If you are the most bored person ever, then you may have looked at the
picture I posted a month ago on Twitter of the puzzle we solved while on vacation:
If you did, then you might realize that this blog post is all about Photoshop. Why? Because we didn't actually solve the puzzle. It was two thousand freaking pieces, most of which were identical sky and ocean, and it took us hours to just do a tiny portion. Here's how far we actually got after about 10 man hours: 
At this point I commented jokingly to Robey that it would be easier for me to make a fake photograph of the solved puzzle than it would be to really solve it. He didn't believe me, so I had to take up the challenge. Here's how I did it: (I should start by pointing out that the most important aspect of a successful Photoshop forgery is that it not arouse suspicion. Since any pixel in the image can be set to any color, it's impossible to convince a skeptical audience of the truth of the image. So once they suspect it, you've already failed. But if your audience is not yet suspicious, your image merely has to avoid raising doubt, which is pretty easy. It doesn't even have to be a really accurate forgery--- it just has to seem plausible to the viewer that the scene you are showing them could exist somewhere in the world.) First I got an image of the completed puzzle from the box top, and a better one online: 
My plan was to fit it onto the real photo of the incomplete puzzle. The problem with the incomplete puzzle photo is that it's full of evidence that the puzzle is incomplete. For example, it's suspicious to have extra puzzle pieces lying around on the table and in the box. So I cropped the box out and rubber-stamped away the puzzle pieces on the table: 
From here, Photoshop's "Distort" mode under free transform made it easy to fit the completed puzzle image onto the photo of the table, like this: 
Now there are several things wrong with this image as-is. The colors look overly perfect and don't match the lighting in the scene, and of course the image doesn't look like it's made of puzzle pieces. To get the appearance of a puzzle I found a handy puzzle grid online, which looked something like this: 
A two thousand piece puzzle is a 40 by 50 grid, so I just repeated this grid in a big rectangle and then distorted it to fit the puzzle area. That looks basically like this: 
By taking the magic wand tool out of "contiguous" mode I was able to select away the white pretty easily, leaving me with a plausible puzzle cut: 
Unfortunately the next major problem is the lighting match. Compared to the real puzzle pieces, the image is far too saturated, and it needs to be lit with the natural daylight of the surrounding room. Here's a sample of the real puzzle, to show how it doesn't match: 
I can fix the tint with a "curves" adjustment layer that greens up the image significantly, and gets rid of the unrealistic pink sky and bright colors in the buildings: 
But the glare from the nearby window makes the real image look much less saturated. I can simulate the glare with a partially transparent radial white gradient fill, centering on the top of the puzzle. I also air-brushed in some black around the corners to emphasize the daylight glare: 
As a final touch, I noticed that the glare from the window also catches the edges of the embossed puzzle pieces. Fortunately photoshop has an "emboss" filter which emulates this effect pretty well, so I add that to the top of the puzzle grid. That plus a little more glare and I have a plausible forgery:
If you saw this image side by side with a real finished puzzle on this table, you would probably know which one was fake. But remember, it doesn't have to look perfect, it just has to look plausible enough to avoid raising suspicion. So did you suspect it? (In my defense, these techniques are not purely nefarious: it's important to be able to convincingly post-process a photo whenever the camera has failed to capture the image you want to share with your viewer. Photoshop can restore the experience of being there, even if your camera lost it.) |
Comments
blog comments powered by Disqus