Hey everyone! I’m back with the second half of this two part series about using Photoshop as a means of extracting a subject from its background, and am very sorry that it took this long to get it done. In Part 1, we undertook a basic review of what channels are, and how they work. If you haven’t read that article already, I suggest that you review it before proceeding here.
Part 2 is (as promised) going to be more of a tutorial type article, in which I am going to guide you step by step through the process of extracting a lovely model from a photograph, and placing her into a relatively simple design. There are quite a few screenshots, but it’s really nothing to worry about.
Before getting started, I would like to thank my 2 awesome sisters-in-law for allowing me to use this awesome photograph as my working example.
The model is of my sister in law. The photographer is another one of my sisters in law. They are both awesome and crazy, and I love them to death. So, thanks ladies for being willing to be a part of this tutorial!
Now, let’s get started with the extraction.
First, it’s important to note that our image is in RGB mode, and that we will be sticking in this mode throughout the course of this tutorial. If you read Part 1 of this series (or if you are already familiar with channel basics), you will recall that every RGB image contains a Red, Green and Blue channel. Let’s start by analyzing the 3 channels to determine which one will provide the best starting point for extracting the model from the photograph. Here are the greyscale representations of each channel.
Since the hair is going to be the most challenging part of the extraction, what we are really looking for is the channel where we have the best contrast between the hair and the background. The red channel would probably work, but it’s a bit lighter than is really ideal, so we’ll scratch that one.
The blue and the green channels are pretty similar in terms of the contrast between the hair and the background, so either of those would probably work. However, I think that the green channel is probably the better option, so let’s go ahead and duplicate it. You can do this by either right-clicking and selecting Duplicate Channel from the menu, or by dragging the green channel down to the new icon in the layers palette.
Now, with our duplicate green channel selected, we are going to adjust the brightness and contrast. Select Image Â» Adjustments Â» Brightness/Contrast from the menu. For this image, I set the brightness to 16 and the contrast to 48.
These numbers will vary drastically from photograph to photograph, but the basic idea is to adjust these values to the point where the background is white and most of the hair is black. The thin strands, however, should retain a certain amount of grey, as you can see in the screenshot above. Also, try to increase the contrast to the necessary minimum, since too much contrast can cause some of the finer details to vanish, and for some of the softer edges to become jagged and rasterized.
Now, while we’ve adjusted our contrast sufficiently on the left side of the woman’s face, the contrast above her head and to the right is not quite good enough. To fix this, use the Rectangle Marquee tool to make a selection around the area in question, and adjust the brightness and contrast again.
Doing so will only apply the adjustment to the selected part of the channel. This may look a bit odd, as it breaks the continuity of the image, but that’s fine. You’ll see why in a bit.
In this step, we want to turn the RGB channel back on so that we can see everything in full colour again. Now, with the the main image layer active, select the Pen Tool from the Toolbox. Use this tool to begin tracing the rest of the model’s jacket. I would recommend starting at the bottom left of the image, as shown here:
Then, just trace the rest of the shape. Later, we will use this shape to establish the layer mask that we will use to extract the woman from the background. If you get to a part of clothing that is a bit softer, like this woolen collar, you can draw your line back a bit and then return later to blend it all together.
In this image, we also have a break in the model’s arm, which we will want to account for.
We want to make sure that we draw in the shapes for any enclosed areas, like these two spaces between her arm and her body.
If you forget to trace this part, the background here will not get extracted. While it’s possible to go back and fix this kind of mistake, it’s always better to take care of it the first time. Also, you can actually create this extra shape right within the same path object, so there’s no need to create two separate path objects!
Next, select the Brush Tool from the toolbox, along with a medium sized, soft brush. What we want to do now is start brushing away all the darker spots in the background, and even some of the edges of the woman herself (but not her hair). This screenshot shows the first stages of this process:
This next shot shows that I have painted away all the darker areas on the top part of the canvas, including the defining lines around her jacket.
Repeat this same process on the bottom half of the image. Again, you can see exactly what I mean in this screenshot, in which I have made the entire background white, and also painted over the edges of the model.
It may seem counter productive to be painting away these edges, but it will actually help us to get a nice, crisp extraction in the next step.
Now, open the Paths Palette. There should only be one path there – the one you created when you traced the woman with the Pen Tool. Command-Click (PC: Ctrl-Click) the preview of that path to create a selection.
Next, making sure that you are still working on the duplicated green channel, fill the selection with black. This will give us back the basic shape of the woman that we brushed away (I told you we’d get it back easily).
Select the Brush Tool again, this time with a slightly smaller size. Set the foreground colour to black, and start painting over the parts of the model’s face that are still visible.
Reduce the brush size, and very carefully start filling in the highlights on her hair, being sure not to paint away any of the finer details around the edges.
Now it’s extraction time. Your duplicated green channel should now look something like this:
The parts of the image that we want to keep (the model) are in black. The parts that we want to discard are in white. Start by revealing the RGB channel again to get all of the original colour back. With the channels palette still open, Command-Click (PC: Ctrl-Click) on the duplicated green channel. This will create a selection based on all of the work we have done so far. Choose Select Â» Modify Â» Feather from the menu, and when the dialogue box comes up, set the feather radius to 2px.
This will soften the edges just a little. Next, press Shift-Command-I (PC: Shift-Ctrl-I) to invert the selection. Lastly, with the model layer selected, create a new Layer Mask. You can do this by simply pressing the Layer Mask button at the bottom of the Layers palette (ignoring that it says Add vector mask – trust me this will add a layer mask – the vector mask option appears after you have already added the layer mask, and is an entirely different topic all on its own):
Once you add the layer mask, the model will be completely extracted from the background.
From this point on, the steps are all about taking the extracted figure and placing her into the the context of a simple design. First, we’re going to start by adding some texture into the background. For this, I am going to be using one of my own watercolour textures, which you can find here.
Open up the image file and immediately select Image Â» Image Size from the menu, to determine the exact dimensions of the texture.
As it turns out, our working document is actually a bit larger than this, so flip back to it and select Image Â» Image Size again, and resample the image so that the entire watercolour texture will fit into the background.
In this case, that means setting the height to 3450 pixels. The width should automatically shrink down to 2249 (narrower than our texture), as long as you have the Constrain Proportions option selected.
Switch back to the texture document. Press Command-A (PC: Ctrl-A) to select the entire image. Copy it to the clipboard, flip back to our working document and paste the texture in as a new layer. Finally, drag this layer behind the model. The document should look something like this:
Let’s work on the background just a little bit more here. To start, we’ll add another layer of texture. If you want to use the exact one that I will be using, head out and pick up my Mega Texture Pack and open this image:
Alternatively, you can use any similar grunge texture that you might have in your repertoire. Copy and paste the texture, just like we did before, and place it at the back of the document (the grunge texture I am using is larger than our working document, so no resizing is necessary). Then, set the blending mode of the watercolour texture to Soft Light.
This gives some extra roughness to the background, which I really like, but ends up muting the colours from the watercolours. To fix this, duplicate the watercolour layer, move the duplicate above the original, set the blending mode to Overlay and the opacity to 50%:
This will bring some of the colour back into the design.
This step is really quick. Select the extracted woman layer and add a choose Layer Â» Layer Style Â» Outer Glow. In the dialogue box, select a medium grey colour, set the blend mode to multiply, the opacity to 55%, the spread to 0 and the size to 125.
This will create a nice, soft shadow around the model, adding just a bit of visual interest to the composition and helping to blend her with the background.
Now, the original image had some subtle shadows in it. Unfortunately, when we extracted the model, we also extracted the shadows. We could probably have tried to do some subtle work to extract the shadow too, but I prefer to just add it back in after the fact. This also has the added benefit of keeping the shadow and the model on separate layers.
Let’s start by duplicating the model layer, dragging it above the original, and deleting the layer mask. Then, set the blending mode to Multiply. This will now be our shadow layer.
Next, head back to the Channels Palette, and create the same selection that we did before, by Command-Clicking (PC: Ctrl-Click) the channel preview. This time, chose Select Â» Modify Â» Contract from the menu, and set the number of pixels to 2.
Return to the Layers Palette, and select the new shadow layer, then apply a new Layer Mask based on the current selection (which we have not inverted). This will knock out all of the shadow on the woman herself, leaving it only on the background.
Now the shadows look a little too dark, so reduce the opacity of the shadow layer. I found that 65% worked well for this image.
We just have a little bit more to do to the model herself. First, start by duplicating her layer again, and drag this one to above both the original, but beneath the shadow layer. Temporarily disable the layer mask by Shift-Clicking on it. Now, add a new Hue/Saturation adjustment layer. Check the Colorize option, then set the hue to 335 and the saturation to 13.
You will also want to create a clipping mask, so that the adjustment will only be applied to a single layer. To make sure that the Hue/Saturation layer is directly above this duplicated model layer, right click on the adjustment layer and select Create Clipping Mask from the contextual menu.
Next, add a basic Gradient Overlay layer style. Use a plain, black and white gradient reducing the opacity to about 44% and setting the blending mode to Multiply. Also, set the angle to 128.
This will add a soft, simulated shadow effect, which will darken the areas toward the bottom right of the image, as you can see here:
Finally, turn the layer mask back on (Shift-Click again), so that the Hue/Saturation is only applied to the model. Set the blending mode to Overlay, and reduce the opacity down to 55%. Altogether, this should give you an image that looks something like this:
That pretty much wraps up this tutorial! Of course, you would probably add a bit more to the design to round it out a little – maybe some simple boxes and text to create a poster kind of like this:
Still, the 11 steps above pretty much cover everything you need to know to extract the woman from her background and build up a simple but interesting design around her. That being said, I want to note that every extraction is somewhat different. This tutorial should help you with a fair number of your complex extraction projects, but the techniques outlined here may not be appropriate for every instance that you need to perform an extraction.
You’ll really have to weight it on a project-by-project basis.
I hope that this article (and Part 1) was valuable to you, and that you learned a thing or two about extracting with channels! And again, thanks to my one sister in law for providing the photo, and to the other for letting me use her face for thus tutorial!