Directional Schemes for Edge Detection Based on B-spline Wavelets

The aim of the present paper is to introduce two efficient robust schemes for edge detection and boundary detection. The main idea is based on the odd-order B-spline wavelets. In the first proposed scheme, high-pass filter of an odd-order B-spline wavelet has been rotated in four directions, and then the best directions for each pixel have been selected through computations. The novelty aspect of this scheme is that unlike to other edge detectors based on wavelets which use wavelet transform modulus value for detecting the edges of the images, each direction information is involved in detecting the singularities of the image independently and then those directions where the singularity in those directions has high absolute value are chosen for detecting the edges. The second scheme, which is a modified active contour model, has been designed for image boundary detection. This model not only is applicable in different scales, but also against the previous active contour models, uses more directional information to guide the motion of the initial contour and is more accurate than previous active contour models for boundary detection or in some cases for segmentation. Moreover, this scheme is not sensitive to the location of initial contour. Experimental results show the accuracy of the proposed schemes in comparison with other state-of-the-art edge detectors like curvelets, shearlets, wavelets and Canny method.