Thermal Noise Removal is an essential preprocessing step in Sentinel-1 SAR data processing. Thermal noise refers to unwanted background noise generated by the radar sensor electronics, which can introduce artificial low-intensity signals across the image. This noise is not related to ground surface characteristics and may affect the accuracy of backscatter analysis.
In this part, the thermal noise component is removed from both VV and VH polarization bands to improve radiometric quality. Removing thermal noise enhances the contrast between real surface backscatter and background noise, particularly in low-backscatter areas such as water bodies. This ensures more reliable results in subsequent steps such as calibration, filtering, and classification.
Figure 1: Left side (BeforeTNR) and Right Side (After TNR)
After Thermal Noise Removal, radiometric calibration was applied to convert the SAR image digital numbers into physically meaningful backscatter values. Raw SAR intensity values are affected by sensor characteristics and acquisition geometry, so they do not directly represent true surface reflectivity.
Calibration transforms the data into standardized backscatter coefficients such as Sigma⁰ (σ⁰), Beta⁰ (β⁰), or Gamma⁰ (γ⁰). In this step, the Beta⁰ band was generated, which represents radar backscatter per unit area in slant range geometry. This process ensures that pixel values accurately reflect the scattering properties of ground surfaces and allows comparison between different scenes or acquisition dates.
Radiometric calibration is essential for quantitative analysis, including classification, change detection, and surface characterization.
After radiometric calibration, Radiometric Terrain Flattening (RTF) was applied to correct terrain-induced radiometric distortions. In mountainous areas such as Salzburg, radar backscatter is strongly influenced by slope and terrain orientation relative to the sensor. Slopes facing the radar appear artificially bright, while slopes facing away appear darker, even if they have the same land cover.
Radiometric Terrain Flattening uses a Digital Elevation Model (SRTM 1Sec HGT in this case) to normalize backscatter values by compensating for topographic effects. This process removes brightness variations caused by terrain geometry and ensures that pixel values represent actual surface scattering properties rather than slope effects.
This step is especially important in mountainous regions, as it improves the consistency and comparability of backscatter values for further analysis and classification.
Multilooking is a speckle reduction technique applied to SAR imagery by averaging neighboring pixels in the range and azimuth directions. Since SAR is a coherent imaging system, it contains speckle noise, which appears as granular variations in brightness. Multilooking reduces this speckle effect by combining multiple looks (independent observations) of the same area.
The main purpose of multilooking is to improve radiometric quality and enhance image interpretability, especially for visual analysis and classification. However, this process slightly reduces spatial resolution because pixel values are averaged. In this step, multilooking was applied to the calibrated (Beta⁰) bands to obtain smoother backscatter images suitable for further analysis.
