Segmentation And Feature Extraction Of Digital Image Of Leukemia Call

At present, medical workers mainly make leukemia diagnosis through morphologicalobserving to the bone marrow smear under microscope, and draw conclusion byexperience. But, there are many drawbacks by this way. In this thesis, Leukemia cell isanalyzed by image processing and analysis, including image segmentation and featureextraction, and image capturing.Image segmentation is based on HSI color model and saturation image isolated. Then, thehistogram of saturation image is analyzed through multiresolution is terms of wavelettransform. Optimum threshold is acquired by introducing (?) toms algorithm instead ofMallat algorithm and applying Gaussian functional second derivative to decompose thehistogram signal. The thesis further discusses the noise reduction by mathematicalmorphology and region noise filter. At the same time, over-segmentation can be restrainedeffectively under two different constraints. At last, some features of cord blood karyocytesin the bone marrow and abnormal morphologic cells are analyzed. There features containnot only traditional morphology but also texture and Invariant moment, which areeffective on classification.