On Tue, 16 Sep 2014 09:12:12 +0100, Martin Brown
wrote:

On 16/09/2014 05:37, John McWilliams wrote:
On 9/15/14 PDT, 7:07 PM, Floyd L. Davidson wrote:
Savageduck wrote:

I got what Floyd was talking about when he was talking
of high pass sharpening, and reversing it by applying
the corresponding reverse parameter blur. However, he
also stated above, "UnSharpMask is not reversible". My
point addressed the fact that for some of us, that is
not an entirely valid statement.

IN an environment that supports a saved original copy so that all edits
are non-destructive then that is true.

You can't reverse a process if you have never executed it. If you make
a copy of an image and edit it, you cant reverse the process of
editing by just hauling out your original image. The edited version of
the image remains edited and in most cases there is nothing you can do
to reverse it.

The water is muddied by the several applications which make use of a
sidecar file of some kind to preserve a list of edits which are only
executed when the image file is exported from the editing environment.
Modifying a sidecar file by deleting an editing process from it does
not make that process reversible. It merely makes that process
asthough it never was.

But the mathematics of blurring and of unsharp masking make it
irreversible if you are only given just the processed image.
(and not some hybrid Photoshop workflow encapsulated format)

That is in fact a valid statement. The USM function is
not reversible.

That isn't a opinion, it's a fact.

For one definition of the word!

This seems to have degenerated into a heated and utterly pointless
semantic argument over the meaning of "reversible" that is in conflict
with normal image processing and mathematical parlance.

A mathematical operation is reversible if a strict inverse function
exists that can exactly get you back to where you started.

Any operation can be made non-destructive simply by saving a copy of the
original before applying the irreversible filter but that isn't very
interesting. Some packages do support this sort of safe workflow.
--

Regards,

Eric Stevens