Using Photoshop and Your Brain to Produce Diorama Illusions
Have you ever built a diorama? It's a depiction of a scene in miniature. Or perhaps you remember owning a dollhouse or train set as a kid. You lay on the floor, inches away from tiny versions of full-size objects. In the background, your brain was busy writing the software that makes this illusion effective.
This fun photo treatment can be performed by almost anyone in short order, once some basics are understood of how the brain processes images. Let's get to it!
Background 1: Understanding the Mind's Eye
The mind is an interesting construct. I'm sure you have spent some time looking at optical illusions. Why do they work? Why is it that we see some things that aren't really there?
At the risk of sounding a bit Zen for an image editing tutorial, the answer becomes clear when you remember that you are not actually experiencing reality when you look at the world. Instead, you experience an internal representation of data from your sensors (in this case your eyes).
When you were very young, your brain wrote the software that it uses to process image data from your eyes. The more reference data this software collected, the more solid its view of how things look became, and (importantly for this effect) how those different things appear in varied situations.
Background 2: Effects of Depth of Field on Perception
Your brain developed a fairly simple set of rules that it uses to determine what looks right with regards to size and position. Here are some of them that are pertinent:
- Objects close to us have higher resolution than objects far away.
- The object that has your central focus is seen clearer than one that is on the periphery of vision – even if the peripheral object is nearer to your eye.
- The further away a scene is from you, the flatter it appears and the less detail it contains (comparatively) to close-by objects.
- Very clearly defined objects are either: very close or very large.
Depth of field is a term that many of you will be familiar with. It describes how far away something can be and remain in focus. You can easily see its effect if you hold your hand flat in front of your eye, so you can see past it. Close an eye, then focus on something across the room and you'll see that your hand fuzzes up a bit, switch focus back to your hand and it will be crisp while the background blurs.
What we do in this effect is selectively alter depth of field so specific parts of the image are blurry when they would not be in a real scene. We either break or flop the processing rules above, and your brain attempts to make sense of what it sees.
Background 3 – Effects of Lighting, Color and Detail on Perception
Also important to your software are the representation of lighting, color, and details in an image. Here are some rules your brain uses when it processes these variables from an image:
- Real scenes are more detailed than man-built models. (Think of the difference between matchbox cars and real cars).
- Real objects vary in their coloration more than model objects, and are often weathered. (toy companies use this often now, in battle-damaged toys, to increase their realism).
- Lighting conditions are different between real objects and model representations. (Most models are seen indoors, under direct artificial light as opposed to in full spectrum sunlight).
Two notes before beginning. First, when I recommend specific feathering values (like 50px) I am basing these values on the images included for PSD Plus members. If you are using your own images, or the 600px wide versions here, your values may need to be lessened somewhat.
Second, I suggest saving the various selections we'll make whenever this tutorial creates a selection area. I've saved them in the PSDs that come with PSD Plus membership, but it is a generally good practice that can save you a lot of time on any project, especially when making complex masks you might need again. The save selection dialog looks like the image below, and is accessed via Select > Save Selection.
Image 1: Applying the Diorama Technique to a Landscape Image
This is a picture I shot in Shinjuku, Tokyo in October 2008. Your brain is telling you a few things about it:
- Almost all the picture is in focus, but there is not a ton of close-up texture and detail; the data was probably collected far away from the image contents.
- It knows that you are mostly looking at large things – cars, buildings, et cetera; since they are small in this image then you must be far away.
- The objects at the fore have slightly higher detail than objects further away, and the loss of detail is consistent with landscape references you've analyzed before.
Combined with the depth of field information it has gathered, your brain comes to this conclusion: You are looking at a scene of large objects, taken from far away.
Now that we know what rules our software uses to come to this conclusion, it's time to break them.
Step 1: Selecting a Focal Point
Choose a portion of the image that you'd like to highlight. There is really just a single rule here: the selected focus should be in the mid-ground. Here, we will use the trucks on the street. Follow the instructions below.
Step 2: Creating and Feathering a Selection
There are several types of selections we can use for this technique, but in this case we'll use a shifted oval. Since Photoshop doesn't allow rotation of selection marquees, we have to use a path and then make it a selection.
Convert the work path to a selection.
Invert the selection and then apply a Feather, as indicated below.
Step 3: Applying Simulated Depth of Field
We'll use the Lens Blur filter to blur the image in the area outside our focal point. The Lens Blur filter works outside the selection (i.e the selected area is left alone while the area outside the selection takes the effect), so do not invert this selection.
Step 4: Applying Lighting Effects
First, we open the curves dialog (Command + M or via Image > Adjustments > Curves). Drag the curve slightly upwards to blow out the image.
Next, we use the Omni light to simulate an all-over artificial light. You can access it from Filter > Render > Lighting Effects.
Step 5: Adjusting Color
Toys and models frequently have deeper color and are more vibrant than their real-world counterparts – partly because they are not subjected to the same dust and weathering and partly because kids like bright colors. This is easy to achieve by pumping up the vibrance and saturation a bit. Set Vibrance to +30 and Saturation to +40.
Step 6: Final Step for Image 1
You might notice that the previous step pumps up the blue in this image to a distracting level. After applying the Vibrance/Saturation settings, we fix this by using auto-color under the Auto-Color menu (Command + B), and our first image is complete!
Image 2: Applying the Technique to a Flightline of Airplanes
This image is an US Air Force file photo of F-22 Raptors on a flightline. It can be found here. What makes this image look real as opposed to a diorama of the same scene?
- There is a high level of detail in all the objects shown.
- The background is very detailed for how far away it is.
- The lighting is consistent with sunlit aircraft.
- The nearest aircraft is the most detailed, with rear aircraft losing details, as expected.
Really, it's no different than our street scene. Where our execution will differ is in the precision and number of our selections. Where in the city scene it was enough to select a single feathered ellipse, here we must take care to mask the details of the aircraft in the middle, which will be our focus.
Step 1: Removing Detail
The image has far more detail than a model would generally have. Use the Clone Tool and Patch Tool to cover up excessive details like rope lines, tarmac imperfections, etc. You can work a quick and dirty in the back and foreground, since heavy blurring is on the way in later steps.
Step 2: Defining our Selections
On this image we will be defining four distinct areas for treatment – the fore, mid, mid-back, and far-backgrounds will each have their own masks.
Making the Foreground Mask
The foreground mask contains the foremost aircraft, as well as the ground beneath it. Use the polygon marquee selection tool to select an area as shown below, then invert the selection (Command + I) to make sure that everything in the foreground is selected.
Use Lens Blur with settings similar to those shown below. I've used a mild specular highlight setting to increase the brightness of some of the higher tones.
Use Command + D to deselect. Select the Pen Tool, and draw a path like below, then turn it into a selection like you did in the "Shinjuku Street" example. Feather the selection 50 pixels, invert the selection, then apply a Lens Blur with a radius of about 47. That's all for the foreground.
Making the Background Masks
The far background mask contains the buildings and grass, but not the rearmost airplanes. The mid-back contains the rear tarmac and the rearmost planes.
Use the Polygon Select Tool to make a selection as shown, then invert the selection (Command + I). You don't have to be super precise here; make sure you leave a little room around the vertical stabilizers of the rear aircraft.
Use the Refine Edge (Command + Alt + R) dialog to make sure that the far background does not intrude on the tails of the aircraft.
Apply a heavy Lens Blur (with a radius of 55) to the far-back selection. Invert the selection again. Use the Polygon Select Tool, with the Alt modifier key held down, to subtract from the selection so it looks like the marquee below. Try to be fairly precise near the vertical stabilizers and wings of the mid-ground aircraft, as we won't be doing too much feathering here.
Feather this selection just 5 pixels, then invert it. Apply Lens Blur again with a 25px radius. Finally, Deselect using Command + D.
Step 3: Lighting and Coloring the Image
Open the Curves Adjustment Tool (Images > Adjustments > Curves or Command + M). Adjust the curve upwards to blow out the image slightly, as in the last exercise. Don't overdo it.
Next, I'm going to use a wide spotlight from the Filters > Render > Lighting Effects menu, with the focal point located on the mid-right aircraft and with the direction of the light pointing the same direction that the sun points in the original shot.
We'll pump the vibrance a little less this time, using the Image > Adjustments > Vibrance dialog. I used a setting of Vibrance 18 to make this final image. I did not modify the saturation, since the tarmac has a significant amount of brown in it as well as gray. Increasing the saturation would make it pop out too much.
Step 5: Final Detail Removal for Image 2
You may wish to spend a moment using the Blur and Clone Tools to reduce additional detail. In this final image, I spent a few seconds blurring out some of the details on the mid-ground airplanes (mostly small coloration differences on the tops), and cloning an unneeded shadow on the roof in the far-back. The final image two is below.
Afterword 1: Which Images Work Best with This Technique? Which Don't?
Images taken from an angle and from above tend to work best for this effect, although I've seen some great examples from straight down as well. Shots taken directly from the side, and which lack good fore-mid-back separation, are not generally good candidates. A crowd of people taken from a balcony will work great. A picture of a person against a brick wall would be a poor choice.
Additionally, images that contain very few recognizable objects are not good. Also, many objects that are of similar size won't usually work either. For the illusion to work, you brain has to process a disconnect, and if there is not anything it can reference, your brain will decide that the image is normal, and see it as such.
A normal person will almost always refer back to what they expect an object to look like. If you take a picture of an abnormally large object (say, a giant prop coffee cup), and photograph it in isolation, very few people will say "oh look, you took a picture of a giant coffee cup!" To the last, anyone asked to guess its dimensions will say it is a normal sized cup of a few inches across without the benefit of a contrasting reference point.
A good example of using this technique in reverse can be seen in any "I Shrunk the Whatever" style movie. You have probably never seen a normal-sized person next to a 40 inch tall coffee cup and donut sitting on a football field sized table. You have seen many reference images of a regular cup and donut. Having been primed that the movie is about people getting shrunk down, your brain creates an instant mental image of a very small person next to these common objects. Sometimes called forced perspective, we are using this technique in reverse when we miniaturize our scene.
Lastly, images with people in them where the people's faces are clear and visible don't lend themselves well to this technique. You are just too well programmed to see faces as real. Heck, we see faces even when they aren't there half the time. When dealing with people in your images, it is preferable to blur their features a bit even if they are the object of focus think about action figures when going for this effect in shots where people are present). But in general, I recommend staying away from using people as your subjects.
Afterword 2: What Ruins the Effect?
Being too aggressive with blurring, hard edges between blurred and unblurred areas, blurring where none is expected (as in the vertical stabs of the middle aircraft in the last example) will all decrease the effectiveness of the illusion.
Being too over the top in your lighting and coloring will drop the brain out of thinking that it is a photo of a miniature scene and into thinking that it is a doctored image. Mostly, it becomes a distraction that the viewer will fixate on instead of giving in to the illusion.
Leaving excessive details can detract from the technique as well. Don't go overboard, but do eliminate or reduce things not commonly seen in miniature – pavement cracks, distressed colors, etc.
In closing, the number one killer is choosing a bad source image. You will know quickly (within the first blur usually, or just on looking at it after you've done a few) whether an image is ripe for miniaturizing. If the effect isn't popping after a short effort, analyze whether it's your technique that is the issue, or just a bad choice of source material.
Thank You & Teach the Teacher!
Other than that, have fun with this! If you do some of these of your own, or have other suggestions on how to improve or add to this technique, please add them to the comments, with links to your creations. Tutorials are a two way affair, and I hope to learn from you, and hope you've learned from me.
Subscribe to the Psdtuts+ RSS Feed for the best Photoshop tuts and articles on the web.