Refinement techniques for animated evolutionary photomosaics using limited tile collections

An animated evolutionary photomosaic is produced from a sequence of still or static photomosaics to evolve a near match to a given target image. A static photomosaic is composed of small digital images or tiles, each having their own aesthetic value. If the tiles are prepared manually, the tile collections are typically small. This potentially limits the visual quality of a photomosaic as there may not be sufficient options for matching tiles. We investigate the use of colour adjustment and tile size variation techniques via genetic programming to improve the animated photomosaics. The results show that colour adjustment improved both visual quality and fitness. However, it can produce strange looking tiles. Tile size variation was able to focus on details in the target image but produced slightly worse fitness values than an equal-sized tiles approach. Combining these techniques revealed that, regardless of the size of tiles, colour adjustment was the dominant refinement. In conclusion, each of these techniques is able to produce an aesthetically different animation effect and presents a better mechanism for generating