Studies On The Classification And Recognition Of Red Blood Cell Microscope-Images Based On Morphology

Using digital image processing technology to classify and recognize red blood cell microscope-Images is an important research task in medical image domain. Currently, most researchesis point to leukocyte micrograph, but it is rare to classify and recognize on red blood cellmicrograph. In this thesis, enclosing red blood cells' modal change with disease diagnosing andfocusing on red blood cells, to deeply investigate noise reduction, segmation and featureextraction. The main contents are listed below:The first, original image is changed from color space to gray, execute noise reduction andthe image is processed by gray level transformations. Then images are segmated based onmorphological watersheds, so that the case of closely clusterd cells and not are segmated.The second, based on red blood cells' characters, we use Hu moments (seven) andgeometry characters (five) to describe the red blood cell microscope-Image.The third, according to 12 characters of red blood cell to train improved BP neural net.Finally, this method could classify and recognise the red blood cells well.