It is often challenging to measure the sizes of small particles from greyscale TEM images.
Similar particles can have very different intensities as a result of diffraction contrast, while the
carbon support has a thickness that changes locally, and variable contrast that affects the mean
intensity values of the particles. Conventionally, a greyscale image is converted into a binary
image, in which pixels with a greyscale value higher than a chosen threshold are separated from
those with values below the threshold. This transformation is applied to the whole picture,
using a single threshold value. This procedure rarely works across an entire image. Here, we
propose using an adaptative threshold that is automatically adjusted in different parts of an
image. Figure 1 shows a comparison between the two approaches.