What's a more efficient way to deep-etch?


Working in Photoshop CS5, I deep-etched an image using a tablet. Where I thought I'd really gotten it quite clean and feathered the selection by 2x before inverting the selection and deleting the background, it in fact left a fair amount of background colour tracing around the edges.

What I would usually do is create a layer mask and use the brush on black to remove what's left. I was wondering if you perhaps know a faster, more efficient way of doing this?

10/9/2012 10:09:00 AM

What I usually do is zoom in to about 500% and manually use the polygonal lasso tool (1px feather). I think of it as the analog to carefully cutting something out with scissors.

One can add and subtract from the selection using combinations of ctrl, alt and shift keys (or the mac equivalents). If you flick the tool off the edges of the document, the view will scroll and the scroll speed is adjustable in preferences IIRC. If you are working on something complex or need to go get lunch etc, an easy way to save the selection is to create a temporary layer and flood-fill it with black. You can recall the selection by ctrl clicking the layer and then hiding it so you can see. Once I have the whole thing set up, I save the mask as a (alpha) channel.

Depending on your desired use, the selection can be converted to a path etc.

The polygonal tool works better than magnet, magic and free-form tools IMO, and it buffers shaky hands. I rarely use brush tools for masks.


If you ctrl+click a layer's thumbnail image in the Layers Palette you will get a selection that conforms to the transparency (alpha) of the layer in some manner (if there are no transparent portions, you wind up with a selection the same size as the document).

Experiment a little: make 2 new layers, make a rectangular selection, fill it with black. repeat on the second layer (but move the selection box). Deselect. Now ctrl+click the thumbnail image of the layer in the layers palette, then hold ctrl+shift and hover the mouse over the thumbnail of the second unselected layer. The mouse icon turns to a box with a plus in it. If you click on the thumbnail, you will be adding to the selection.

ctrl+alt "= subtract"; ctrl+alt+shift = ??profit?? (XOR or something)

My primary use for this is to mask off an area of a photo for printing purposes. An alpha channel is a way to save the mask in a non-destructive way. When I place it in layout software, I see the full photo, then I set a clipping path or mask within the layout software, referencing the alpha channel as the source to use. The primary advantage of this method is that the "background color" can be set in software and need not be known at the masking stage. If you are merely using "white out" on the original image, you will need to mask it again if you want to reuse it on black and you certainly will never be able to add back in the top of Grandmom's head you accidentally painted over...

I am working within a long-maintained library of images and I usually save them as CMYK TIFF with alpha (when applicable) mainly because TIFF is software agnostic.

I would only finalize the alpha and save the tiff at the end. Until then I would work with layer and PSD files (and save that source so I can return to it if needed).

Note that I use alphas a lot because of the nature of items I work with (photos of artifacts). I have had some printers complain they aren't paths and for them I make adjustments, but the paths are often too "squirrely" and clip off too much. Too many alpha channels being used as clipping paths in a single document can/will cause problems when "RIPing" plates/separations.

10/9/2012 7:50:00 PM