An optimum or a high heat treatment of milk before yogurt preparation was applied according to an experimental design with the deliberate aim of making stirred yogurts with varying properties. Images of sections of yogurt samples were obtained using a transmission electron microscope. These were digitized and segmented via a threshold determination to separate the protein structure from the background. Four spatial statistical estimates were determined in order to evaluate their applicability for characterization of yogurt microstructure. The volume content was estimated and compared with the protein content obtained through chemical analysis. Heat treatment had no significant effect on volume content, as expected. The surface area of the protein aggregates was a useful measure to describe the difference between casein networks obtained by an optimum and a high heat treatment. The star volume gave a measure of the average local size of pores in the yogurt gel. The optimum heat treatment resulted in yogurt with small values for the star volume, indicating a less coarse structure of the casein network compared with yogurt prepared from high heat milk. The covariance function was able to differentiate between yogurts made using two types of heat treatment.