Abstract: 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 {unction 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.