Dark-field microscopic image stitching method for surface defects evaluation of large fine optics

Area of Science:

• Optical engineering
• Metrology
• Image processing

Background:

• Evaluating surface defects on large optics (tens/hundreds of mm) presents challenges, particularly for micron-sized defects.
• Sub-aperture scanning and stitching is a practical approach, but traditional methods struggle with sparse defect data, leading to image mismatch.

Purpose of the Study:

• To develop a robust image stitching algorithm for large optics with sparse surface defects.
• To overcome the mismatch problem in traditional stitching methods when few features are present in overlapping sub-aperture images.

Main Methods:

• A feature-based multi-cycle image stitching algorithm is proposed.
• Overlapping sub-aperture image areas are categorized by feature content.
• Stitching is performed in multiple cycles using different methods tailored to overlapping area types, mimicking a region-growing process.

Main Results:

• The proposed algorithm successfully stitches sub-aperture images even with minimal feature information in overlapping regions.
• It demonstrates high suitability for stitching dark-field microscopic images of large optics.
• The stitching trace adapts based on detected features, improving accuracy.

Conclusions:

• The feature-based multi-cycle stitching algorithm provides an effective solution for surface defect evaluation of large fine optics.
• This method enhances the accuracy and reliability of full-aperture image reconstruction in scenarios with sparse defect data.