photomosaic¶
Tile images from a large collection into a mosaic which, viewed at a distance, gives the impression of one large photo.
This implementation has several distinguishing features:
Mix tiles of different sizes and place them in irregular or overlapping positions — not confined to the usual rectangular grid.
Partition an image into tiles adaptively, subdividing regions of high detail into smaller tiles.
Include a mask, and tile partitioning can roughly trace its edge, giving the appearance of puzzle pieces scattered on a table in some shape.
To maximize contrast and reduce tile repetition, adapt the color palette of the target image to the palette of colors available in the tile pool.
Obtain large collections of images to use as tiles using the official Flickr API. Easily collect attribution info and filter results by license.
Employ perceptually-accurate color matching via the colorspacious package.
Development¶
The code is factored with maximum flexibility in mind and uses only standard data structures: numpy arrays, Python slice objects, and other built-in Python types. It is easy to experiment with custom strategies for partitioning tiles, characterizing colors, choosing matches, or drawing the final result.
You are cheerfully invited to contribute to this project on GitHub. In particular, these topics might be interesting:
Alternative matching algorithms that minimize tile repetition
Alternative tile-partitioning algorithms
Additional image-harvesting tools (continuing to give due attention to licenses, attribution)