A comparative analysis of pairwise image stitching techniques for microscopy images.

Area of Science:

• Computational pathology
• Digital biology
• Microscopy image analysis

Background:

• Microscopic image stitching combines overlapping images into a whole slide image.
• Accurate pairwise registration is critical for stitching computational load and overall success.
• Selecting an efficient, accurate, and robust registration method is essential for whole slide imaging.

Purpose of the Study:

• To conduct a detailed comparative analysis of various pairwise registration techniques for microscopic image stitching.
• To evaluate these methods based on execution time and image quality.
• To identify the most effective technique for whole slide image creation.

Main Methods:

• Evaluated feature-based methods: Harris, Shi-Thomasi, FAST, ORB, BRISK, SURF, SIFT, KAZE, MSER, and SuperPoint.
• Analyzed region-based methods: Normalized Cross-Correlation (NCC) and Phase Correlation with NCC.
• Tested methods on bright-field, phase-contrast, and fluorescence microscopy images.

Main Results:

• Feature-based methods demonstrated high robustness against uneven illumination.
• Some feature-based methods outperformed region-based methods in accuracy and speed.
• SURF features were identified as the most effective pairwise registration technique.

Conclusions:

• The choice of pairwise registration method significantly impacts whole slide image quality and creation efficiency.
• SURF features offer a superior balance of speed, accuracy, and robustness for microscopy image stitching.
• This research provides crucial insights for advancing computational pathology and biology through improved whole slide imaging techniques.