System and method for encoding multidimensional data using hierarchical self-organizing cluster vector quantization

A method for encoding a hyper-spectral image datacube using vector quantisation, wherein the encoded hyper-spectral data is compressed data, the method comprising the steps of:

• a) determining a plurality of codevectors;
• b) encoding each spectral vector of the hyper-spectral data by determining a codevector within the plurality of codevectors that approximates the spectral vector within the hyper-spectral data;
• and,
• c) repeating the steps of:
• determining another plurality of codevectors relating to a subset of codevectors within a previously determined codebook, encoding each spectral vector within the hyper-spectral data that is associated with codevectors within the subset using the other plurality of codevectors,
• until a desired number of codevectors exist.

A system for encoding a hyper-spectral image datacube using vector quantisation, wherein the encoded hyper-spectral data is compressed data, the system comprising:

• a first port for receiving the hyper-spectral image data;
• a processor for:
• a) determining a codebook having a plurality of codevectors;
• b) encoding each spectral vector of the hyper-spectral data by determining a codevector within the codebook that approximates the spectral vector within the hyper-spectral data;
• c) determining a fidelity of a cluster of vectors within the encoded hyper-spectral data;
• and,
• d) when the fidelity of a cluster of spectral vectors encoded by a codevector is below a predetermined fidelity, determining another codebook based on the cluster of spectral vectors, encoding each spectral vector of the cluster ofspectral vectors within the hyper-spectral data that is associated with codevectors within the subset using the other codebook, and returning to step (c);
• memory for storing data during execution of steps a) to d);
• and,
• a second port for providing the encoded hyper-spectral data.

A system for encoding a hyper-spectral image datacube using vector quantisation, wherein the encoded hyper-spectral data is compressed data, the system comprising:

• a first port for receiving the hyper-spectral image data;
• a processor for:
• a) determining a plurality of codevectors;
• b) encoding each spectral vector of the hyper-spectral data by determining a codevector within the plurality of codevectors that approximates the spectral vector within the hyper-spectral data;
• and,
• c) repeating the steps of:
• determining another plurality of codevectors based on spectral vectors within the hyper-spectral data and associated with a codevector, encoding each spectral vector within the hyper-spectral data that is associated with the codevector using the other plurality of codevectors,
• until a desired number of codevectors exist within each of the pluralities of codevectors;
• memory for storing data during execution of steps (a) to (c);
• and,
• a second port for providing the encoded hyper-spectral data.