Automation of Supported Nanoparticle Recognition in Low Contrast STEM Images

In this study, high angular annular dark field (HAADF) scanning transmission electron microscopy (STEM) results in Z-contrast, ultimately mass-thickness effects will also contribute to the relative contrast within images. In images with multiple species, such as supported nanoparticles. we must consider measurable contrast of a singular object (nanoparticle) as the relative difference in intensity between the closest contour outside the object and the contour of the object. The low contrast images, neither manual thresholding nor using the multi-Otsu algorithm are capable of finding the true contours,as shown. The sample investigated here consists of CoCuGa alloy nanoparticles with different sizes and composition on SiO2 support. This makes it a difficult system to analyze, especially with a fully automatic algorithm. HAADF-STEM image of CoCuGa alloy nanoparticles supported on SiO2.The method presented here occurs as follows: the set of all contours in an image is found by applying a threshold at all pixel values and determining the contours for each threshold. Within this set, the true contours corresponding to the morphology of the target nanoparticles are present, however they co-exist alongside contours from the support, background and noise. To separate the nanoparticles from the rest of the contours, a set of constraints is defined that are independent from the characteristics that are to be determined.