Preprocessing And Feature Extraction Of Tumor Cells' Image

This thesis is intended to develop a computer-aided diagnosis system of tumor cells based on the flow of image processing and recognition. Computer-aided diagnosis (CAD) of cell smears is realized by utilizing a color CCD video camera to capture cell images, making use of digital image processing and pattern recognizing techniques to classify cells. With a view to preprocessing of cell images, feature extraction of tumor cells'image, a great deal of study is devoted and satisfying results are achieved.In the aspect of cell image preprocessing, it improved that the improved morphology opening and closing filter method is better than others in the aspect of tumor cell's image denoise, by comparing some filtering algorithms. It can remove the noise and the natural discrete cells in the image,and reserve all characters of remanent cells , in order to increase the processing speed of laster step.In the aspect of feature extraction, tumor cells have many characteristics, the most important characters are nucleocytoplasmic ratio, the size of nucleus, of cell's conglobation. The last two characters are few used in papers. The hyperchromasia can be easily extracted from the image of the nuclei by the well-known Morphological Top Hat Transform. And the character of adjacent nuclei center distance can be used new method that is mentioned in this thesis.The conception of cell image segmenting is defined. Segment the cell image to small pieces by polymerization of the cells. And feature extraction is aimed at these small pieces. It can gained the most exact value of characters, and it is the foundation of the cells image processing and recognition.