Evaluation of Focus Measures for Hyperspectral Imaging Microscopy Using Principal Component Analysis

Area of Science:

• * Microscopy and Imaging Science
• * Spectroscopy and Data Analysis

Background:

• * Automatic focusing systems are vital for automated microscopy, optimizing lens-to-object distance by maximizing focus measure (FM) values.
• * Hyperspectral imaging systems (HSIS) require reliable autofocus methods for accurate chemical and spatial information extraction from hyperspectral datacubes.
• * Traditional FM evaluation relies on visual sample features, which are often insufficient in HSIS due to a lack of human-comprehensible visual information.

Purpose of the Study:

• * To develop and evaluate reliable autofocus methods for hyperspectral imaging systems.
• * To establish a benchmark for assessing the performance of various focus measures in HSIS.
• * To identify the most effective focus measures for hyperspectral image autofocusing.

Main Methods:

• * Development of autofocus methods tailored for hyperspectral imaging microscope systems.
• * Utilized Principal Component Analysis (PCA) to define optimal ("ideal") optical focus positions as a benchmark.
• * Evaluated 22 common focus measures using hyperspectral images across visible and near-infrared bands in four different HSIS setups with varying magnifications.

Main Results:

• * Principal Component Analysis (PCA) provided a reliable method for determining optimal focus positions in HSIS.
• * Evaluated 22 focus measures across diverse hyperspectral datasets and imaging configurations.
• * Gradient-based focus measures demonstrated superior speed and reliability compared to other methods.

Conclusions:

• * Gradient-based focus measures are the most effective autofocus operators for hyperspectral imaging systems.
• * The PCA-based benchmark provides a robust framework for evaluating autofocus performance in HSIS.
• * Reliable autofocusing is critical for enhancing the accuracy of chemical and spatial information retrieval in hyperspectral imaging.