Attempts to solve autostereograms programmatically.
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Magic Eye

Attempts to solve autostereograms programmatically.


How it works

The algorithm for finding a solution is relatively straightforward.

The first step involves finding the offset at which the autostereogram repeats. This is accomplished by calculating the total pixel difference between the image and an overlap of itself at various offsets over a reasonable subsection of the image. The minimal diff is chosen as the offset.

The second step applies the offset to the image and calculates the absolute diff for the remainder of the image. Borders are added to compensate for the loss of an offset's worth of pixels.

Although simple, this algorithm seems to work relatively well and is efficient to execute, although there is a small chance that it won't determine the correct offset for the autostereogram.


# Install dependencies.
brew install opencv

# Create a build directory.
mkdir /tmp/build
cd /tmp/build

# Build and run.
cmake path/to/src
./magic_eye acorn.jpg


If you're using Vim, you can speed this up by mapping the build script to a key:

:nmap - :w \| !mkdir -p /tmp/build && cd /tmp/build && cmake path/to/src && make && ./magic_eye test.jpg