Automatic generation of photo albums: Collective photographic/videographic memory over the family-area-network

It is not necessary to press any button on the ``lookpainting camera'' because it can automatically determine when images are in the same orbit of the projective group of coordinate transformations. This is done by analysis of the error terms in (12), or, by comparing, for each pair of images in the sequence, $\sum_{\bf x}(\hat{I}_i - p_{ij}I_j)^2$with a threshold to determine whether or not they are in the same orbit. The choice of thereshold is adaptive, as suggested by a Likelihood Ratio, Neyman-Pearson detection strategy, or CFAR (constant false alarm rate) detector in which the user can, for example, select a rate at which new candidate images are presented.

It is then assumed by the algorithm, that if more than some number (say, a hundred or so, or a number based on orbital distance) images are captured within the same orbit, that the subject matter is of sufficient interest to begin building an environment map. The building process stops as soon as the incoming images are no longer in the same orbit, and if enough new images arrive to form a second orbit, a second lookpainting is generated, and so on. All of the lookpaintings may be posted to a World Wide Web page, or disseminated over a smaller ``family-area-network''. In this way, when, for example, when one or more family members goes on vacation, wearing the special sunglasses, the family's photo album is generated automatically without the need for any conscious thought or effort on the part of family members.