| There are many kinds of medical micrographs such as cell smear, tissue slice and so on need to be observed and analyzed. By using different kinds of microscope, the microstructure of cell or tissue can be observed, and configuration can be analyzed, which provide enough information for diagnosis. But observing the microstructure through microscope not only makes the observer's eyes tire, but also adulterates some subjective factor and decreases the objective quantitative which causes bigger error. With the development of the technology, medical check-up objectivity is demanded higher and higher, the means of check-up develops from manpower to objectivity. So, a new technique is needed to meet the consumer's requirement.This research focuses on the image processing and counting the number of red blood cell in urine image captured under microscope. After preprocessing the urine image by denoising and image segmentation, red blood cell features are extracted and chosen. And then effective classification method (Nearest Neighbor, NN) is used to classify and count the red cell accurately. This research supplies a reliable and faster detecting method for the doctor.In this paper, the cell image is preprocessed using the sobel operator to detect the edge of red cell image. In order to be sure the single edge image is obtained, some edge image must be thinned. There are some scattered points which can not compose a round are regarded as noise and will be taken out.Second, the principle and characteristic of cell image segmentation methods is compared. Our research adopts the Hough transform to detect the centre of red cell in urine image because the shape of red blood cell in image is circle. So, transcendental knowledge (the radius of red blood cell) can be used to confirm the position and the center of red blood cell in image.At last, Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) are used to classify and count of red blood cell. |
PDF Full Text Request