A Novel Edge Detecting And Tracing Strategy Within 2D And 3D Images

Edge detection is one of the most important issues of image processing and analyzing. It is a basic characteristic of image processing which is widely used in image recognition, image segmentation, image enhancement, image restoration and image compress. Edge detection and extraction is the research hotspot in the technical field of image processing and analyzing. Since the sep-like edge is the most common edge in 2D images and the step-like edge surfaces usually correspond to the edge surface including the anatomical structure within 3D images. So the detection of step-like edge curve (surface) is a main research direction of edge detection technology. Step-like edge curve (surface) can be regard as discrete samples of continuous functions separated by 2D (3D) regular data. As a result, this dissertation discusses a novel edge detecting and tracking strategy based on the connectivity characteristics of this sort of implicit curve and surface within 2D and 3D images.Firstly, a novel edge curve detecting and tracing algorithm is proposed in this dissertation based the connectivity characteristics of the edge curve in 2D images. In the algorithm, we first detect the edge grids that contain edge curve sheet. During the tracing process, we choose a certain number of edge grids as seed grids, then, base on the connectivity principles of two adjacent grids, we track and obtain all the other edge grids that contain edge curve from the seed grids. At the last step of the algorithm we find all the edge pixels from the edge grids got yet. We do lots of emulational experiments. It turns out that the algorithm proposed in this dissertation can overcome two complex issues, selecting proper gradient thresholds and detecting weak boundaries, very effectively. Also, this algorithm does good performance in the judgement of edge detector including continuity, smoothness, thinness, location and noise proof.In addition, this dissertation proposes an extension of the edge detecting and tracing strategy base on the connectivity of edge curve to 3D images. Naturally, the algorithm in 3D images based on the connectivity of edge surface. Similar with the algorithm in 2D images, we first detect the edge cubes that contain edge surface sheet. Then based on the connectivity principles of the two adjacent cubes, and obtain all the other edge cubes included edge surface. The algorithm in 3D images can solve the issue that it is hard to determine the tracing direction when using gradient direction for trace effectively. The result of the experiments also demonstrates that the algorithm overcomes the flaws obviously, and can acquire high-precision (sub-voxel accuracy) edge surface.The edge curve, surface tracing algorithm in 2D, 3D images proposed in this dissertation both acquired high-precision edge. The experimental results show the effectiveness and advantages of these two algorithms. May all the work in this paper is of some value to research and applications in the image edge detection.