The effect of speckle filtering on scale-dependent texture estimation of a forested scene

Abstract

Abstract-Spatial fluctuations in microwave backscatter may be an important piece of information in discriminating tree stands. However, the presence of speckle in synthetic aperture radar (SAR) image data is a barrier to the exploitation of image texture. We explored a new methodology that combines a recent adaptive speckle reduction algorithm by Lopes et al.[12] with a generic texture estimation scheme. We investigated the claim that this filter was capable of preserving backscatter texture. To understand if speckle reduction was destroying backscatter texture, we compared the strength of the relationship between forest inventory parameters and image texture as a function of spatial scale for both filtered and unfiltered images. We used Radarsat Fine mode image data: single look resolution is approximately 8.5 m, and pixel spacing is 3 m. Our study area was northern Vancouver Island, B.C., on the west coast of Canada. For the unfiltered data, we found that the ability of image texture to predict the forest parameters decreased as the texture scale increased from 3 to 13 m, suggesting greater information content in the small scale texture. For the filtered data, this relationship was much weaker at small scales and was not a function of distance. Our results suggest that the speckle filter was not retaining small scale texture, which is consistent with the theoretical hypotheses underlying its multiplicative noise model. We also show that there is significant information in small scale SAR image texture that may be used as an adjunct to other spatial information for discriminating tree stands in the temperate rain forest.

A SAR-Based Index for Landscape Changes in African Savannas

“…Pre-processing included radiometric calibration to radar brightness (Beta Naught, β • [87]), multi-looking (n = 2), terrain flattening to normalize radiometric effects caused by different incidence angles (Flattened Gamma Naught, γ • [88]) and Range-Doppler terrain correction to adjust topographic distortions using a digital elevation model (1 Arc-Second SRTM) [89]. All rasters were resampled to a common ground resolution of 30 m. No speckle filtering was applied in order to conserve image texture as accuracies of classifications based on SAR texture alone are reported to decrease when speckle was removed [90,91].…”

Combining SAR and Optical Data for Environmental Assessments Around Refugee Camps

“…In order to preserve the original image textures, no speckle removal was applied. In texture-based classifications the accuracy has been observed to decrease when speckle filtering was applied before classification (PRASAD & GUPTA 1998, COLLINS et al 2000. In contrast to L band radar such as ALOS, C band sensors are more sensitive to leaves and small branches, and have a lower penetration depth.…”

“…However, we additionally used different window sizes for each texture. As studies have shown that especially small structures exhibit valuable textures for classification (COLLINS et al 2000), we chose 3, 5 and 9 pixels as window sizes. This resulted in a total of 12 texture parameters per polarization which could additionally be used for the classification.…”

Coincident high resolution optical-SAR image analysis for surface albedo estimation of first-year sea ice during summer melt

“…A 5-by-5 pixel size enhanced Lee adaptive filter was applied to each image channel in order to remove speckle, an effectively random interference pattern which produces graininess in radar imagery (Woodhouse, 2006). This filter was chosen for this application based on two criteria: (1) the filter makes use of the coefficient of variation (CV) within a scene so as to preserve its radiometric information (i.e., σ°) while also preserving scene heterogeneity, or textural information (Lopes et al, 1990); and (2) it is important to preserve the spatial heterogeneity of a scene when the scale of the scene elements approaches that of the resolution of the sensor (Collins et al, 2000;Lopes et al, 1990). Embedded ground control points (GCPs) were then used to georeference images to a UTM Z15 (North) map grid.…”