This study proposes a novel approach for automating the identification of diplococci in microscopic images by utilizing adapted image descriptors designed to capture microbial samples’ distinct structural and morphological features. The method focuses on analyzing live, non-fixed samples, ensuring the descriptors are optimized for the dynamic and intricate nature of microbial imaging. By integrating these specialized descriptors, the approach not only improves classification accuracy but also provides interpretable features that facilitate understanding of the underlying biological patterns. To validate the effectiveness of this technique, an annotated dataset was developed and used to benchmark the performance against multiple classification algorithms, including models based on deep learning. Comparative results showed that the proposed solution demonstrated outstanding performance, achieving a precision of 0.898, a recall of 0.933, and an F1-score of 0.915. These findings emphasize the potential of this methodology to advance automated diagnostics in microbiology while maintaining computational efficiency and interpretability.