<p>Complete blood cell (CBC) classification is one of the primary biomedical examinations of a human physiological condition. Conventional examination methods are susceptible to errors, while modern biological image processing techniques offer an efficient approach to CBC classification. To enhance the accuracy of blood cell classification, numerous image sensing techniques based on machine learning (ML) and deep learning (DL) have been developed in recent years. Among such techniques, the YOLO algorithms have demonstrated significant potential in object interpretation in the images with their remarkable accuracy and speed. In this work, an image processing technique with YOLOv8 framework is employed with two key enhancements to classify the blood cells of the blood smear images. The first enhancement is to include Convolution layers with a Transformer (C3TR) module in the backbone, enabling the precise detection and analysis of complex biological cell structures. It also stabilizes the image by removing pre-existing data noise, shifting focus from surroundings which are important advancements in biomedical imaging. The second is to incorporate the Small Object Detection (SOD) module in the neck to aid in accurate detection and localization of minute image features with precise spatial context. With these improvements, the proposed framework showed superior performance on the BCCD-derived dataset with 98.0% accuracy, 96.0% precision, and 96.0% recall compared with the existing models over the BCCD or its derived datasets. The proposed framework also operates up to three times faster than most CNN-based models, while maintaining a tiny architecture with only 11.2 million parameters. Furthermore, the existing image processing models classify only 3 or 5 types of blood cells. This work overcomes the limitations of various state-of-the art models with comprehensive 7-cell classifications with enhanced accuracy and machine vision applicability.</p>