Quite frankly you appear to be overlooking some factors here and you seem to be using the term »Filter« for what is actually called »Blend Mode«.
If there’s a (fringe) pixel of, say, 128/128/128 Photoshop can quickly calculate the result of that with an underlying pixel’s colors with a certain Blend Mode – but whether it is supposed to be a black pixel with 50% transparency or a solid gray pixel is less easy to determine.
So many designers have to work with logos & whatnot that are on a white or otherwise colored background
If they have to do that it’s probably because customers provide inferior, inappropriate logo-material instead of the vector-data or at least layered files, which one would assume usually do exist somewhere.
Blend Mode is correct. Thanks.
In the real world, you know, you have to deal with people and they don't always have vector files handy--and, if they're not handy then they may not look for them at their letterhead vendor, local newspaper, or whoever might have such a file. That's really missing the point of the post here.
That's really missing the point of the post here.
I feel comfortable going off on tangents in these Fora.
Moreover it seems justified when you ask Adobe to solve problems caused by the incompetent, unwilling and those who would have the knowledge and the willingness to apply it but are kept from doing so by, for example, unreasonable deadlines.
But back to your request:
The way I see it You make an observation (Blending Modes like multiplying takes care of certain problems under some circumstances) but draw a conclusion (that this could be used to create useable transparency) that I think is inadmissible for many cases.
If you talk about purely black (or basically one-color) logos where the only problem is anti-aliased edge-pixels one can easily enough lift the transparency of the appropriate channel and use it as a Layer Mask for a Solid Color Layer – this could be automated with Actions or Scripts.
But for Logos with (possibly overlapping) elements of different colors, gradients, soft edges etc. the matter is less easily analyzed – wherefore I think automation is likely to produce »incorrect« results for intersections etc.
Then again I may be wrong …
I've used Multiply or Darken to take a logo that is imbedded on a white bg and overlay it on a gradient or textured bg. It works very well and is, for all practical purposes, making the logo act as if it is on a transparent bg. So, I'm just saying that there may be a fairly simple way to use this to actually take that bg out.
It's like a green-screen that isn't just a screen. It would make the logo more versatile.
You're right that logos that are on an unevenly colored bg wouldn't work as well, if at all, but the majority of cases out there are gifs or jpgs that have a white bg.
I'm using logos as the main example but I'm sure that (if this works) the creative types out there would find other uses for it.
As to the tangential part of this discussion--I understand your point but I have extracted logos from bgs and/or recreated them far too often and taken much to much time in doing so in order to ensure the highest quality end-result. The client is not necessarily incompetent or unwilling. I don't expect all small business owners to understand or even care about the technical issues that designers must deal with regularly--they do their job and we do ours. We can only educate them along the way so that things get easier and more standard but that is a slow and, I suspect, endless process.
It seems that there would be a relatively simple tool that could use these as a way of removing the background of an img
…there probably is. Look into your blending options, and the "Blend-if" sliders.
These will actually make the areas transparent, rather than 'blending' them. Option clicking them allows graduated drop off. For high end work these produce too crude a result, but for clipping out a background that kind of thing, they are good enough. You can also clip based on the color channels