Software approaches to hyperspectral data compression
The invention comprises proprietary techniques for data compression of hyperspectral datacubes, which are typically 500 MB in size. Data compression is useful for processing, transmitting and/or storing these large data files.
Vector Quantization (VQ) can provide high compression ratios based on relatively simple structures. The technique provides an opportunity for a data compression system that takes advantage of correlation in the spectral domain and the 2-D spatial correlation between adjacent image pixels.
The CSA has developed two innovative hyperspectral data compression techniques: Successive Approximation Multi-Stage Vector Quantization (SAMVQ) and Hierarchical Self-Organizing Cluster Vector Quantization (HSOCVQ). For both techniques, a new codebook is generated for each datacube that is to be compressed, and this codebook is transmitted as overhead with the compressed data. Both of these techniques use very small codebooks to generate this first approximation and in this respect, they are refinements over the general VQ procedure. HSOCVQ and SAMVQ provide two different methods for moving from this first approximation to a more refined and compressed data file that more accurately represents the spectral vectors.
Both SAMVQ and HSOCVQ can compress data in a manner that is near-lossless, if the fidelity threshold parameter is set to a value such that the error induced during the compression is at the same level as the intrinsic noise in the original data. This property is usually not available in conventional compression techniques. The techniques are implemented in software, using conventional programming tools.
These innovative data compression techniques have potential to be used in other areas where compression of multi-dimensional data is beneficial, such as three-dimensional medical imagery (MRI, positron emission tomography, enhanced ultrasound and computed tomography modalities), enabling faster processing times when used in conjunction with the CSA's hyperspectral browser technology. In principle, the technology may find application in any field in which the data have three dimensions (two spatial dimensions and a third dimension that has other information). Such applications include optical, radar and other forms of images.
The Business Opportunity
The invention has direct application to the processing of hyperspectral datasets from spaceborne or airborne remote sensing. The invention can improve the processing time of datasets for end-users in government, university and industry (e.g., in agriculture, mining and forestry). An additional application is archival storage of large three-dimensional datasets in a compressed form. The technology will be of interest to companies that provide hyperspectral image analysis software or customized data processing services for end-users.
Technology Transfer Details
These technologies are available for licensing.
The business opportunity may be referred to by its CSA case ID: 50525-50528
- Number: 6,724,940
- Number: 6,701,021
- Status: Issued Patents
- Source: United States Patent and Trademark Office (USPTO)
- Number: EP1209627 (A2)
- Status: Pending patent application
- Source: European Patent Office (EPO)
- Date modified: