Research Of A Chromosome Recognition Algorithm

In the clinical research of cell genetics, chromosome analysis is one of the most essential tasks of genetic syndrome diagnosis. Correct karyotype image, which could be obtained through chromosome image processing, together with precise chromosome analysis and recognition, provides valuable information for doctor in the diagnosing of human illness. This paper mainly studies image preprocessing, feature extraction and classification analysis, accordingly realizes the chromosome recognition.Before image preprocessing, human-segmentation need to be applied to the whole chromosome image in order to get single chromosomes, and then, straighten up and normalize the chromosomes by image processing software. During the preprocessing, clearing "salt and pepper" noise by median filtering, binarizing image, smoothing boundary by morphological cleaning and gray stretching altogether lead to a perfect image.The length of chromosome, centromeric index and G-band—serving as three pattern features—are used for chromosome recognition. The length and centromeric index can be measured by image processing software, and G-band stands for the average gray resulting from the media axis.In the study of chromosome recognition algorithm, we use average gray distribution profile to denote G-band, bring forward a concept of chromosome fingerprint, select cosine of angle method to calculate the similarity, and match the homologous choromosomes. Meanwhile, better results root in the resolve of different lengths of chromosomes by profile straightening, as well as the solving of peak shift by profile moving. Illustrating G-band by gray gradient profile, rather than average gray distribution profile, is much more effective.