Maximizing Image Utilization in Photomosaics

Abstract

Non-photo realistic rendering (NPR) is an appealing subject in computer graphics with a wide array of applications. As opposed to photo realistic rendering, NPR focuses on highlighting features and artistic traits instead of physical accuracy. Photo mosaic generation is one of the most popular NPR techniques, where a single image is assembled from several smaller ones. Visual responses change depending on the proximity to the photo mosaic, leading to many creative prospects for publicity. Synthesizing photo mosaics typically requires very large image databases in order to produce pleasing results. Moreover, repetitions are allowed to occur which may locally bias the mosaic. This paper provides alternatives to prevent repetitions while still being robust enough to work with coarse image subsets. Three approaches were devised for the matching stage of photo mosaics: a greedy-based procedural algorithm, simulated annealing and Soft Assign. We found that the latter two approaches deliver adequate arrangements in cases where only a restricted number of images is available.