Image analysis with rapid and accurate two-dimensional Gaussian fitting.

Area of Science:

• Image Analysis
• Computational Physics
• Scientific Imaging

Background:

• Accurate particle localization in images is crucial for scientific research.
• Existing methods like least-squares fitting are precise but too slow for large datasets.
• Faster approximations often sacrifice significant precision.

Purpose of the Study:

• To develop a computationally rapid image analysis method for 2D Gaussian fitting.
• To achieve subpixel resolution for particle localization.
• To improve upon the speed and precision of existing methods for large-scale image analysis.

Main Methods:

• Weighted overdetermined regression for 2D Gaussian fitting.
• Analysis of particle location from pixelized light intensity images.
• Quantitative comparison with least-squares, nine-point regression, and moment methods.

Main Results:

• The new method offers over 100-fold improvement in computational time compared to least-squares fitting.
• Precision is equivalent to least-squares at high signal-to-noise ratios and approaches it at low ratios.
• Improved precision and speed compared to nine-point regression and moment methods for both circular and elliptical distributions.

Conclusions:

• Weighted overdetermined regression provides a computationally efficient and precise alternative for 2D Gaussian fitting.
• This method is suitable for large datasets in fields like astronomy and microscopy.
• Facilitates real-time quantification of microscopy images and advances in particle tracking.