A novel method is proposed for the accurate isolation of the erythrocytes infected with plasmodium parasites in images acquired from Giemsa-stained blood samples using conventional light microscopes for diagnosis of Leukemia . As Leukemia is an infectious disease so rapid diagnosis is necessary for accurate medication. The automated system is designed to positively isolate Leukemia parasites in microscopic images. We applied image processing techniques namely morphological operations, edge detection, image normalization and image segmentation. The robustness and effectiveness of our method have been assessed through the comparison with more than thirty microscopic blood images. The experimental results show that the proposed method is more robust, highly accurate in true isolation of infected cells, extremely fast and simple to execute. This work is useful in medical applications and can help to identify Leukemia disease and construct a computer- aided diagnosis system in rural areas where health care manpower is limited. Future work is outlined.