| Image fingerprinting is the act of generating a unique digest for an image. Unlike cryptographical hashing, slight differences in the input to the hashing function do not create significant differences in the digest. This property makes image fingerprinting useful in identifying near-duplicates of an input image. This paper describes a novel technique for generating an image fingerprint using Self-Organising Maps (SOM) with ranks higher than 2. The method is compared to a selection of more traditional fingerprinting algorithms and against a further variation on the proposed technique using a more conventional rank 2 Self-Organising Map. |
*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.