Automatic White Blood Cell Segmentation Based on Color Segmentation and Active Contour Model

The identification of white blood cells (WBC) is essential since it can be used to diagnose various diseases. But, there are many complex features in the image of blood smear associated with signs of disorder in white blood cell. So, an accurate and reliable method to detect white blood cell diseases is needed. This study proposes an automatic WBC segmentation method using color-based segmentation with active contour model as the final step. We analyze the effectiveness of utilization various color spaces including RGB, HSV, YC
_B
C
_R
, CieLab, and grayscale in the preprocessing stage of the image. We do analyze 352 images in the dataset which consist of mononuclear and polynuclear WBC images. Since we do not need to analyze all of the existing color space elements, the method of channel selection of color space might be the solution to make the computation more efficient. The experimental results indicate that HSV color space is suggested as the most suitable color space for WBC image segmentation.