Hyperspectral Image Destriping via Pixel Variance Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional hyperspectral imaging systems, particularly those using Mercury-Cadmium-Telluride focal plane arrays, suffer from sensor calibration artifacts that result in false alarms and reduced target detection accuracy due to residual gain and offset errors, manifesting as columns or striping artifacts in detection images.

Innovation Solution

A method and system for processing hyperspectral images that involves selecting regions with striped columns, computing a global mean, and adjusting pixel values within these regions to align with the mean, thereby reducing variance and minimizing false alarms by normalizing pixel values.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional HSI sensors are used for target detection, then imaging capability is provided, but sensor calibration artifacts create false alarms and reduce detection accuracy

Engineering Contradiction:
Improvetarget detection accuracyVSAvoidsensor calibration artifacts
Core Design Contradiction:
ReliabilityVSObject-generated harmful factors

Solution Approach 1:

The patent extracts and removes the harmful striping artifacts from the detection image by identifying abnormal pixels and replacing their values with averages from neighboring pixels, effectively separating the harmful artifacts from the useful target detection information

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent converts the harmful striping artifacts into beneficial information by using the abnormal pixel identification and replacement process to enhance the overall image quality and reduce false alarms, turning the calibration errors into an opportunity for artifact reduction

Inventive Principle:
Principle #22Blessing in disguise (Convert harm into benefit)

2Manufacturing precision

If linear time invariant approximation is used to estimate gain and offset for destriping, then some artifact reduction is achieved, but nonlinear effects of clipping introduce skew into calculations reducing effectiveness

Engineering Contradiction:
Improvedestriping accuracyVSAvoidcalculation complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent changes the approach from linear time invariant approximation to a method that accounts for nonlinear effects by using abnormal pixel identification and replacement, adjusting the processing parameters to handle clipping effects properly

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent uses neighboring pixels to create replacement values for abnormal pixels, copying information from adjacent valid pixels to restore the corrupted pixel values and reduce striping artifacts

Inventive Principle:
Principle #26Copying

3Manufacturing precision

If abnormal pixels are identified and replaced with average of neighboring pixels, then striping artifacts are reduced, but correct identification of abnormal pixels is required to avoid destriping errors

Engineering Contradiction:
Improveartifact reduction accuracyVSAvoidabnormal pixel identification
Core Design Contradiction:
Manufacturing precisionVSDifficulty of detecting and measuring

Solution Approach 1:

The patent implements a feedback mechanism where the detection filter output is analyzed to identify abnormal pixels, and the replacement process uses feedback from neighboring pixel values to iteratively reduce artifacts while maintaining image quality

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP2836984B1Post-detection artifact reduction and removal from images
Publication Date: 2018.07.11 RAYTHEON CO
  • EP2836984B1 patent drawingFigure 1
  • EP2836984B1 patent drawingFigure 2A~2C
  • EP2836984B1 patent drawingFigure 2D~2E

AI summary

In accordance with various aspects of the disclosure, a system, a method, and computer readable medium having instructions for processing images is disclosed. For example, the method includes selecting, at an image processor, a region of a first image comprising a plurality of pixels. A mean value of pixels in the selected region is computed. From a plurality of sets of pixels in the region, a first subset of pixels in the region containing artifacts therein is selected. A value of each pixel in the first subset is compared with the mean value. The value of each pixel is adjusted based upon the comparing. The first image is reconstructed based upon the adjusted value of each pixel in the first subset, such that a variance of pixel values in the reconstructed image is lower than a variance of pixel values in the first image.