In this paper, an approach to extract edges from noisy images within a wavelet framework is described. The edge detection is one of the most important processing in image analysis or image understanding and therefore various methods have been proposed. However, most of them are sensitive to noise existence due to their local differential operations, and consequently they often tend to extract spurious edges from noisy images. We aim to solve this problem by incorporating wavelet transform into edge detection process. The main idea of our method is to modify wavelet coefficients adaptively, that is, coefficients associated with low frequency and noise components are both removed whereas coefficients associated with edge components are kept. Accordingly, we can remove noise and enhance edges simultaneously and thus extract edge information stably and efficiently. Experimental results show that our method performs well to noisy images in comparison with conventional edge detection methods.
展开▼
