Five Groups Differential Counting System Of Leukocyte

Blood cell composition reveals important diagnostic information about the patients. The counting result gives important information about patient's health status and plays an important role in diagnosis. By present, electric resistances and lights spread methods are widely adopted in automatic leukocyte counting systems. Under the two above situations where the human vision can't arrive, we can only perform regular classification. However, the image processing method, which is similar to the way in which human judgment system works when recognizing cells, can not only complete the counting but also distinguish the abnormal cells alone from the normal ones. This method can learn from the pathologist's diagnostic experience, also take the advantage of the large capacitance and high speed of computers. It greatly improves the efficiency, detecting accuracy and individual case analysis ability. Therefore, to develop an automatic white blood cell counting system based on image processing skills is of great significance. At the same time, the digitalization is a trend of development for the biology science. Hence, researches in this project do help other aspects in biology and other fields.The auto-analysis system of blood cells is the combination of computer science, image processing, pattern recognition and artificial intelligence. It is a " brain -eye" system which mimics human being's behavior. In this thesis, a retrospection of the domestic and international researching in automatic leukocyte analysis system is made, and an applied system structure using a PC, a microscope and a color CCD camera is proposed. According to this structure, we made a research of how to segment color blood cell image. Firstly, The original image is transformed from RGB (Red, Green, Blue) color space to HSI (Hue, Saturation, Intensity) space. Because the H component of the smeared image contains more information than I and S components, the iterative OTSU approach based on circular histogram is applied here in order to binarize this image. Then, mathematical morphological operations are applied to fill the holes and gaps, erase the dots at the same time. In the subsequent leukocyte labeling steps, a fast method for labeling connected component based on edge tracking is put forward, which reduces largely the scanning time and improves the real-time ability. Again, another OTSU segmentation step within each leukocyte is performed to get its nucleus. Finally we extract the shape, color and texture characteristics of each leukocyte and its nucleus.Because there are not sufficient samples of leukocytes, a new kind of machine learning...