1

I have a photo of a person (blurred for privacy) and I used the adjustment brush on the eyes to experiment a bit. While switching the adjustment brush on and off I noticed that not only the eyes changed, but the background was noisier when the adjustment brush was on.
I then did a test with the images shown below.
The left most image (image1) is a JPEG export of the RAW with the adjustment brush turned off, the middle image (image2) is the same image exported to JPEG with the adjustment brush turned on and the right image is the result of placing image2 above image1 in Photoshop and setting the blending mode to divide.

My expectation would be that only the eyes would show a difference (and they do, there are two black spots where the eyes have been adjusted). However, the backgrounds shows a lot of difference as well. This is unexpected.

Could someone explain to me why this is happening?

Photo: enter image description here

Saaru Lindestøkke
  • 8,648
  • 3
  • 39
  • 94
  • 1
    Did you export them with the same sharpening setting? I'm not sure what you did with the adjustment brush, but I did some adjustment with the adjustment brush in Lightroom 4.3 and found no changes outside of the area I used the brush. – AJ Henderson May 19 '13 at 19:40
  • 5
    Try it again with TIFFs. Since JPEG is lossy, there will always be pixel-to-pixel differences, even at very high quality settings, if the pictures aren't completely identical. (Not saying you're not seeing something real, just that JPEG's the wrong test conditions.) –  May 19 '13 at 20:00
  • I'm on Lightroom 3.6. I don't apply any sharpening at export, so the sharpening settings are the same for each frame. I've tried it now with TIFF, but the result is similar to what is obtained with JPEG. Visually I only see the background noise increase in LR. When I export and compare the two TIFF's (or JPEG's) side-by-side there seems no difference, except for the adjusted area. However, dividing one image by the other in PS does give this weird noise. @AJHenderson, when you say found no changes did you try dividing one image by the other as well or did you have another method? – Saaru Lindestøkke May 20 '13 at 00:03
  • @BartArondson - yes, I reproduced your technique exactly. I exported jpeg, made an adjustment with the adjustment brush (darkened an eye) and then exported again. I opened both JPEG files in Photoshop and copied one to a new layer on top of the other. I set blending to difference and had a black screen with an eye showing. – AJ Henderson May 20 '13 at 03:48
  • @AJHenderson I've set the blending mode to divide not difference. I've tried it with difference and then I also don't see any noise. Maybe I'm misunderstanding what the divide blending mode does, but I expected that dividing two identical pixels one would get a white pixel as a result. – Saaru Lindestøkke May 20 '13 at 10:08
  • @BartArondson - ok, I'll try it with divide later. I have the test images still. I forget if I just mislabeled divide as difference in my post since I posted after I had done the test, but I'll verify the result again. – AJ Henderson May 20 '13 at 12:52
  • @BartArondson - I saw your flag to close the question, but I'm not sure that I agree with you since I think it would be easy for others to make the same misapprehensions. However, you do have the ability to vote to close your own question if you really feel it should be. – Joanne C May 21 '13 at 10:36
  • @Bart: The use of difference mode would be appropriate here. Divide has a different purpose, and wouldn't really serve your goal of determining if there is any "difference" between the two photos. I agree with John, I think this question should be left open...but it is up to the community, and if you wish to close, you can always cast your vote. – jrista May 21 '13 at 13:41

1 Answers1

2

I believe most of what you're seeing here is some kind of dithering. It's a common step when writing out processed data where the input data range is much larger than is needed for the final output. It has a way of masking tiny errors that creep in through a processing chain.

I don't think this is anything to worry about. Pretty much every music track produced in the past two decades or so has been dithered during the mix-down process from high resolution masters (e.g. 96 kHz 24-bit per channel) to delivery formats like CD and MP3. You can't hear its presence, though you might hear its absence! Since your ears are more sensitive to small differences than your eyes, I wouldn't worry about any dithering added by Lightroom.

Now, I said "most" of what you're seeing here is the dithering. You're also seeing some artifacts due purely to problems in your test method:

  1. JPEG is the wrong format to use when pixel-peeping unless your aim is to look for differences introduced by the JPEG algorithm. JPEG is lossy: a difference in one pixel is likely to affect the values in 63 of its neighbors, however slightly.

  2. Divide is the wrong tool for detecting image differences. Suffice it to say, there are several ways for it to give false negatives, which is to say that it can indicate that two pixels are the same when they are different. There is also at least one case I can think of where Divide will indicate that two pixels are different when they are the same!

    A much better choice here is the Difference blend mode. It always gives a pure black image when the upper layer is exactly the same as the layer(s) below it.

    You can also use the Apply Image command to do pixel-level forensics on an image.

    Another superior option is ImageMagick's compare tool.

Warren Young
  • 5,224
  • 1
  • 25
  • 35
  • Thank you for your time in pointing out my bad science. I am not concerned on this at all as this small anomaly surpasses the precision of my workflow (I don't have colour calibrated monitors, I don't shoot professionally, etc...). It was just out of curiosity. – Saaru Lindestøkke May 21 '13 at 07:12