Article Preview
TopIntroduction
Remote sensing is the science and art to collect information about an entity, region, or phenomenon by analysing data acquired by a computer system that is not in contact with the object, area, or phenomenon being investigated (Yesserie, 1999). In all kind of remote sensing, electromagnetic energy source is used. The energy waves traveling in waves. These waves are computed in wavelength and velocity assessed by its frequency. In general, Remote sensing categorized in two part. Passive remote sensing, in which passive sensor used to captures the emitted electromagnetic energy from natural sources. In the passive remote sensing, microwave radiometer and magnetic sensors are used in non-imaging sensors whereas, cameras, spectrometer, microwave radiometers are used for imaging sensors. In the active remote sensing, which is used to detect reflected responses from objects irradiated by artificially generated energy sources. In this approach, microwave radiometer and laser are used for non-imaging sensors and real aperture radar and synthetic aperture radar are used for imaging sensors.
In the area of remote sensing, SAR (Synthetic Aperture Radar) imagery systems are growing tremendously which attracted attention of various researchers (Jun Wang et al., 2018; Xu et al., 2019). In order to produce SAR images, an active sensor sends out the microwave signals to the observation location, and it receives back the signal or backscattered signal from that location (Zhai et al., 2019). During the flow of the scattering, simulation of SAR images can consist of two components. The first component is the signal level simulation which is focused on electromagnetic scattering and the second component is focused on hypothetical distribution (Huang et al., 2019).
Typically, SAR images are captured by moving objects like satellite and aircraft. The widespread use of this technology is in various remote sensing applications because of its all-time and all-weather collection capability (Moreira et al., 2013) and it produces images in very high resolutions. SAR imagery systems are most effective in interference effects and depend on the coherence properties of the scattered signals. Due to the consistent nature of these waves and the subsequent unified processing, SAR images are distorted by a particular strong noise, called “speckle” and most of the SAR image gets affected by speckle noise. Thus, it is very important to remove such type of noise from the images before processing it. The information in SAR images has great importance but due to speckle noise, the quality of this information degrades severely. Which can also degrade the performance of SAR image based applications.
The de-speckling approach aims at eliminating speckle noise and retaining all textural features in SAR images. The de-speckled image is used to identify texture material further, to distinguish artefacts and boundaries. Therefore, the speckle noise reduction algorithms from SAR images need to be analysed and developed which can preserve important features.
A strong adaptive speckle filter should have the following properties:
- 1.
Reduction of the speckles in statistically homogeneous regions.
- 2.
Feature preservation (such as edges & real textural variations).
- 3.
Radiometric preservation.
A variety of de-speckling approaches have been proposed to restore SAR images for decades. In (Argenti et al., 2013), authors have explained the applications of different filters for despeckling process. It includes classical filtering such as Kuan, Lee and Frost filtering. A non-local algorithm was proposed in (Jin Wang et al., 2006), where the authors have applied the repetition of tiny image patches surrounded in a natural images which is further extended through replacing euclidean distance by noise distribution based similarity measures. A non-local and wavelet shrinkage based methodology has been proposed in (Parrilli et al., 2012), where the SAR oriented block matching 3-D has been used.