Neural Upsampling of Decompressed Time-Series Data for SAR Quality Recovery

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current data compression methods for complex-valued SAR images, such as those used in synthetic aperture radar, face challenges in preserving image quality due to noise sensitivity and limited real-time processing capabilities, particularly in compressing phase images which require high bits-per-pixel, leading to impractical compression and potential loss of information.

Innovation Solution

A neural network-based system that incorporates a novel AI deblocking network with recurrent layers for feature extraction and a channel-wise transformer with attention to capture complex inter-channel dependencies, enabling effective upsampling of decompressed time-series data and improving compression quality by mitigating compression artifacts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If lossy compression is applied to SAR images to reduce bandwidth and storage requirements, then compression ratio is improved, but image quality and information preservation deteriorate

Engineering Contradiction:

Improvedata volumeVSAvoidinformation loss

Core Design Contradiction:

Quantity of substanceVSLoss of information

Solution Approach 1:

A neural network upsampler is introduced as an intermediary component between the lossy compressor and the final output. This mediator recovers lost high-frequency information by learning from correlated time-series datasets, enabling the system to achieve both high compression ratios and high fidelity reconstruction of SAR images

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system performs preliminary compression with aggressive lossy algorithms first, then applies neural network upsampling as a post-processing step. This two-stage approach allows the system to benefit from high compression ratios while subsequently recovering the information that was lost during compression

Inventive Principle:

Principle #10Preliminary action

2Ease of operation

If conventional compression algorithms are used for phase images, then ease of operation is improved, but compression efficiency deteriorates due to noise sensitivity

Engineering Contradiction:

Improveease of compressionVSAvoidcompression efficiency

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The system changes the operational parameters by transitioning from traditional pixel-based compression to neural network-based compression that operates in the feature space. The neural network learns optimal compression parameters automatically from training data, adapting to the noise characteristics of phase images while maintaining compression efficiency

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces conventional mechanical compression algorithms with a neural network-based system. The neural network substitutes traditional signal processing operations with learned representations, enabling efficient compression of noise-sensitive phase images through pattern recognition rather than fixed algorithmic operations

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Manufacturing precision

If high bits-per-pixel are used for phase image compression to maintain quality, then image quality is improved, but device complexity and processing requirements worsen

Engineering Contradiction:

Improveimage qualityVSAvoidprocessing complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The compression system is segmented into two distinct stages: a first stage using conventional lossy compression for bulk data reduction, and a second stage using neural network upsampling for quality recovery. This segmentation allows each component to be optimized independently, avoiding the need for a single complex high-bit-rate compressor

Inventive Principle:

Principle #1Segmentation

4Productivity

If lossy compression is applied to time-series data, then productivity is improved through faster processing, but measurement precision deteriorates due to information loss

Engineering Contradiction:

Improveprocessing speedVSAvoiddata accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The neural network upsampler uses feedback from correlated time-series datasets to recover lost information. By learning temporal and spatial correlations from multiple datasets, the system can predict and reconstruct missing high-frequency components, maintaining measurement precision while enabling fast lossy compression

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12068761B1System and methods for upsampling of decompressed time-series data using a neural network

Publication Date: 2024.08.20 ATOMBEAM TECH INC

AI summary

A system and methods for upsampling of decompressed time-series data after lossy compression using a neural network that integrates AI-based techniques to enhance compression quality. It incorporates a novel AI deblocking network composed of recurrent layers for feature extraction and a channel-wise transformer with attention to capture complex inter-channel dependencies. The recurrent layers extract multi-dimensional features from the two or more correlated datasets, while the channel-wise transformer learns global inter-channel relationships. This hybrid approach addresses both local and global features, mitigating compression artifacts and improving decompressed data quality. The model's outputs enable effective data reconstruction, achieving advanced compression while preserving crucial information for accurate analysis.