Color and morphological based techniques on white blood cells segmentation

The conventional way of leukemia diagnosis is microscopic observation of the blood slide by hematologists which can be tedious, time consuming and expensive due to the complex contents of the blood slide and the expertise required. This paper proposes a combination of color and morphological based segmentation techniques for blood cells segmentation on microscopic images. K-means clustering is implemented for color based clustering and primary segmentation of the image. The clustering is cascaded into two layers for cell recognition and background elimination. Morphological based segmentation is conducted using watershed transform based on gradient magnitude and skeleton by influence zone. The proposed method overcomes the problem of previous work which is over-segmentation and efficiently segments the WBC cells and keep cytoplasm intact.