Roundness curve for classification of cell phase on microscopic image

Object shape recognition is important for cell microscopic image. Previous works have used circularity to describe the shape feature for different objects. In this paper, we propose an alternative shape feature, roundness curve, to represent the shape of object and classify cell microscopic image. The roundness curve is extracted based on the central projection method. Cell images are first segmented by applying the mathematical morphology method to binary images, which are converted from original microscopic images. Then, centroid of each object is calculated. The centroid projects radials to acquire projection value in pre-defined directions. The roundness curve in an orthogonal coordinate system is set up with the projection values and their indices on its two axes respectively. Experimental results prove our proposed method the effectiveness and robustness, it can be used to recognize and count cells of different phases using high-content screening images.