Mask Optimization for High-Precision Extraction of Geometric Features in Microscopic Scenes.

Area of Science:

• Microscopic imaging
• Geometric feature extraction
• Image segmentation

Background:

• Regular geometric targets (microspheres, micropores, microtubes) in microscopic scenes present challenges due to small scales, low contrast, and degraded boundaries.
• Existing segmentation methods often produce masks unsuitable for high-precision geometric parameter measurement.

Purpose of the Study:

• To propose and validate a mask optimization method for high-precision extraction of regular geometric features in microscopic scenes.
• To improve the accuracy and robustness of geometric measurements from microscopic images.

Main Methods:

• A mask optimization framework integrating initial mask generation and geometric consistency refinement.
• Mask initialization using segmentation and adaptive super-resolution (SR) under low annotation constraints.
• Iterative optimization fusing multi-dimensional pixel features with regular geometric priors for mask correction.

Main Results:

• The proposed method achieved lower geometric errors and significantly improved the fitting success rate on a sphere-tube assembly dataset.
• Ablation studies confirmed the critical roles of dynamic SR and iterative mask optimization in enhancing precision and stability.
• Generated target masks exhibited continuous boundaries satisfying stringent geometric constraints.

Conclusions:

• Integrating geometric-consistency constraints into mask optimization is effective for microscopic regular geometric measurement tasks.
• The method enhances both the accuracy and robustness of geometric feature extraction in challenging microscopic imaging scenarios.