A framework for improvement in homogeneity of fluorescence and bright field live cell images using fractional derivatives

Cell segmentation has been an important area in modern biological image processing applications. The most commonly used cell segmentation algorithms are region based and rely on the homogeneity value of the image intensities in the region of interest to be segmented. But the highly inhomogeneous behavior of cell region and background causes feature overlapping between the two leading to misclassification and poor segmentation results. This paper proposes a method to improve the homogeneity of the cell images. The existing clustering criterion for bias correction has been improved upon by introducing fractional differential in the algorithm. The proposed method has been tested on two different sets of 2D cell images, and improved performance results over the existing method are obtained.