MULTI-SCALE EDGE DETECTION WITH GAUSSIAN FILTERS.

The authors give a possible solution to the multiscale edge detection problem. For ease of presentation only one-dimensional results are given; however, the proposed multiscale detector could be extended to include general two-dimensional images. Edge and inflection points are defined for 1-D images, and a simple statistical analysis of Gaussian filtering is given to show how noise removal is improved as the variance of a Gaussian filter increases. It is shown that the Gaussian filtering first suggested by D. Marr (1982) provides a good choice, but not a unique choice, for multiscale edge detectors.