Visual Information Representation Using Sparse Dictionary Ensembles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge of handling and processing large data sizes in high-dimensional visual information, such as light field data, is exacerbated by inefficient storage and infrastructure requirements, necessitating improved data compression and reconstruction methods.
Innovation Solution
A computer-implemented method that transforms n-dimensional visual information data elements into a transformation domain using an ensemble of dictionaries and basis functions, sparsifies the coefficients, and compresses them to achieve efficient storage and reconstruction, utilizing non-local clustering and machine-learning assisted dictionary creation for improved compression and selective reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional data compression methods are used for high-dimensional visual information, then storage requirements are reduced, but compression rate is insufficient and reconstruction error increases
Solution Approach 1:
The patent segments high-dimensional visual information into multiple data elements (e.g., angular slices in light field data) and processes each element separately through dictionary assignment and transformation. This segmentation enables more efficient compression while maintaining reconstruction quality by applying specialized basis functions to each segment.
Solution Approach 2:
The patent transforms data from the spatial domain to a transformation domain using learned basis functions, changing the representation parameters. This transformation reveals sparse structures in the data that enable more effective compression while preserving essential visual information for accurate reconstruction.
2Quantity of substance
If data compression is applied to reduce storage requirements, then storage efficiency improves, but compression rate remains insufficient for very large data sets
Solution Approach 1:
The patent performs preliminary actions by learning ensembles of dictionaries from training data before actual compression. These pre-learned basis functions are then applied to new data, enabling faster compression rates without sacrificing compression effectiveness. The preliminary dictionary learning phase creates reusable transformation tools.
Solution Approach 2:
The patent creates compressed representations (copies) of the original high-dimensional data in a transformation domain. These compressed copies retain essential visual information while occupying significantly less storage space, achieving both reduced storage requirements and high compression rates.
3Device complexity
If conventional compression methods are used, then infrastructure requirements are reduced, but the system cannot handle very large data sizes efficiently
Solution Approach 1:
The patent transforms data from n-dimensional space to an m-dimensional transformation domain where m ≥ n. This dimensionality change allows the data to be represented in a form that is more amenable to compression and can be processed with reduced infrastructure requirements while handling very large data sizes effectively.
4Measurement precision
If high-dimensional visual information is stored without compression, then reconstruction quality is maintained, but storage and processing efficiency deteriorates
Solution Approach 1:
The patent introduces an intermediary transformation domain that bridges the original spatial domain and the compressed representation. Basis functions act as intermediaries that transform data into a sparse representation, enabling efficient compression while preserving visual information quality for accurate reconstruction when needed.
Data Source
AI summary
The present disclosure relates to a computer implemented method for representing a data set comprising at least one n dimensional data element representing visual information, said method comprising obtaining (210) said data set, obtaining (220) a dictionary ensemble comprising a plurality of dictionaries each comprising at least one basis function (102), assigning (230) each at least one data element to a dictionary, wherein a set of basis functions represents an m dimensional transformation domain, transforming (240) the at least one data element with the corresponding dictionary of basis functions to the transformation domain wherein each data element is defined by an associated coefficient set, sparsifying (250) the coefficient sets, forming (260) the representation of the visual information comprising a coefficient data set and the corresponding dictionaries.