Cell distributions between cancer tissue and normal tissue images are different. In this study, in order to distinguish the cancer image, image processing is used to extract cell nucleus from the images by techniques of an edge extraction and an elliptic ring template matching. The matching method is improved by one-pixel step moving and by calculating matching rates for both on-ring and in-ring. As the result, typical 4 cells in a normal tissue image have been successfully extracted with matching threshold margin of 3%, when the cell nucleus edges are extracted by thresholding an edge gradient of 4.0.
展开▼