Directional multiscale edge detection using the contourlet transform

Wavelet multiresolution analysis allows us to detect edges at different scales, also to obtain other important aspects of the extracted edges. However, due to the usual two-dimensional tensor product, wavelet transform is not optimal for representing images. The main problem in edge detection using wavelet transform is that it can only capture point-singularities, and the extracted edges are not continuous. In order to solve that problem, we propose a new image edge detection method based on the contourlet transform. The directional multiresolution representation Contourlet takes advantages of the intrinsic geometrical structure of images, and is appropriate for the analysis of the image edges. Using the modulus maxima detection, an image edge detection method based on contourlet transform is proposed. To suppress the image noise effect on edge detection, the scale multiplication in contourlet domain is also proposed. Through real images experiments, the proposed edge detection method's performance for the extracted edges is analyzed and compared with other two edge detection methods. The experiment result proves that the proposed edge detection method improves over wavelet-based techniques and Canny detector, and also works well for noisy images.