Wavelet Operators for Multi-scale Edge and Corner Detection

• Main Authors: Ku, Cheng-Tung, 辜正同
• Other Authors: Hung, King-Chu
• Format: Others
• Language: zh-TW
• Published: 1998
• Online Access: http://ndltd.ncl.edu.tw/handle/96740208439922423880

碩士 === 義守大學 === 電機工程研究所 === 86 === 　　Based on the two-dimensional (2-D) wavelet transform operator, two new edge and corner detectors are presented in The thesis. The two detectors are called the wavelet edge operator and the wavelet corner operator, respectively. In terms of the two operators, a new algorithm of multi-scale edge and corner detection is proposed. The algorithm has characteristics of simple, fast, and high noise resistibility. 　　In edge detection, one applies different scales' operators to detect multi-scale dege points. The multi-scale edge point is defined as the pixel which appears the characteristic of local maximum gradient across scales. The detection hence can be performed by an simple logic AND operation. The algorithm has been applied to several noisy images. The experimental results reveals that the proposed algorithm can have much better detection performance in comparison with the Canny and Marr methods. 　　With the similar approach, in corner detection, different scale's wavelet corner operators are applied to detect multi-scale corners. The main merit of the new corner detection algorithm is that it is unnecessary to perform an edge detection process befreo the corner detection. The proposed new algorithm can be thereform simpler and faster than the conventional approaches. In addition, the new corner detection algorithm is also resistive to noisy environment.