![]() SAR Images FormatsUsually, one-look images are given in the "complex" format, in order to preserve the knowledge of the phase of each pixel in the image. The following figure shows the representation of a pixel in the complex format, where Uq and Ui are the real and imaginary components, respectively; A
is the module of the complex number, representing the amplitude of the pixel and Representation of a pixel in the complex format. The multi-look images are usually represented by Amplitude images, that is: The amplitude images are also known as linearly detected images. It is common, for this type of image, to use a representation of 16 or 8 bits per pixel. The images can also be represented in Intensity, that is: The Intensity images are of quadratic detection and need, in general, 32 bits for representing each pixel. Radiometric CorrectionsThe radiometric quality of a SAR data is affected by factors inherent to the instrument, as well as to the illumination geometry. The two main causes of the radiometric distortions that affect the interpretation of SAR images are: speckle noise and the antenna pattern effect.
Speckle NoiseThe Speckle noise is one of the main factors that degrade the quality of the SAR images. The Speckle is a multiplicative noise, proportional to the intensity of the received signal. The visual effect of this noise is a granular texture that might difficult the SAR image interpretation, reducing the separability among land use classes, lithological types, etc. There exist two methods for reducing the Speckle noise: filtering and multi-look processing. The filters should maintain the mean value of the radar return (backscatter), preserve the edges between different areas and the textural information. The Speckle noise is always associated to coherent imaging systems, such as microwaves, laser and ultrasonography. It is assumed that inside a resolution cell of an imaged scene, there exist a large number of scatterer elements, randomly distributed, such that these elements can produce constructive and destructive interferences on one another, causing abrupt variations on the intensity of the image, characterizing the Speckle noise (see figure below).
Figure - Resolution cell and the resulting backscatter Antenna Pattern EffectLow frequencies variations in the image, in the range direction, are mainly caused by the loss of power related to the side looking geometry, which decays with 1/R3 in the case of SAR images, where R is the distance of the antenna to a given point in the image. This problem becomes worse in images acquired by airborne sensors, since the ratio of Rmin (beginning of the swath) and Rmax (end of the swath) is smaller when compared to the images acquired by satellites, where the ratio is practically 1 (see figure below). Figure - Variation in range for aircrafts and satellite This loss in power is corrected in the instant of each pulse acquisition, by the STC (Sensitivity Time Control). However, due to imperfections in the STC system, or other radar electronic perturbations (variations on the gain of the amplifier during the echo acquisition time), mechanic or electric, the correction is not perfect, remaining residual variations.
Antenna Pattern CorrectionThe algorithm consists in generating a pattern by averaging the columns of the image. The columns averages should be taken in regions (windows) as homogeneous as possible. It should be guaranteed that the average is obtained for the entire range direction. The Pattern obtained by column average should be filtered (adjusted) in order to obtain only low frequencies variations. Two methods can be used for filtering (adjustment). The first is the moving average method, which consists in filtering the Pattern by the mean filter, where the number of points to compute the mean is defined by the size of the window in the interface. The second method is a polynomial fitting, where the degree of the polynomial is selected in the interface. After filtering the Pattern, the image can be corrected. One of the correction methods is the multiplicative, which consists in multiplying the pixel being processed, V(i,j), by a factor given by the ratio of the mean value of the Pattern, P, and the adjusted Pattern value, P(j), that is: where j is the image column index (range direction). The following figures presents the original image (a) (image from the SAR-580 system of Tapajós River, Brazil, gathered during the SAREX-1992 mission), and the corrected image (b). Figure (c) shows the plots of the original and the adjusted (smoother) Patterns used in the correction. (a) (b) Figure - (a) Original and (b) corrected images
Figure - Original Pattern with high frequencies fluctuations, and adjusted (by a polynomial of degree 8) to remove the fluctuations. See how to correct the Antenna Pattern
![]() |