Study On Wavelet Multi-scale Product And B-Spline Edge Detection Algorithm

Edge detection technology plays an important role in the Image Engineering, it is core content of Image Segmentation which is key linkage between image analysis and image processing and also the foundation of amage understanding. So edge detection is a hot point for researchers in the field of image process, which results in an amount of algorithms for edge detection. However, in practice of actual application and development, it is discovered that current available edge detection algorithms, which are applied to target detection and recognition, can't obtain satisfying results. Thereby, the researches on edge detection and target measurement combining demand of actual project which is ICF((Inertia Constrain Fusion) image processing system has been explored in this paper. In ICF system high accuracy position space coordinate of the target is one of the core technique, and the key of resolving this problem depends on the high accuracy detection, recognition and measurement of the target by making use of image segmentation technique. In this thesis several problems are discussed. Firstly, by the analysis of wavelet multi-scale edge detection principle, and different regulation of wavelet transformation to the noise and edge, we bring up wavelet multi-scale product the edge detection algorithm which can availably repress the noise. Especially, the algorithm can effectively recognize the target district at the big scale without considering the accuracy. Secondly, making use of the advantage of the B-Spline function to the Gauss function, we developed Canny-based improvement edge detection algorithm, which can increase the edge clearness and continuity, and has the ability and advantage of anti-noise by comparing with the Canny algorithm. Thirdly, because the analysis technique of sub-pixel is non-substitute in the high accuracy image measurement, the research on the new technique of sub-pixel position is developed in this paper. Finally, we make use of the above method, and bring up the automatic detection, recognition and measurement algorithm.