Hyperspectral Preview Image Selection for Real-Time Contrast
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for previewing hyperspectral data during acquisition require manual selection of frequency bands or rely on selecting bands with maximum intensity, which often fail to showcase the sample's chemical or physical properties of interest, lacking a fast and high-contrast live preview.
Innovation Solution
A computer-implemented method scores each single channel image based on contrast using convolutions with pre-defined filters, selects high-contrast images, and generates a live preview by applying augmentation methods to enhance the image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual selection of frequency bands is used for preview, then the preview can be customized to show specific chemical or physical properties, but the process requires a priori knowledge of the sample properties and cannot be performed in real-time during data acquisition
Solution Approach 1:
The system automatically selects and processes frequency bands without requiring manual intervention or a priori knowledge of sample properties. The automated contrast scoring mechanism self-evaluates each frequency band and selects the most informative ones for display, enabling real-time preview during data acquisition.
Solution Approach 2:
The system dynamically changes the selection criteria from static manual band selection to dynamic automated contrast-based selection. By calculating contrast scores for each frequency band and selecting bands with highest contrast, the system adapts to unknown sample properties in real-time.
2Productivity
If frequency bands with maximum intensity are selected for display, then the preview generation is simple and fast, but the selected bands often do not show the sample's chemical or physical properties of interest
Solution Approach 1:
The system changes the selection parameter from maximum intensity to maximum contrast. By calculating contrast scores that measure the distinguishability of features in each frequency band, the system selects bands that best showcase sample properties rather than simply the brightest bands.
Solution Approach 2:
The system replaces the simple mechanical approach of selecting maximum intensity bands with a more sophisticated contrast-based selection mechanism that evaluates the informational content of each frequency band, substituting brute-force intensity maximization with intelligent contrast optimization.
3Loss of information
If all channel images are processed and displayed to ensure comprehensive coverage of sample properties, then no information is lost, but the processing time increases and real-time preview becomes unachievable
Solution Approach 1:
The system extracts only the most informative frequency bands by calculating contrast scores and selecting the top-performing bands. This extraction approach retains the critical information needed for sample analysis while discarding redundant low-contrast bands, enabling real-time processing.
Solution Approach 2:
Instead of processing all channel images equally (excessive action), the system applies partial action by selectively processing only the highest contrast bands. This partial processing approach maintains information completeness for the most relevant properties while significantly reducing processing time.
Data Source
AI summary
Real-time preview of hyperspectral data during data acquisition is provided, in which the hyperspectral data represents chemical and/or physical properties of a sample and is represented by a set of channel images. For each channel image of the set, an edge score is determined using a plurality of edge kernels, a noise score is determined using a noise kernel, and a contrast score is determined for the channel image based on the ratio of the determined edge score to the respective determined noise score. Channel images are selected in descending order of associated contrast scores, as a plurality of high contrast candidate images. One or more augmented candidate images are generated by applying one or more augmentation methods, and a contrast score is determined for each augmented candidate image. The candidate image with the highest contrast score is provided as preview image during data acquisition.