Fast Multiscale Edge Detection and Fiber Enhancement

Meirav Galun, Ronen Basri and Achi Brandt

ICCV 2007

image 1
image 2 image 3
image 4 image 5

Weak edges can be detected only over sufficient extent

image 6

We need: Differences of oriented means at all lengths and orientations…

Our Goal

Fast detection of faint edges in noisy images


We present a multiscale algorithm for edge detection suitable for both natural as well as noisy images. Our method is based on efficient multiscale utilization of elongated filters measuring the difference of oriented means of various lengths and orientations, along with a theoretical estimation of the effect of noise on the response of such filters. We use a multiscale adaptive threshold along with a recursive decision process to reveal the significant edges of all lengths and orientations and to localize them accurately. We further use this algorithm for fiber detection and enhancement by utilizing stochastic completion-like process from both sides of a fiber. Our algorithm complexity is O(N log N), N=#Pixels.


Algorithm outline

  1. Calculate differences of oriented means at multiple lengths & orientations
  2. Detect significant responses using a multiscale adaptive threshold
  3. Identify coherent edges by a recursive decision process

Calculating all responses:

  • How many? Haw fast? O(N log N),N=#pixels (Brandt & Dym 1995)
  • Recursive computation: long responses accumulated from short ones

Multiscale adaptive threshold:

  • Estimating pixel noise
  • Noise is averaged out by long responses,
  • Accordingly, determine a multiscale adaptive threshold:


title={Multiscale edge detection and fiber enhancement using differences of oriented means},
author={Galun, Meirav and Basri, Ronen and Brandt, Achi},
booktitle={2007 IEEE 11th International Conference on Computer Vision},