Hyperspectral Image Compression Using Orthonormal Basis Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing compression techniques for hyperspectral image data either result in loss of valuable information or fail to achieve significant reduction in data size, leading to latency issues and inefficiencies in processing and transmission.
Innovation Solution
A method that utilizes orthonormal basis vectors to compress hyperspectral image data by selecting a minimal set of coefficients based on error magnitude and data size, allowing for efficient reduction of data volume while preserving information within a specified error or data size constraint.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If lossy compression methods are applied to hyperspectral data sets, then data size is reduced, but valuable information is removed that is needed for computer or mathematical processing
Solution Approach 1:
The patent transforms the hyperspectral data from its original form into a different parameter space using orthonormal basis vectors. By changing the representation parameters (from pixel values to coefficients in an orthonormal basis), the data maintains all information content while achieving compression. The transformation parameters (basis vectors) are computed once and reused, enabling lossless compression with significant data size reduction.
2Loss of information
If general purpose lossless compression algorithms are used on hyperspectral images, then data integrity is maintained, but significant compression is not achieved and decompressed data may be larger than original
Solution Approach 1:
Instead of applying general-purpose lossless compression to the raw hyperspectral data, the patent changes the parameters by transforming the data into an orthonormal basis representation. This parameter transformation reveals the inherent structure and redundancy in hyperspectral data, enabling much more effective compression while maintaining complete information integrity.
Solution Approach 2:
The patent creates a compressed representation (copy) of the hyperspectral data using orthonormal basis vectors and coefficients. This compressed copy contains all the information of the original data but in a more compact form, achieving lossless compression with significantly reduced data size compared to the original hyperspectral image.
3Productivity
If hyperspectral sensors generate large amounts of data at high transmission rates, then more information is captured, but latency problems occur in near real-time processing
Solution Approach 1:
The patent applies orthonormal transformation to change the parameter representation of the hyperspectral data, which reveals the underlying structure and enables efficient compression. This parameter change allows the system to handle large data volumes from high-rate sensors by reducing the amount of data that needs to be transmitted and processed in real-time, thereby reducing latency while maintaining full information content.
Data Source
AI summary
Methods for compressing hyperspectral image data include receiving sets of coefficients associated with each pixel of the hyperspectral image data, a set of basis vectors utilized to generate the dimensionally reduced data from the hyperspectral image, and either a maximum error value or maximum data size. The methods include associating the coefficients with a subset of the basis vectors, and storing the association. Methods of decompressing the compressed hyperspectral image data are also disclosed, utilizing the association.