A method based on rank-ordered filter to detect edges in cellular image

To extract morphological features about nuclei from microscopy cellular image, it is usually required to find the edges of nuclei at first. Standard edge detection methods may not produce satisfactory results due to the varying brightness and background in cellular image. It is important to extract close, smooth, and correct edges in order to compute features like compactness, convexity, roundness, and etc. We present a new method to detect edges of nuclei in microscopy images. The method is based on using median filtering to compute the total variation with respect to the central pixel in a filter window. This step exploits one important feature of median filter, i.e., within the filter window, the total variation (TV) with respect to the median is always less than or equal to the TV with respect to the original center pixel. In other words, median filtering looks for the output that minimizes the total variation within the filtering window. The resulting image has enhanced contrast along the boundary of nuclei. As the final step, we use popular edge detection methods such as Canny detector and Laplacian of Gaussian to find edges of nuclei in the image. Examples from processing real cellular image obtained by light microscope show that the method obtains better edges in terms of connectivity, smoothness, and closely following the boundary of nuclei.