A comparison of reconstruction methods for undersampled atomic force microscopy images.

Area of Science:

• Atomic Force Microscopy
• Image Processing
• Signal Recovery

Background:

• Non-raster scanning and undersampling in Atomic Force Microscopy (AFM) enhance imaging rates and minimize tip-sample interactions.
• Image reconstruction from undersampled data relies on interpolation or optimization techniques, significantly impacting final image quality.
• The optimal reconstruction method is sample-dependent, necessitating a comparative analysis.

Purpose of the Study:

• To compare the efficacy of inpainting algorithms (interpolation) versus basis pursuit (optimization) for reconstructing non-raster AFM images.
• To evaluate reconstruction performance across diverse sampling patterns: row subsampling, spiral, Lissajous, and random scanning.
• To develop a classification strategy for selecting the appropriate reconstruction algorithm based on image characteristics.

Main Methods:

• Subsampling existing AFM images using four distinct non-raster patterns.
• Reconstructing images using both inpainting algorithms and the basis pursuit algorithm.
• Classifying image frequency content and sparsity using support vector machines (SVMs).
• Developing and testing a decision strategy for algorithm selection based on SVM classification.

Main Results:

• Inpainting generally yields superior results for images with predominantly low-frequency content.
• Basis pursuit demonstrates better performance for images with sparse, mixed-frequency content.
• A classification-based decision strategy effectively guides the selection of reconstruction algorithms for subsampled AFM data.

Conclusions:

• The choice between inpainting and basis pursuit for non-raster AFM image reconstruction is contingent on the image's frequency content and sparsity.
• Support vector machine classification enables a data-driven approach to optimize reconstruction method selection.
• This work provides a framework for improving the quality and efficiency of AFM image acquisition and processing.