Biomedical Image Segmentation Based on Shape Stability

Biomedical image segmentation remains a challenging task mainly due to the weak edges and unevenly distributed color intensity of the objects and background. We present a novel unsupervised segmentation method to extract nuclei region from background. Our method, called shape stability algorithm, is a multiscale local adaptive threshold method. A modified weighted filter which serves as preprocessing method is also introduced. The presented algorithm is applied for segmentation of a number of Pap Smear images as well as bone marrow cell images. The results indicate the successful performance of the presented segmentation algorithm in segmentation of both Pap Smear and bone marrow samples.