A Novel and Maximum-likelihood Segmentation Algorithm for Touching and Overlapping Human Chromosome Images

The chromosome abnormality in human is a vital issue. These abnormalities happen due to touching or overlapping chromosomes in human beings. It may cause breast cancer, improper structure or functions in body metabolism, birth defects, down syndrome, turner syndrome etc. To overcome these problems a stringent screening and diagnosing must be followed during earlier stage of pregnancy. In this paper Novel and Maximum-Likelihood segmentation algorithms were used to segment the overlapping and touching human chromosome images. Either the gray scale image was converted into color image or a color image is directly applied as the input to the Novel as well as Maximum-Likelihood segmentation algorithms. After obtaining threshold from binary, the watershed transform was applied. The output of watershed was improper. So after threshold the iterations were applied, followed by edge detection and corresponding segmentation. The Novel algorithm worked better only for touching chromosomes images but failed to work for overlapping images. Then the segmentation was followed using Maximum-Likelihood segmentation algorithm. By comparing with Novel algorithm, the Maximum-likelihood segmentation algorithm works better for both touching and overlapping chromosome images.