An image-processing algorithm for morphological characterisation of soot agglomerates from TEM micrographs: Development and functional description

Abstract

The inspection of soot agglomerates from microscopy images usually relies on manual human measurements, whereas image processing tools for faster analysis are highly demanded. In this study, an automated algorithm for the extraction of morphological soot descriptors from transmission electron micrographs is presented. The proposed algorithm involves the detection of the image scale (the conversion from pixels to nanometres) using a Hough transform and an optical character recognition process. Primary particles are identified through a two-step circle Hough transform combining phase-coding and edge-based approaches, whereas size descriptors are obtained through spatial and frequency filtering. Finally, the fractal dimension is obtained for each agglomerate as a projected-area derived measurement due to an iterative process. Results were validated by comparison of a set of micrographs taken at three different magnifications with manual image processing, obtaining p-values greater than 0.05 and around 91.5% time saving.