Alright this will be a very long story with things I always keep in mind when making one myself..
I take an image I personally like and make sure that:
One: the image has a ratio of at least 1.33:1 or more. Images that have less are possible, but unwise considering the 3DS screen is which ratio is 1.66:1. If you have a lower aspect ratio then 1.66:1 you will get pillar boxing, that's not bad if the image is between 1.66:1 - 1.33:1, but anything lower will be quite hard to see on the 3DS screen (I've learned from my
second try). Anything higher then 1.66:1 though is a lot easier, although I wouldn't try to go further then 2:1, as then the letterboxing will take it's toll.
Two: there is something that I can work with in terms of depth in the image. For example,
this image was an excellent choice because I have a total of four layers: The front guy, the bottom guy, the top on the ground and the background. Unlike
this image in which I can't really do anything with depth. I can still do it, but it won't matter much, it would be waste of the effort honestly.
Three: it's not censored beyond usage. Self explanatory I assume.
Once I have the image(s) I want to convert I open them up in Photoshop, I copy one of them as a base of one perspective, and use the copy to separate the characters from the background, but I do keep everything I cut for later. After the characters are separated from the background, I cut the characters into pieces all for separate layers (front, middle back). Once every piece has been cut for each separate layer, I (sometimes) make a couple of extra maps for each layer: one for the foreground, the middle ground and the background. This isn't necessary, but it's useful to keep the (many) layers ordered.
Once I have everything separated from each other, orderd in separate maps and made sure there is an extra .PSD saved in case I may screw up. I shift the layers towards the left or the right (depending how the base image looks), and the further the layer has to be, the more I shift it (example front layer 2 pixels, second layer 7 pixels, and background 10 pixels). Well naturally this will cause holes in the image, which I fill them up on separate layers, I don't draw on the existing once just in-case.
After it's done, I save the images (the original and the edited one) as.jpg once everything is filled up and open a small program called:
StereoPhoto Maker. Click and drag both files into the program and click the Anaglyph button, doesn't matter if it's the Black and white, or the one with color (although for testing I use black and white). I put on my "awesome "3D glasses and switch the images if the depth perception doesn't work properly (front in the back..? Flip the image and it's solved!). If the depth works the image looks very, very 2D-ish with depth. This is because all the layers are still pretty much identical.
After that in Photoshop I: skew, distort, wrap or whatever the separate layers to make the image appear more 3 dimensional although not always does it work perfectly. This is the point where I very often save my progress as a .jpg and check again if the image is improving in terms of depth using
StereoPhoto Maker. I keep doing this until I am satisfied with the end result, and if I am I save it as a .MPO, upload that to a site and use my 3DS to see if it works like it should (as Anaglyph and "real" 3D sometimes differ), if it's good even on my 3DS, I upload the image as a .jpg and as a Anaglyphic image and upload them here.
It's very time consuming and very confusing at times. Here is an image (url is dead) of layers of the feet one in the image above. And you shouldn't forget that I did this one twice (29 layers * 2) due to not entirely being satisfied with using the original perspective... I seriously
do not recommend anyone to try it, and if you want to try it, try creating multiple images before starting to upload . My first couple of attempts have some serious problems in the conversions and layering, and I'm still learning to improve and perfecting even after 25 conversions.
So that's the entire story really... Like I said, a long story, and I hope people aren't scared off. But like I said, it's quite a bit of work, and it doesn't always pay off. I like to do it because it's relevant to my interest (3D modelling), but if you just like to see 3D images but don't have more "feelings" for it. Then seriously don't try it