Absolute Difference

[Marpel]

I have always thought the best way to determine how significant the difference between an image and its derivative (when making very subtle changes in tone/hue etc) was to use Composite with Absolute Difference blend mode. In the black result image, white suggests the difference. At least that is the way I have always done it.

However, I had an image that the only change was Colour Balance. The resultant image visually looked a bit different than the original but only when each was clicked on repeatedly back and forth.

The difference was so slight that I did not know if my eyes were tricking me, so I used the a/n method. The result image from A.D. was completely black, and as I have done in the past, I ran it through Levels and Colour then maxed the Dynamic Range (in the past this has brought out any white, so I know there is some, albeit slight, change). The result was pure black again.

The weird thing is, if I click the Histogram Tool on both images, there is a slight change to its shape, which does suggest there is something different between the two.

So, my questions - Is the a/n method the best to determine extremely slight differences in an image (especially to identify the exact location within the image)? - Is there some floor, below which the A.D. cannot detect a difference? - If A.D. is not the best, which method is?

I read the Help file on Composite, A.D. but it is way above my pay grade, so it was not much help.

[jsachs]

When doing Absolute Difference you may need to increase the Strength slider all the way up to 100% to amplify very small differences. If that doesn't work, try using Autorange (not Levels and Color) on the absolute difference image with black and white thresholds set to 0%. Also, double check that you are not computing the difference between the image and itself.

[Marpel]

Jonathan,
Thanks for the reply. I am sure I was not computing the difference of the image with itself. And I will give a go with your suggestions to see what happens.