Edge detection using directional wavelet transform

In this paper we propose a new wavelet-based approach for solving the edge detection problem. A primary knowledge about the magnitude and the direction of the edges extracted by the ordinary wavelet-based edge detector is used to classify the image pixels as edge points, candidate edge points, or not edge points by double thresholding using extended Otsu's thresholding method explained here. Then based on the computed direction of the candidate edge points, the directional wavelet-based edge detector is applied on these pixels and the new magnitude of the wavelet transform coefficients is computed and then the pixel classified as edge points or not edge points. Using this method we can enhance the quality and performance of the wavelet-based edge detection in the presence of noise.