On a Software Framework for Automated Pore Identification and Quantification for SEM Images of Metals

The functional performance of porous metals and alloys is dictated by pore features such as size, connectivity, and morphology. While methods like mercury porosimetry or gas pycnometry provide cumulative information, direct observation via scanning electron microscopy (SEM) offers detailed insights unavailable through other means, especially for microscale or nanoscale pores. Each scanned image can contain hundreds or thousands of pores, making efficient identification, classification, and quantification challenging due to the processing time required for pixel-level edge recognition. Traditionally, pore outlines on scanned images were hand-traced and analyzed using image-processing software, a process that is time-consuming and often inconsistent for capturing both large and small pores while accurately removing noise. In this work, a software framework was developed that leverages modern computing tools and methodologies for automated image processing for pore identification, classification, and quantification. Vectorization was implemented as the final step to utilize the direction and magnitude of unconnected endpoints to reconstruct incomplete or broken edges. Combined with other existing pore analysis methods, this automated approach reduces manual effort dramatically, reducing analysis time from multiple hours per image to only minutes, while maintaining acceptable accuracy in quantified pore metrics.