| The essential parts of ocular fundus images processing system, including image restoratiorn radiant calibration~ geometric calibrations mergence and 3 Dimensions display, are involved in this thesis. Image degradation exists in some sorts of imaging systems. Adaptive correction of regularization parameter given by Chen Wufan is adopted for problem of image restoration in space domain. According to the characters of symmetrical and positive definite matrix, inverse matrix is calculated approximately. Then the algorithm of fast approximate inverse matrix is elicited. This algorithm could effectively improve quality of restoration, and eliminate better over-sharpening phenomenon in restored image than others. Radiant distortion occurs in imaging by lens, especially in ocular fundus camera. A radiant calibration algorithm is presented to satisfy visual and psychology estimation. Details submersed in margin region could be recovered. In imaging system of ocular fundus, geometric distortion might be unavoidable, which is brought by taking photo at off angle. Based on anatomical and physiological characters of oculus, a mathematical model, suitable to medical visualization, is constructed. According to the cause of distortion, a new calibration method is presented. Mergence of multi ocular fundus images is often needed for clinic and research. Wavelet multiscale analysis is applied to image mergence algorithm. By matching of approximation coefficients matrixes decomposited by wavelet transformation at high level, the huge computation cost is cut down. Subsequently, a novel automatic mergence algorithm is presented with geometric calibration. Based on the mathematical model of ocular ellipsoid, we put forward a imaging method of 2-D image on 3-D retina surface. Certainly, the proposed method will help a lot for getting physical parameters as well as for clinical use. |
