Conventional blood analysis using blood smear image were performed manually by experts in hematology is tedious and highly depending on the level of experience. Currently, computer-assist technology is developed to reduce the time-consuming process and improved accuracy. As an example, various image processing techniques used to quantify such as white blood cells (WBCs) morphological conditions or classification in the blood smear image, which assist experts in developing confidence decision making in the analysis of cells conditions linked to the specific diseases. However, the WBCs shape features are arbitrary than the red blood cells (RBCs) because of the maturation state, cell orientations or positions, cell color variations, and the quality of the image captured influences the performance of classification accuracy. Therefore, we proposed a scale and rotation invariance feature for WBCs classification using speed up robust feature (SURF). SURF is suitable to be applied in identifying objects even though the orientation, scale, and position are varying, such as WBCs in microscopic blood smear images. We analyzed the classification performances using a support vector machine (SVM) and an artificial neural network (ANN) of WBCs types in the microscopic image based on the cell nucleus. The results show that the purposed SURF feature method has an excellent performance of accuracy for both methods and suitable to be utilized for the application of cell types classification.
