Abstract

Detecting edges is a fundamental problem in computervision with many applications, some involving very noisyimages. While most edge detection methods are fast, theyperform well only on relatively clean images. Unfortunately, sophisticated methods that are robust to high levels of noise are quite slow. In this paper we develop a novel multiscale method to detect curved edges in noisy images. Even though our algorithm searches for edges over an exponentially large set of candidate curves, its runtime is nearlylinear in the total number of image pixels. As we demon-strate experimentally, our algorithm is orders of magnitudefaster than previous methods designed to deal with highnoise levels. At the same time it obtains comparable andoften superior results to existing methods on a variety ofchallenging noisy images.