| Summary: | 碩士 === 國立成功大學 === 工程科學系 === 87 === Due to the nondestructive, noninvasive and real-time display capabilities, the ultrasonic imaging system becomes an important inspection tool for clinic medicine. However, the ultrasonogram is noisy. It is difficult to find out the contours of 2-D ultrasonic images.In this thesis, an algorithm has been developed to find out the contours from noisy ultrasonic imasges. The spirit of the algorithm is similar to the Snake algorithms. But insteading of the energy function, Variation of local graylevel, and the changing of curvature are considered. Firstly, a closed curve which encloses the object on the image is manually drawn by mouse. Then sample points, with equal distance, on the closed curve are set. For each sample point, it is tested if the point is an edge point of the object according to the verifying coorditions which conducted from the graylevel of pixels locally. For these sample points, that are not belonged to the edge points, candidate edge point is selected to instead of the current sample point, The amount of variation of the graylevel and the changing of curvature are used to select the candidate edge point. To find out the contours of a series 2-D images, the algorithm is applied semi-automatically to these 2-D images, To genrate the 3-D reconstruction of a series of 2-D images, conventional rendering technologies are used
|
|---|
