中心B样条二进小波多尺度边缘提取  

CENTER B SPLINE DYADIC WAVELETMULTISCALE EDGE DETECTION

作　　者：贾天旭[1] 郑南宁[1] 张元亮[1] 

机构地区：[1]西安交通大学人工智能与机器人研究所

出　　处：《自动化学报》1998年第2期192-199,共8页Acta Automatica Sinica

摘　　要：分析了截尾的Ｃａｎｎｙ算子在多尺度边缘提取时对运算速度造成的影响，提出了中心Ｂ样条二进小波多尺度边缘提取．详尽地研究了Ｃａｎｎｙ算子与中心Ｂ样条函数的若干性质，中心Ｂ样条函数具有紧支集，以极快的速度逼近高斯函数，四阶中心Ｂ样条函数的导数比Ｃａｎｎｙ算子更接近最佳边缘检测滤波器，四阶中心Ｂ样条函数是二阶平滑问题的唯一最优解，并且它的时频测不准关系值非常接近时频测不准关系下界．从对计算结果的讨论中也得出中心Ｂ样条二进小波优于Ｃａｎｎｙ算子的结论．,We first analyse the effect of cutting off the Canny operator on the calculation speed for multiscale edge detection, and then propose the center B spline spline dyadic wavelet multiscale edge detection algorithm. Some properties of the Canny operator and center B function are studied in detail: 1) the center B spline function has compact support and the Canny operator does not; 2) the center B spline function converges to the Gaussian function and the derivative of center B spline function converges to Canny operator; 3) the derivative of the fourth order center B spline function is more like the optimal edge detection filter than the Canny operator is; and 4) the fourth order center B spline function is the only solution to the second order smoothing problem and its time frequency uncertainty value is very close to the optimal value. The result of the multiscale edge detection shows that the center B spline dyadic wavelet is superior to the Canny operator.

关 键 词：边缘提取 B样条函数 二进小波 图像处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]