A new efficient similarity metric generic computation strategy for pattern-based very low bit-rate video coding

In the context of very low bit-rate video coding, patternrepresentations of a moving region (MR) in block-based motionestimation and compensation has become increasingly attractive.Generally, all existing pattern-matching algorithms apply asimilarity metric involving elementary operations, to computethe mismatch between a MR and a particular fixed pattern inorder to select the best-matching pattern from a fixed-sizecodebook of predefined patterns. In this paper, an efficientsimilarity metric together a new generic computation strategy ispresented by considering only the mismatch areas of MRs. It istheoretically proven that for a specific MR in a macroblock, thenew similarity metric selects exactly the same pattern as existingmetrics, while the resulting computational coding efficiency isimproved by between 21% and 58% compared with the H.263low bit-rate coding standard.