A new method of wavelet transform-based edge detection

In many edge detection methods, Finding a proper threshold is an unavoidable step. In this paper, a new algorithm of edge detection is proposed based on wavelet transform. After multiplying the DWT coefficients in the adjacent scale, a new method is proposed to calculate the proper threshold, which is used to separate the coefficients come from wavelet transform. After the multiplying the coefficients in adjacent scale, the product coming from noise are small and accounts for the most part of the data, while there are less product, whose amplitudes are bigger, coming from edge. Thus, we statistics the product and get the interval in which the amount of the product is the biggest. The threshold is the upper bound of the interval. A scheme is then designed to synthesis the two edge maps obtained in two orthometric directions. A set of the experiments demonstrate the effective of the approach.