Image processing tools
Image Enhancement Tools
The radar data processing software package includes the following tools intended for image enhancement: speckle-noise filters and edge detection (detection of local brightness variables) ones.
Suppression of speckle-noise on the radar images could be performed via different local linear and non-linear
filters such as mean, median, Lee, Frost, Kuan, Sigma-filter. There is another tool that provide possibility to
perform speckle-noise filtering using global approach. This filter is based on the wavelet decomposition with
different basis (D4 Daubechies, CDF 9/7, CDF 11/7, Villasenor, Odegard). User's interface of local filters allows
the selection of sliding window size.
For wavelet filters, user's interface permits the exploiting of different number of decomposition levels and
adjustment of weighting coefficients for particular level.
Edge detection tool includes number of processing algorithms (Sobel, Laplas, Prewitt, Kirsh, weighted,
3/5 gradation, vertical/horizontal shift, user defined mask and others). Both speckle filtering and edge
detection interfaces have two windows: first one shows the input image quick-look and second shows the preview
of processed subset. User can select the interesting area on input quick-look image interactively.
Feature Extraction Tools
Software tools intended for extraction specific information from the radar images. It allows user to detect
difference texture characteristics of scene and to form number of surface classes over one or several input images
under selected criteria. Set of input images could be compounded from multifrequency, multipolarization, or
multitemporal acquisitions. All input images must be previously coregistrated and to have same sizes.
Software tool for classification allows to perform an estimation procedure under criteria of minimum distance.
User's interface has possibility for selection of desirable metric (Euclidean, Mahalanobis, Bayesian metrics are
realized currently), number of classes for which processing is executed, and number of processing iterations.
Textures analysis tool allows to extract the one-dimensional histogram statistics such as mean, entropy,
standard deviation, energy, skewness, kurtosis and the two-dimension histogram statistics such as absolute value,
entropy, contrast, energy, maximum probability. User can to change the size of sliding window and processing
parameters (direction, distance). The quick-look of input file and the result of processing over interactively selected area can be viewed on interface.