On the basis of the Marr edge detection method and the selection of the first and second order derivative of the Gaussian function as the wavelet

we proposed a method of multiscale edge detection that uses two variable parametes. The grade of the wavelet decomposition can be regarded as the control parameter for the viewing distance

and the standard deviation R of the Gaussian function co rresponds to the degree of the details of the detected edges. This method is suitable for the searching strategy about the“Pre attention”and the“Post attention”in active computer vision.