IMAGE CLASSIFICATION EMPLOYING COMPRESSED IMAGE VECTORS USING VECTORIAL QUANTIFICATION

摘要

A storage medium, method and apparatus are described. Local descriptors are extracted (32) from an image (10) and an image vector having vector elements that indicate mixing model component parameters of a mixture model representing the extracted local descriptors is generated ( 14). The image vector is subdivided (42) into a plurality of sub-vectors, wherein each sub-vector comprises at least two vector elements. Each sub-vector is compressed (46) independently using a vector quantization algorithm. Further, the compressed subvectors are concatenated to generate a compressed image vector representing the image (18), wherein the extraction, generation, subdivision, compression and concatenation are implemented by a processor digital.