This is a fairly basic example, in tutorial format. It's intended to let people step through the actual edit, understand what the result of each step is supposed to be (and why), and ultimately get the same end result.

spoiler(note: text in spoilers is extra detail and notes; skip if your attention span is stretched)

More advanced discussion is in forum #3342.

Here's a simple edit runthrough: post #106627 (raw: post #106635). It's pretty easy; it's AM screening, but the screening is high-resolution, and it's a light-colored image (descreening is harder in darks).

The first thing I do is to straighten it, since there's a solid horizontal line to reference. (See the last paragraph of forum #2588 for info.) If you're following along at home and just want the same adjustment, it's 0.1 degree clockwise.



#1: Raw in the background layer.

#2: Copy the background layer (Control-J). Name this layer "light". Run greycstoration with: -p 0.6 -a 0.6 -alpha 1 -sigma 1 -gauss 0.8 -interp 1 -fast -alt. When I'm tuning greyc, this is where I start. The .8 blur is used because we're working at 600dpi, higher than the screening itself. The other important parameters are -p, -a and -alpha; the rest I don't change very often.

This gives us a good result for most of the image. We can smooth out the darker color in the eyes (green) and fabric (orange) a bit more:

3: Copy the background layer again, and put it on top. Name this layer "heavy".

spoilerI copied the background, not the layer we just filtered. The parts we want to clean up have some anisotropic artifacts (the weird swirly patterns, see zoom-in in #999). Instead of trying to filter away artifacts I just created, I'll do these sections from scratch and avoid creating them in the first place.

On the "heavy" layer, run the same greyc filter, but with -alpha set to 2 instead of 1. This strengthens the filter, giving us a cleaner result for the darker areas (compare #2 and #3).

spoiler

4: We want to use the "light" layer for brighter parts of the image, and "heavy" on darker parts. We'll use a luminance mask for this. Create a mask on "heavy" (that's why we put it on top). Select the layer by clicking on it, and run Image->Apply Image. Turn on "invert", and use defaults for the rest.

Then, we'll adjust the levels of the mask layer. This adjusts wihch brightness levels will get the "heavy" layer. Select the levels layer. (You can alt-click on it to view it directly, which can make this step make more sense.) Run Levels (Control-L). I happened to use black: 0, gamma: 0.38, white: 247. spoilerThis isn't carefully tuned for this image, it's just levels I pulled in from a previous edit; don't worry about the exact numbers, just what it does to the mask. Now, the heavy layer will be used even less in the brighter parts of the image.

Merge the heavy and light layers. You'll end up with two layers: "light" merged from the above, and background (the raw).

5: A common detail: we can do a better job on the blush. The edges are being faded away, even in the light filtering. It's a minor issue here (the difference is subtle), but in many images this is critical.

Copy the background layer again. Leave this layer below "light"; don't move it to the top. I've named this layer "detail". Run greycstoration with the above parameters, with -p 0.8 -a 0.8 -a 1.0. full list-p 0.8 -a 0.8 -alpha 1 -sigma 1 -gauss 0.8 -interp 1 -fast -alt This is like the "light" filtering, but with -p and -a increased from 0.6 to 0.8, which makes the filtering weaker, and to preserve edges more. spoilerThis part is often trickier, and I often use two greyc passes here to preserve the detail without preserving noise, but it's not necessary this time.

6: Create a mask on the "light" layer. Select the brush tool, set the brush color to black, and just mask away the blush, showing through to "detail". The image is shown with the layers below it hidden, to show what I've masked out. Flatten the image; we don't need the raw anymore.

spoilerIt's a subtle difference, but it's often not, so I make a habit of doing it. It only takes a minute to make sure we preserve one of the most commonly-overfiltered details.

Finally, crop, dust cleanup (mostly the spots on the right edge), then save to TIFF, and with XnView, resize and convert to PNG. spoiler(Actually, for this one I got lazy and resized in PS: image size -> DPI -> 350, but I normally do it as part of the XnView conversion.)

A couple other images in the pool have more darks, and I'd probably increase #3's -alpha from 2 to 2.5 or 3, so the darks aren't too noisy.

Humm... i never though of masking the blush. On mine i use the history brush at 30-50% opacity with a selection tool like mask/polygone tool/pen and add it back after the filtering.

I'll try this method next time :)

Thanks petopeto, but could you write something about color adjustment? Or maybe in forum #3342

blooregardo said:
Thanks petopeto, but could you write something about color adjustment? Or maybe in forum #3342

forum #4619

image_editing_help updated

I used the settings as prescribed, but ended up with some swirlies in the dark parts because the -a is set kind of high, though it's not as obvious when downscaled. What do you usually do about those?

I need the raw to tell what's happening.

Try to change -a in greyc... it always causes artifacts like that

It doesn't. I very rarely set anisotropy below .6 (or above .8). Lower starts to blur a lot.

(Use a heavy filter & heavy-noise picture to make it more visible) I think this artifacts is because of principle of work of greyc filter..

I can't tell what you're doing. Remember to include the raw, and the full parameters you're using (click Copy).

OK, since I'm trying to learn to edit because I want to help out, I was working on this tutorial. The part about the luminance mask drove me batty, probably because I'm not very familiar with masks and stuff in Photoshop anyway, and also I think Adobe may have changed things around a bit in CS4 vs. the version petopeto wrote the tutorial based on. I finally sorted it out, and thought I'd share for others who may be confused too, this is step 4:

Create a Layer Mask on the Dark layer you made in Step 3. Make sure you have the Dark layer selected, then either click on the Add Mask button in the Layers box, or do Layer > Layer Mask > Reveal All.



Now make sure you have the mask selected and click on Image > Apply Image. Select invert in the dialogue box, but leave everything else as defaults. If you did it right the mask should change to an inverted black and white image.



Again making sure you have the mask selected, adjust levels (Control-L or Image > Adjustments > Levels) If you do it right the mask will get a bit darker, with less white lines showing. It's really subtle so not worth a screenshot.

Finally, apply the mask to the Dark level by making sure you have the Dark layer selected, then using Layer > Layer Mask > Apply. If everything went properly the Dark layer will become mostly transparent, with only the darkest parts still in it ready to be merged.



Now you can merge the layers, but Photoshop will most likely name the new layer Dark because Dark was on top. Don't let this confuse you in the next step.

Hopefully this will be useful to someone. ヾ(＾∇＾)

You don't have to apply masks in order to merge it into something.
Peto's tutorial should apply to any version as long you know the terminology. ( ´・ω・｀)

Using apply mask is optional. You can use it, or you can directory merge it. The result is the same.
BTW, there is another way to apply mask I'm using; change the mask to selection. it allows to expand/contract the selection. I use it with USM2 to sharpen around dark edges (expanding a dark selection)