PhotoSynthesizer uses an entirely new method of mosaicing which uses your existing photos to resynthesize a target photo. The user first picks a target photo which will be resynthesized. Then the user picks a database of images which the app uses as the palette of paint for the resynthesis. The final stage is viewing the rendering of the synthesis. This stage shows the rendering of the image starting as a black canvas and slowly paints each region as though they were brush strokes. The final synthesis can take up to one minute to complete but is an entertaining process to watch if you’ve never seen an automated rendering before.
How Does It Work?
Simply, PhotoSynthesizer learns the appears of objects in the same way as Google Search by Image may learn the appearance of images. It finds and clusters these objects together and when presented a target image, it recreates the target image using only regions from the database of images. In this way, the target image provides the form and the resynthesis is an interpretation by the algorithm.