Use of anisotropic diffusion filters in Synthetic Aperture Radar (SAR) imagery

PDF

Authors
  1. Barrie, G.
Corporate Authors
Defence Research and Development Canada, Ottawa Research Centre, Ottawa ON (CAN)
Abstract
The Defence R&D Canada (DRDC) Ottawa Research Center, ISR Data Exploitation Group, specializes in development of algorithms and tools enabling uptake of Synthetic Aperture Radar (SAR) data for Canadian Armed Forces (CAF) planning and operational purposes. This scientific report addresses the extraction of coastlines in SAR images exhibiting poor contrast between land and water backscatter. In place of traditional speckle filtering, we apply a nonlinear filter whose design principles are rooted in the physics of diffusion, where image blurring is treated in a manner analogous to diffusing particles. After a brief discussion of how the diffusion equation is applied to image processing, we introduce the Speckle Reducing Anisotropic Diffusion (SRAD) filter. SRAD is shown to be related to more commonly used speckle filters, which constitute a special case of isotropic diffusion. The SRAD filter is first used to process an artificial coastline consisting of two adjacent blocks of RADARSAT-2 (RS-2) data of water and land backscatter forming an idealized linear coastline. This construct allows us to establish easily quantifiable ground truth, and perform statistical analysis on both individual and combined land and water pixel distributions. Following this, SRAD is applied to large (~2 GB) Wide Ultra-Fine RS-2 data sets to examine its use in poor (0.6 dB) contrast imagery. Results indicate this is an effective filter, and when used in conjunction with a traditional edge detector such

Il y a un résumé en français ici.

Keywords
Synthetic Aperture Radar (SAR);radar data exploitation;anisotropic diffusion;speckle filtering;shoreline extraction;edge detection
Report Number
DRDC-RDDC-2018-R171 — Scientific Report
Date of publication
01 Jun 2018
Number of Pages
37
DSTKIM No
CA046773
CANDIS No
807368
Format(s):
Electronic Document(PDF)

Permanent link

Document 1 of 1

Date modified: