Image Indexing Based on STIR and MPEG4

Image Indexing System or Invariant Associative Image Memory is an essential part of various multimedia, security and image processing systems. To create an Invariant Associative Image Memory two key problems occur: selection of effective invariant features and effective coding of the reference images. The paper gives the results of the development work related to design an Invariant Associative Image Memory based on the use of STIR (Scale, Translation and In-plane Rotation) invariant transform and MPEG4-like codec. The memory was implemented as a software package on a powerful PC and tested in various experiments with recognition a large set of human face photographs, which were arbitrary shifted, rotated, scaled, partly covered and corrupted by noise.