Multiple partial solutions for the point-to-point correspondence problem in three views

Abstract

This paper presents a method of point-to-point correspondence analysis based on a combination of two techniques: (1) correspondence of multiple points through similarity of invariant features in three views by using a standard feature method, and (2) a combination of multiple partial solutions through the trifocal geometry. This method allows the determination of point-to-point geometric correspondence by means of the intersection of multiple partial solutions that are weighted through the MLESAC algorithm. The main advantage of our method is the extension of the algorithms based on the correspondence of invariant descriptors, generalizing the problem of correspondence to a geometric model in multiple views. For all the images analyzed, we showed that the point-to-point correspondence can be generated through a multiple geometric relation between three views. An important characteristic of our method is that can be used in sequences of images that have a low signal-to-noise ratio.