Abstract

Given a set of poorly aligned images of the same visual concept without any annotations, we propose an algorithm to jointly bring them into pixel-wise correspondence by estimating a FlowWeb representation of the image set. FlowWeb is a fully-connected correspondence flflow graph with each node representing an image, and each edge representing the correspondence flflow fifield between a pair of images, i.e. a vector fifield indicating how each pixel in one image can fifind a corresponding pixel in the other image. Correspondence flflow is related to optical flflow but allows for correspondences between visually dissimilar regions if there is evidence they correspond transitively on the graph. Our algorithm starts by initializing all edges of this complete graph with an off-the-shelf, pairwise flflow method. We then iteratively update the graph to force it to be more selfconsistent. Once the algorithm converges, dense, globallyconsistent correspondences can be read off the graph. Our results suggest that FlowWeb improves alignment accuracy over previous pairwise as well as joint alignment methods