Automatic blood cell classification based on joint histogram based feature and bhattacharya kernel

Sharmin Nilufar, Nilanjan Ray, Hong Zhang

Abstract: We propose a blood cell classification method with the aim of designing an automatic differential blood count system, which can help cancer diagnosis. The proposed system contains two automated steps: an active contour-based segmentation of blood cells from microscopy images and their classification. For classification we investigate several joint histogram-based features extracted from the segmented blood cells. We use support vector machine with a proposed kernel based on the Bhattacharya coefficient of joint histograms. Experimental results show the effectiveness of our system. Furthermore, comparative study illustrates that the proposed system outperforms other existing classification approaches in terms of classification accuracy.