Much of our knowledge concerned with microbial cells is based on population-based analysis of cultures, which give useful insights into average responses but neither on individual cells nor subpopulations. In this work we demonstrate how to access and utilise large amounts of valuable information concerned with cell populations contained in laser scanning microscopy images. To this aim we carried out quantitative characterization of selected strains of Saccharomyces yeast by image and statistical analysis of the Laser Scanning Microscopy images. Features such as cell size, entropy and intensity of the cell ensembles were extracted and analysed using a method appropriate for datasets with a high degree of variability. The empirical cumulative distribution functions (ecdfs) were compared using the Kolmogorov-Smirnov test, which confirmed that the ecdfs for size, intensity and entropy were statistically different for each of the studied strains at standard confidence levels. Further, we used this technique to investigate the evolution of cell features with culture age between 24 h and 72 h, the latter corresponding to a stationary phase. Moreover, in mixed cultures we were able to estimate the fraction of each pure strain, within about 5% accuracy. We thus demonstrate how the information from the cell ensembles can be extracted by data mining of microscopy images and utilised to support objective judgements about strain identity and differentiation.