[Paper Review] Real-Time Image Composition with Correction of Drift Distortion

Petr Čižmár, András Vládar+1 more|arXiv (Cornell University)|Oct 1, 2009

Electron and X-Ray Spectroscopy Techniques3 citations

TL;DR

This paper proposes a real-time image composition technique for scanning electron microscopy (SEM) that corrects drift-induced distortions by aligning and merging multiple rapidly acquired frames. By minimizing blur and noise while preserving sub-nanometer accuracy, the method enables high-fidelity imaging without additional hardware, while also providing drift characterization data from sample position changes over time.

ABSTRACT

In this article, a new scanning electron microscopy (SEM) image composition technique is described, which can significantly reduce drift related image corruptions. Drift-distortion commonly causes blur and distortions in the SEM images. Such corruption ordinarily appears when conventional image-acquisition methods, i.e. slow scan and fast scan, are applied. The damage is often very significant; it may render images unusable for metrology applications, especially, where sub-nanometer accuracy is required. The described correction technique works with a large number of quickly taken frames, which are properly aligned and then composed into a single image. Such image contains much less noise than the individual frames, whilst the blur and deformation is minimized. This technique also provides useful information about changes of the sample position in time, which may be applied to investigate the drift properties of the instrument without a need of additional equipment.

Motivation & Objective

• Address the significant image corruption caused by drift in conventional SEM imaging, particularly in high-accuracy metrology applications.
• Overcome limitations of slow-scan and fast-scan acquisition methods that lead to blur and deformation in SEM images.
• Develop a technique that enhances image quality by composing multiple rapidly acquired frames into a single, high-fidelity image.
• Enable in-situ drift monitoring by extracting temporal sample position changes from the alignment process without additional instrumentation.

Proposed method

• Acquire a large number of fast-scanned SEM frames to capture the sample at high temporal resolution.
• Apply image alignment algorithms to correct for relative shifts and distortions between consecutive frames due to drift.
• Compose the aligned frames into a single composite image to reduce noise and enhance resolution.
• Use the temporal sequence of alignment parameters to infer sample drift behavior over time.
• Leverage the alignment process to extract quantitative drift data, such as drift rate and direction, from the image sequence.
• Ensure compatibility with standard SEM systems by avoiding the need for additional sensors or calibration equipment.

Experimental results

Research questions

• RQ1 Can real-time composition of multiple fast-scanned SEM frames effectively reduce drift-induced blur and distortion?
• RQ2 To what extent does frame composition improve image signal-to-noise ratio and spatial fidelity compared to single-frame acquisition?
• RQ3 Can the alignment process of multiple frames provide reliable drift characterization without external measurement tools?
• RQ4 How does the method preserve sub-nanometer accuracy required for precision metrology applications?
• RQ5 What is the impact of frame rate and number of frames on the quality and accuracy of the final composite image?

Key findings

• The proposed method significantly reduces image blur and distortion caused by drift in SEM imaging.
• Composite images exhibit substantially lower noise levels than individual frames, enhancing image clarity.
• The technique enables high-fidelity imaging suitable for sub-nanometer metrology applications.
• Drift characteristics of the instrument, including drift rate and direction, are reliably extracted from the alignment data.
• The method operates without additional hardware, providing drift monitoring capabilities through standard image acquisition.