Quantitative analysis of white blood cells (WBC) or leukocytes is crucial in the diagnosis of various blood diseases including blood cancers like Leukemia and Myeloma. They are an integralconstituent of the immune system and the production of excess white blood cells is one of the initial steps in the defence mechanism by the human body against any disease. An abnormal WBC count or an increase in one type of white blood cells can be an indication of an infectious or inflammatory diseases. Hence, an analysis and identification of the different blood cells in a patient’s blood is pivotal for an early diagnosis of diseases. The manual methods adopted for this purpose are time-consuming and vulnerable to errors. Automating this process involves careful feature extraction for an increased accuracy. In this research, Convolutional neural network (CNN) models are exploited for automatic feature extraction and the images are classified using traditional classifiers such as Linear discriminant analysis (LDA) and Logistic Regression (LR). The different hybrid models obtained by using VGG19, Inception V3 and DenseNet169 with the above mentioned classifiers are evaluated and analysed to find the most suitable approach for this application. DenseNet169 – LDA model was shown to have the best performance in classifying the white blood cell images into neutrophils, lymphocytes, monocytes and eosinophils.