument.

 

[A.T.]

Ah yes (with the help of a spoon) the highlight moves less and not more as you move. Highlight position wouldn’t change enough to offset by a noticeable amount so reality is believable - clinched by the parallax argument.

Yes, you have two effects:

- The highlight in the mirror image is in a different physical location of the sphere (closer to the nose).

- The mirror image shows the sphere from a different direction (more from the nose side).

For the visual impression (relative position of the highlight on the dark spot) those two effects work in opposite directions. So it's barley noticeable, but it looks slightly different if you look closely.