|Abstract||The information content of SAR imagery is often seriously degraded by the noise or speckle inherent in the imagery. The presence of this speckle or noise will effectively limit the amount of usable
information. Filtering is an effective preprocessing technique to reduce the effects of speckle.|
A great number of filtering algorithms have been developed, although few have been compared for a combination of speckle reduction and image
integrity. The preservation of certain image qualities, i.e. edges and strong returns, is imperative in the processing of SAT images (Thomson, et al, 1987). Thus the objective of this study was to assess the median, mean and adaptive filters, and in
particular to evaluate the size of the filter in relation to the pixel spacing and image resolution. Within the agricultural context, filtering has become an important preprocessing step to allow the reduction of within field variation, while
retaining the edges of field boundaries.
The data set used in this study was acquired over an agricultural test site near Melfort Saskatchewan. The C-VV imagery was acquired on July 31, 1983 using the Canada Centre for Remote Sensing Convair 580
aircraft. Using a subarea of the original data, mean, median and adaptive filters were passed over the imagery. Statistical and graphical information were calculated using training areas representative of 5 canola, 5 wheat and 5 fallow fields.
The data was entered into a statistical data base to allow for a graphical representation of the data.
The degree of correlation between pixels introduced by using the mean and median filters was investigated. The results indicate that the 5x5
median or mean filters provided a good combination of speckle reduction and preservation of image integrity.