nomore wrote: makc wrote:
Okay... To end up this discussion, I've made an image where you can actually see the bee itself
, but it seems victorengel
was first to achieve this level of quality for bee shape...
Here are steps to re-produce image:
1. mix 50/50 before-after image, and filter it (Gaussian blur, 0.5 pix)
2. apply S-shaped Curves transform to bright-up bee (result
3. put the result on top and swith to "lighter" layer mode (to remove lots of jpeg artifacts)
4. merge it with underlying layer copy to get it back to "normal"
5. now put the layer in the difference mode. it will be mostly too black to see anything, but don't be scared
6. select near-pitch-black color range (preview ON, GrayScale)
voila - there the bee is.
There is no logic to this process.
It is only a means to an end.
Why don't you just add a bee to the image?
That would bee just as scientific as this approach.
Disagree. What's illogical?
1 - we need a background w/o unknown something
2 - we need to make pixels affected by something
more "outstanding"; yes, the S-transform did affected background, too, but - there's step 3 - we keep only enhanced pixels affected by something
(and, unavoidably, light jpeg artifacts)
4 - we just store the result of step 3 in separate layer
5 - by substracting background we see how each pixel is affected
6 - we select least-affected pixels that match background the best and, therefore, yield near-pitch-black in step 5
what we see on the final image is how close to black (zero) is the difference of images with something
and without it: white marks pixels that going to be selected (which became black after step 5), and black marks pixels that ain't going to be selected - i.e., too different from black - that's merely the way photoshop works.
the fact that it shapes the bee
is clearly explained by the theory about flash light reflection from the bee, and, on that basis, we can
conclude that something
is actually the light reflected from the bee.
what in God's name is wrong with that? the only assumption I took in the step 2 (and then used it in 3), is that something
makes pixels lighter than background. but, I think it's pretty obvious without any shopping at all.