Talk:Percentage difference between images: Difference between revisions

m
 
Line 43:
:: I actually liked the example (REBOL) that produced the differenced image. That would have been a bit more interesting. Unfortunately the author didn't explain what the code that's enhancing the contrast is doing at the pixel level so not too many of us could actually repeat it if we had to code it --[[User:Dgamey|Dgamey]] 22:12, 1 September 2011 (UTC)
::: Simpliest way to do contrast is find min and max pixel values, then rescale all colors such that min is 0 and max is 1 (or 255). If you want to be fancier, you can also gamma correct it, which is a little questionable. --[[User:Ledrug|Ledrug]] 02:10, 2 September 2011 (UTC)
:::: I'd thought of that, but one white and black pixel forces something else. Scaling by some number of std dev from mean perhaps. Maybe based on luminance. I was thinking that if a task to produce a difference image and intensified one would need to have a comparable methodology. Or produce the diff image and explain the intensified one. Something like that. But I digress. --[[User:Dgamey|Dgamey]] 13:11, 2 September 2011 (UTC)
Anonymous user