Image enhancement and segmentation system for chromosome images
Loading...
Date
2017-06
Authors
Choo, Hooi Yang
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In karyotype, the chromosome images acquired have poor contrast which will contribute to the possible bad performance of the process or analysis in the future stages. The chromosomes are also usually overlapping or touching in the image, which require a specific methods or algorithm to segment and distinguish them. This study proposes a pre-processing and segmenting system that integrates several techniques from these two stages to enhance the image quality and segments the chromosome images including the overlapping and touching chromosomes. In pre-processing stage, an algorithm is developed to remove unwanted blobs. Three enhancement techniques namely Contrast Stretching, Adaptive Histogram Equalization (AHE) and Edge Sharpening are implemented to enhance the image quality. In segmentation stage, Global Otsu Threshold method is employed to segment the chromosome. To separate the overlapping and touching chromosomes, a geometrical approach is used by finding the corner points of the intersecting chromosome and segment them based on the points. The proposed algorithm was tested on 36 sample chromosome images, which consists of 1656 chromosomes. Among these chromosomes, there are 42 pairs of overlapping chromosomes and 45 pairs of touching chromosomes. 31 out of of 42 overlapping chromosomes can be segmented correctly (i.e. 73.81%), while 28 out of 45 touching chromosomes can be segmented correctly (i.e. 62.22%). Overall, 1566 out of 1656 chromosomes were successfully segmented, resulting in the success rate of the system to be 94.56%.