Region Growing Segmentation of Chromatin Clumps of Ovarian Cells Using Adaptive Gradients

An iterative region growing algorithm for segmentation within the chromatin of ovarian cells is presented. The growing procedure starts with small seed regions that are relatively easier to segment accurately by an intensity thresholding process. In each of the iterations, the growth decision is made based on whether a given test is passed. The growth test is set to pass more easily in the first iterations when the intensities are low and the gradients are high. As the region grows, the test gradually becomes more difficult to pass. The growth stops when it reaches the boundaries where the pixel intensities are high and the gradients decrease to zero. To reduce the noise effects that may interrupt the region growth, a preprocessing median filter is applied to smooth the original image and suppress the image noise. Results are demonstrated on ovarian cells.