Registration of plantar pressure images

Abstract

We present an analysis of four different algorithms used to register plantar pressure images: a first one based on the matching of the external contours of the feet, a second algorithm based on the technique of phase correlation, a third one based on the direct optimization of the cross-correlation (CC) and using the Fourier transform, and a fourth and last algorithm that is based on the iterative optimization of an intensity (dis)similarity measure. In terms of accuracy, the later algorithm achieved the best registration results; on the other hand, the algorithm based on the matching of contours was the fastest, but its accuracy was inferior to the accuracy of the remaining algorithms.