PDF version of this entire document

Surface Signatures

The currently-worked-on approach will try to measure distances between images in hyperspace based on their parameterised version, where these parameters are basically a small set of distances, each (hopefully) encompassing a sort of concise digital signature corresponding to a person's facial surface alone. Currently, 60 such parameters are planned for use, centered around points in the eye and nose regions (less prone to change due to expression).

The sketch shows the approach tested so far. It is a brute force implementation that measures many geodesic distances and then compares surfaces based on distance-to-distance subtractions. It is not particularly clever, but the results of recognition tests are not too bad, either. They help validate the premise that by measuring Euclidean distances in XY, YZ, and XZ (based upon geodesic operators like FMM) we are able to carve out the surfaces and extract meaningful measures from the sub-surfaces.

Figure: Brute force implementation that measures many geodesic distances

Another figure, Figure , shows the next step.

Figure: This figure visualises the idea of encoding surfaces as a vector not of surface vertices but an ordered list of Euclidean-upon-geodesic distances, which are fast to compute and sensitive to isometric/mildly detectable alterations