Adapted Generalized Unsharp Masking Algorithm for Sharpness Improvement of Scanning Electron Microscopy Images

Abstract

Scanning electron microscope (SEM) images are highly valuable in different scientific applications because they can depict extremely small entities. SEM images are sometimes obtained blurry, in that such an issue reduces the clarity and hampers the detection of important features in the image. One way of processing the unwanted blurring effect is to use image sharpening, which aims at emphasizing the edges so that the output appears more significant to the observer with better-highlighted details. Many image sharpening methods exist, but not all are efficient as they may introduce artifacts, unnatural appearance, contrast/brightness modifications, or can be complicated and require a high computational cost. One algorithm of interest is the generalized unsharp mask (GUSM), which avoids the generation of artifacts that many sharpening methods own and have a somewhat simple structure. Still, when the GUSM algorithm is applied to different SEM images, it provides an unnatural sharpness and modifies the contrast and brightness as well. This is undesirable because proper sharpening is required for SEM images as they depict important information. Hence, an adapted GUSM algorithm is introduced in this article in that it provides a more natural sharpening without modifying the brightness or contrast of the filtered images. The main contribution of this study is to remove the contrast enhancement procedure and replace the smoothing process to deliver more natural sharpness. The developed AGUSM algorithm is verified with different real-unclear SEM images, its performance is appraised against different image sharpening methods, and the outcomes of comparisons are evaluated by utilizing advanced metrics. For the performed experiments, the AGUSM provided satisfying performances as the outcomes appear to have more acutance and look more natural when compared to the original counterparts and the outcomes of the comparison methods.