Integrating Passive and Active Remote Sensing Data with

Spatial Filters for Urban Growth Analysis in Urmia, Iran

Vahid Isazade1 , Esmail Isazade2 , Abdul Baser Qasimi*,3 , and Ahmed Serwa4
1 Department of department of geographical science, Kharazmi University, Tehran, Iran
2 Department of Urban planning, Tarbiat Modares University, Tehran, Iran
3 Department of geography, Faculty of Education, Samangan university, Samangan, Afghanistan
4 Faculty of Engineering in El-Mataria, Helwan University, Cairo, Egypt

* Correspondence to: Abdul Baser Qasimi, [email protected].

Abstract: Active remote sensing and related technologies are one of the new tools recently used
to monitor complications and urban growth. This research aims to investigate the effect of spatial
filters on urban complications. The aim of this paper is to compare Lee, Frost and Average spatial
filters with Landsat 8 satellite images and radar images with HH/HV polarization to investigate and
identify urban features in the west of Urmia City. The results showed that Filterelli with the kernel
3 × 3 had reduced the spike noise in Alus Palsard satellite radar images in identifying the growth
of urban tolls. Also, the results of K-means classification, the Lee filter with kernel size 3 × 3 more
accurately identifies the urban features of the west of Urmia City. The kappa coefficient was 0.96%,
and the overall accuracy of this filter was 97.36%. Therefore, Lee’s spatial filter has successfully
identified the urban features of west Urmia with high accuracy. This system can be implemented in
any other field due to its generality and reliability. This system may be a step towards remote sensing
automation.

Keywords: Urban growth, Active remote sensing, Passive remote sensing, Lee filter, Urmia.

Citation: Isazade, V., E. Isazade, A. B. Qasimi, and A. Serwa (2023), Integrating Passive and Active
Remote Sensing Data with Spatial Filters for Urban Growth Analysis in Urmia, Iran, Russian Journal
of Earth Sciences, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861

1. Introduction
Fusing different remote sensing data types (active and passive) can improve the
performance of any spatial system [Serwa and Saleh, 2021]. The world is developing fast the
motion Toward urbanization is built mainly through updating the texture maps [Seto et al.,
2012; Sun et al., 2007]. Land use/cover maps are suitable for planning, environmental
administration, disaster management, and public education on global development status
[Qasimi et al., 2022]. The impacts of the growth of urbanization on the environment
consist of triggering possibly destructive feedback such as a massive change in climate,
water pollution, and destruction of nature by developing settlement areas [Ghanbari et al.,
Research Article 2022; Mohan and Kandya, 2015; Zia et al., 2015]. Urbanization accelerates significantly,
and the number of sizable metropolises continues to grow [Hu et al., 2016]. For multiple
Received: 15 May 2023 years remote sensing of the satellite has been used for continuous mapping and observing
Accepted: 7 July 2023 various earth countenance consisting of urban objects [Isazade et al., 2022; Yang et al., 2016].
Published: 30 December 2023 Methods and techniques are desired over manual view methods because of additional
effectiveness and accuracy, like time and spatial coverage [Isazade et al., 2022; NISAR, 2018;
Podest, 2018]. Some international spaces agencies, including India and the European place
Copyright: © 2023. The Authors.
agency the National Aeronautics and place command [Attarchi et al., 2020; Podest, 2018],
This article is an open access article
have successfully developed and designed several optical satellites to observe and measure
distributed under the terms and the features of the earth’s surface, such as Landsat and IRS series [Kraus, 2003; Li et al.,
conditions of the Creative Commons 2020; NISAR, 2018], Sentinel 2 along with artificial crevice radar (SAR) satellites so as
Attribution (CC BY) license (https:// RISAT-1, Sentinel 1-6, MODIS Aqua, Terra. In the last few years, optical remote sensing
creativecommons.org/licenses/by/ observations are getting outmoded, as expected by their ineffectiveness in piercing past the
4.0/). dense cloudy cover [Kraus, 2003; Li et al., 2020].

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 https://rjes.ru/
Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Nowadays, optical satellites are generally utilized for some schematization and moni-
toring operations. Though in some positions, it is essential for remote sensing to survey
and observe the status of the earth’s surface condition in the period of cloud cover and
terrible climate conditions [Isazade et al., 2021]. Multispectral satellite images are more
general than optical ones due to the variety of spectral resolutions and bands [Serwa and
Elbialy, 2021]. Using active sensing data perfectly performs to obtain spatial data [Serwa
and Saleh, 2021]. Monitoring surface situations using radar equipment has a different
view and provides several states for feature recognition [Alves et al., 2020; Leiva-Murillo
et al., 2013]. However, acquiring any radar images comes up with multi-dimensional issues
containing quality of the information, cost of acquisition, preprocessing of the data or
methods of correction, and enhanced distortion. This data collection becomes offender the
efficacy cited by some caused mistakes that need information processing [Betbeder et al.,
2015]. A prominent section of the processing functions underscores the drop of extended
contortion or noise from the accepted SAR images. In this issue, the European charge holds
attended to institute the Program of Copernicus, which designed a novel model mutation
toward the accessibility and availability of information to hand over different earth-seeing
services by the support of satellites and in site information down six present Copernicus
services [Kaushik and Kumar, 2020]. Lines and edges are commonly easy visages to be
isolated from a satellite image so that they can be operated for subsequent studies in the
domains of image processing, computer perspective, remote sensing, and photogrammetry
[Rasche, 2018; Seo, 2018, 2019a,b]. Zakeri et al. [2017] utilized S1 and ALOS- 2 PALSAR-2
images for land cover classification in Tehran. Utilizing the backscattering worth on S1
images for the support vector machine classification of land cover by five categories, Kappa
stood at 0.30 and the OA 45.70%. Additionally, elements that had surfaces were selected
through the PCA and plugged with the backscattering polarized satellite imagery.
Osgouei et al. [2019] used enhanced image classification to map the land cover for
Istanbul, acquired from sentinel 2A data, employing SVM. The normalized deference
tillage index was utilized to determine built-up elements as the domain of a three-band
synthesized image, comprising the vermilion-edging-based normalized dissimilarity vege-
tation index and the corrected, normalized distinction water index. Camargo et al. [2019]
and Lapini et al. [2020] employed SAR imagery for Brazil’s equatorial savannah land cover
classes and the Mediterranean region’s forest classes.
Previous studies also offered MLP, RBF and Support vector machine (SVM) as the
exact arranger. Also, the decision tree exhibited a refined version for observing specific
land cover categories (e.g., vegetation). Deep learning classifiers are proven perfect [Serwa,
2020]. Li et al. [2020], Mun et al. [2019], and Routray et al. [2020] focused instead on filtering
out the performance and worth keeping of an image. Speckle noise can bring down the
modality of the image. Due to this reasoning, the assessment has carried a position among
untroubled and filtered imagery. The spot degradation is accomplished at the damage by
protection data.
Established at this, an equivalent extent has been given to estimating the speckle
filtering and filtering efficiency and retinue in detail. Wei and Blaschke [2018] created a
collation between supervised classification, object-based, and pixel-based unsupervised
category, and the object-based process is recommended for the category of increased-
solution images. Abdollahi et al. [2017] illustrated the road extraction, acquired by SVM,
level set method and Otsu’s algorithm with Comprehensiveness and feasible Correctness.
However, it is limited to images with only a spatial resolution of 0.21 m. Dai et al. [2020]
defined path exploitation acquired utilizing Preprocessing, smoothing, segmentation, and
eventually and eventually flexible path factor acquiring per a specific histogram MLSOH is
intended to extract the road and its orientation by Correctness of 99.54%, with a Totality
97.73%. However, it is not reasonable for the extensive shadow area, and the image is
regularly essential to be enhanced. Shi et al. [2019] extracted constructions using artefact
features well-found in building measures, figures, tissue, paint, shade, and roof textile.
In common, the broadly utilized methodology for remote sensing feature detection is

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 2 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

classification-based. Building extraction is done by separating its elements and applying

classification techniques like Support vector machine or RF.
Isazade et al. [2021], spectral indices and constructed and non-constructed urban
features and their comparison with surface temperature using Landsat satellite images for
the city of Tehran were evaluated. Their results showed that VbSWIR_BI and IBI indices
have the highest accuracy compared to other indices. With the increase in construction,
the temperature has risen in the central and northern parts of the Tehran metropolitan.
Ghorbanian et al. [2020] demonstrated the acceptable granular noise decrease to dissolve
the difficulty relevant to the SAR data used for the broad application areas. A comparative
analysis of filters is highlighted to figure out the performance of the filters and the ability
to extract edges, lines, and points. Dass and Yadav [2020] Examined Radarset-2 images for
the speckle containment with the Lee, Lee-sigma, Frost and Gamma-map. In common, the
median & mean filters are also contained Alves et al. [2020], Liu et al. [2019], all filters are
evaluated and show that some filters perform well in reducing the effect of granular noise
and conserving the details of the edge.
The study aims to compare different filters’ performance on AlosPalsar satellite bands
for noise reduction and feature extraction in urban environments, focusing on the northwest
region of Urmia City, Iran. Specifically, the Lee, Gamma-Map, Frost, and Average filters
will be compared using K-means classification to identify the filter that provides the closest
results to surface phenomena. Additionally, the study seeks to investigate the impact of
the HV band’s polarization properties and sensitivity on urban feature evaluation and
resolution. Moreover, the research aims to determine the suitability of the SSI index for
estimating noise-free gray levels. Overall, the study aims to contribute to improving remote
sensing automation and monitoring of earth changes in sensitive areas such as natural
resources, the environment, and agriculture.
The innovation of this paper lies in its application and evaluation of different spatial
filters, specifically Lee, Frost, and Average filters, for monitoring urban complications using
active remote sensing technologies. The paper focuses on comparing the effectiveness of
these filters in identifying and characterizing urban features in the west of Urmia City
using Landsat 8 satellite images and radar images with HH/HV polarization.

2. Study Area
Urmia locates in the central part of the province of West Azerbaijan, covering an
area of about 5227 km2 (Figure 1). It is the most populous city in the province, with a
relative share of 31.28%. The study area shares borders with Turkey in the west, Lake
Urmia in the east, the cities of Naqadeh, Mahabad, and Oshnavieh in the south, and Salmas
in the north. According to the 2016 census, it has 736,224 inhabitants [Fazel et al., 2017].
According to environmental considerations, the Urmia Basin’s cities are vital in agricultural
and industrial activities. There are 43 urban centers and 520 villages with about 7.3 million
canteens, 9 to 10% of the total population. [Fazel et al., 2017; Nouri et al., 2017; Shokati and
Feizizadeh, 2018].

3. Data and Methodology

3.1. Data
AlosPalsar radar images ( HH and HV) sensors for the research area are used with a
rare angle of 34.3 degrees. The second data used in this study is Landsat 8 bands (Blue, Near
infrared, and Mid-infrared bands) for 2020 west Urmia City. Specifications of AlosPalsar
Landsat 8 images are presented in Tables 1 and 2.

3.2. Geometric Correction and Calibration

In order to apply the geometric corrections for the radar images of the west Urmia
City, the Snap 5.3 software was used. Then its pixel size was changed to 30 meters and
the color image (HH-HV-HH/HV) RGB After generating the comment, the polarimetric
images (HV) and (HH) are subject to geometric correction.The results showed that spectral

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 3 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Figure 1. Study area west Urmia City.

Table 1. Profile of AlosPalsar radar images (https://www.eorc.jaxa.jp/ALOS/en/alos/sensor/palsar_

e.htm)

Polarimetric
Mode Fine ScanSAR
(Experimental mode)
Center Frequency 1270 MHz (L-band)
Chirp Bandwidth 28 MHz 14 MHz 14 MHz, 28 MHz 14 MHz
Polarization HH or VV HH+HV or VV+VH HH or VV HH+HV+VH+VV
Incident angle 8 to 60° 8 to 60° 18 to 43° 8 to 30°
100 m
Range Resolution 7 to 44 m 14 to 88 m 24 to 89 m
(multi look)
Observation Swath 40 to 70 km 40 to 70 km 250 to 350 km 20 to 65 km
Bit Length 5 bits 5 bits 5 bits 3 or 5 bits
Data rate 240 Mbps 240 Mbps 120 Mbps, 240 Mbps 240 Mbps
< −23 dB (Swath Width 70 km)
NE sigma zero < −25 dB < −29 dB
< −25 dB (Swath Width 60 km)
> 16 dB (Swath Width 70 km)
S/A > 21 dB > 19 dB
> 21 dB (Swath Width 60 km)
Radiometric accuracy scene: 1 dB / orbit: 1.5 dB

noise was observed in both produced images. The replay values of the HH/HV sensor
polarizations are presented in Table 3. Methodology for identifying urban landscapes using
radar and land images (Figure 2).

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 4 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Table 2. Landsat satellite bands 8 OLI sensors (http://mehdigis.blogfa.com/post/266)

Spectral Bands Wavelength in µm Resolution in m

Coastal / Aerosol 0.433–0.453 30
Blue 0.450–0.515 30
Green 0.525–0.600 30
Red 0.630–0.680 30
Near Infrared 0.845–0.885 30
Short Wavelength Infrared 1.560–1.660 30
Short Wavelength Infrared 2.100–2.300 30
Panchromatic 0.500–0.680 15
Cirrus 1.360–1.390 30

 

   

 - 

   
   
   

   

_
_ 

 _  
_ 
   
    
 
  

 

     

Figure 2. Workflow of the methodology.

Table 3. Sensor polarization replay values

Polarized Min Max Mean StdDev

HH 0.0005 22.5697 0.0987 0.4116
HV 0.0002 21.2345 0.0325 0.0721

4. Results and Discussion

In order to improve the range of radar images and to distribute them in a better range,
they were converted to decibels after calibration. The results showed that by applying
this algorithm on each band, the output images of west Urmia City, had better brightness,
and the phenomena that were not clear became somewhat more transparent. Calibrated
imagery by the HH and HV bands is displayed in Figures 3 and 4.

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 5 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Figure 3. (a), HH and (b), for the HV band (calibrated) west of Urmia City.

Figure 4. (a), calibrated HH band and, Sigma_0 (b), calibrated HV band Sigma 0 and db west of
Urmia City.

Figure 3, shows A of the HH band, with intensities ranging from (0.01 to 3.62%), and
Figure B of the HV band, which has intensities ranging from (0.001 to 0.157%), which
are brighter than the HH band. The donor is that image B of the HV band evaluates and
identifies urban landscapes with greater accuracy.
While Figure 4a shows the calibrated HH band Sigma 0 converted to decibels, its
values range from – (17.27% to 5.61%), and Figure 4b shows the calibrated HH band
Sigma 0 converted to decibels. Bell is converted, and its values are between (−25.94% and
−8.0%).
Therefore, it is assumed that after converting to Sigma0 and converting it to decibels,
its intensity values are distributed and stretched in such a manner that it causes better and
more precise detection of urban west of Urmia City features in the image. Additionally,
it makes it possible to detect urban features. It Improves the image using the relevant
histograms shown in Figure 5, where the distribution and elongation of the values can be
observed.

4.1. Average Filter

The Average filter is a regular averaging, which changes the central pixel’s value with
the average of the surrounding pixel’s values. This filter has an excellent ability to smooth
out noise. However, treating the moderation action the same will result in a reduction in
resolution around sharp edges. Kernel windows with sizes of 3 × 3 or 5 × 5 are typically
used to reduce this problem [Shanthi and Valarmathi, 2013].
Figure 6, presents the application of the average filter on the HH band with the kernel
size 3 × 3 × 5 × 9 × 9. Larger sizes are mixed, reducing the accuracy of the segregation of
urban toll assessments. In addition to reducing spectral noise, selecting a 3 × 3 kernel size

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 6 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Figure 5. (a) sigma 0 calibrated HH band histogram converted to decibels, (b) and Sigma 0 calibrated
HV band histogram converted to decibels.

Figure 6. Average filter with kernel size (a) 3 × 3, (b) 5 × 5, (c) 9 × 9, applied on HH band (Sigma 0 –
db) west of Urmia City.

also displays the existing urban west of Urmia city features in the image in smaller pixel
sizes. It makes the resolution of the features more accurate.
Figure 7, which, with expanding the kernel size of histogram values, has increased
from (0 to 12.5%), which causes the smoothing of urban west of Urmia city features in the
image Figure 8.
In previous studies, Urban features show themselves more accurately with minor
changes in the values within the specified location. Alenin and Tyo [2014] examined a
different despeckling algorithm for (SAR) data based on the shrinkage of the wavelet and
non-local concepts. LaCasse et al. [2011] conducted research that combined two methods for
constructing images based on mathematical structure. They proved that the demosaicking
method corresponds with the channel filtering technique. However, when utilizing the
implied filters in that technique, the constructed image with available objects in a higher
part of frequencies resulted in a higher-quality image.

4.2. Lee Filter

Jong-Sen-Lee established the Lee filter in 1982; the base concepts of these filters are
Gaussian dispersion. In general, the statistical distribution of the image is the output of
these filters. We acquire some statistical information such as median, Mean and Standard
Deviation. This information provides the knowledge to assess the performance of speckle
noise. The kernel size is also essential in speckle noise reduction and varies from 5 × 5
to 17 × 17 in some software such as snap. The size of kernels is odd numbers [Lee, 1986].

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 7 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Figure 7. Color combination of average filter with kernel size (a) 3 × 3, (b) 5 × 5, (c) 9 × 9, applied on
HH band (Sigma 0 – db) west of Urmia City.

×
×

Figure 8. Mean histogram filter with kernel size (a) 3 × 3, (b) 5 × 5, (c) 9 × 9, applied on HH band
(Sigma 0 – db).

Figure 9, presents the application of the Lee filter on the HH band with a kernel size
3 × 3 × 5 × 9 × 9.
The image in the pixel is mixed in larger sizes, which reduces the resolution of the
features with low density.
However, almost all three selected kernel sizes can accurately distinguish high-density
urban west of Urmia City features. In the image, it should be displayed in smaller pixel
sizes, and the resolution of the effects should be made with better accuracy. In Figure 10,
the color combination of the Lee filter applied on the HH band with the kernel size 3 × 3,
5 × 5, 9 × 9 confirms this result. However, the results obtained from Lee filter histograms on
the HH band with kernel size 3 × 3, 5 × 5, 9 × 9 show Figure 11 that with increasing kernel
size, the histogram values from 0 to 12.5, also decrease to some extent. Finds that make the
image Smooth.
Therefore, it could be mentioned that the more changes in the values in the mentioned
range, the better the effects. In addition to the mentioned range of kernel size values 3 × 3
and 5 × 5, other values are available that have unique functions in differentiating the effects.
However, these values range between from −17.5 to 10. The kernel size histogram 9 × 9 has

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 8 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

values ranging from −15.0 to 10. Expanding the kernel size results in a smoother image
but reduces the accuracy of detecting complications with low density.
Lee et al. [2015] used the boxcar and sigma filters as initial filters. Nevertheless,
extending the intended approach to most available filtering methods is uncomplicated.
Moreover, the spatial filters vary commonly in the way of determining the adjacent uniform
pixels [Lee et al., 1999], the scattering model-based method [Lee et al., 2006], the sigma
range [Lee et al., 2015], the scattering mechanism, and the intensity information [Wang
et al., 2016], homogeneity measurement [Lang et al., 2015], NL [Chen et al., 2011; Lang et al.,
2014], region growing [Vasile et al., 2006].

Figure 9. Lee filter with kernel size (a) 3 × 3, (b) 5 × 5, (c) 9 × 9, applied on the HH band (Sigma 0 –
db) west of Urmia City.

Figure 10. Combined Lee color filter with kernel size (a) 3 × 3, (b) 5 × 5, (c) 9 × 9, applied on HH band
(Sigma 0 – db) west of Urmia City.

4.3. Frost Filter

The Frost filter is a widely used filter with variable parameters. It uses a weighted
average on kernel pixels to soften images. The weight assigned to each pixel is a function
of the kernel adjustment coefficient, mean, variance, and center pixel distance from the
mean in the kernel gray [Frost et al., 1982].
Figure 12, indicates the application of the Frost filter on the HH band with a kernel size
3 × 3 × 5 × 9 × 9; this increases the separation accuracy of high-density features. Nevertheless,
almost all three selected kernel sizes can accurately distinguish high-density features.
However, it is impossible to separate low-density tolls, and it causes urban tolls to be mixed.
In addition to reducing separation accuracy, it also causes classification errors. In this filter,
as in the previous filter, selecting a 3 × 3 kernel size causes the effects in the image to be
displayed in smaller pixel sizes and the resolution of the effects to be done more accurately,

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 9 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Figure 11. Lee filter histogram with kernel size (a) 3 × 3, (b) 5 × 5, (c) 9 × 9, applied on HH band
(Sigma0 – db).

in addition to reducing spectral noise. In Figure 13, the color combination of the Frost
filter applied on the HH band with the kernel size 3 × 3, 5 × 5, 9 × 9 confirms this result.
However, the results obtained from the Frost filter histograms on the HH band with
The kernel size 3×3, 5×5, 9×9 show Figure 14 with increasing the kernel size, the histogram
values decrease from 0 to 12.), to some extent, which makes the image smoother. Therefore,
it can be concluded that more changes in the values in the mentioned range show better
effects on extracting urban west of Urmia City features. Also, kernel size values 3 × 3 and
5 × 5, other than the mentioned range, differ, with values between 17.5.
They have ten and have their unique functions in differentiating different effects.
According to the kernel size histogram 9×9, the values of which range from −15.0 to 10, the
distribution of values of the histogram in this image becomes more compact. However, this
indicates a more compact distribution of values in the histogram. It is worth mentioning
that expanding the kernel size makes the image smoother and reduces the accuracy of
detecting features with low density, making the distribution of intensity values more
compact.
In previous studies, Improved AGGD-based noise reduction is presented to enhance
the quality of the reconstructed image. Moreover, for quality enhancement, the cost of
implementation is lower than the alternative methods (improved AGGD [Golilarz et al.,
2019], HHO [Golilarz et al., 2019], and Bayes [Chang et al., 2000].

Figure 12. Frost filter with kernel size (a) 3 × 3, (b) 5 × 5, (c) 9 × 9, applied on HH band (Sigma 0 – db)
west of Urmia City.

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 10 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Figure 13. (a) frost filter with kernel size 3 × 3, (b) 5 × 5, (c) 9 × 9 applied on HH band (Sigma 0 – db)
west of Urmia City.

4.4. Quantitative SSI Parameter

The quantitative SSI parameter in radar images is used to evaluate the filtering perfor-
mance. The SSI parameter is a standard for quantifying and evaluating the performance
of radar filters in which the average values of playback and standard deviation of im-
ages before and after application are used. Values less than one of these criteria indicate
that the filter performed better. The SSI quantitative parameter was used to evaluate
the performance of LEE, Average, and Frost filters, the results of which are presented in
Table 4.

Table 4. SSI index values in different filter dimensions Average, Lee, Forst

SSI Parameter Filters

0.563 Lee 3 × 3
0.675 Lee 5 × 5
0.896 Lee 7 × 7
0.590 Frost 3 × 3
0.601 Frost 5 × 5
0.789 Frost 9 × 9
0.683 Average 3 × 3
0.630 Average 5 × 5
0.785 Average 9 × 9

4.5. Classification of Images by K-Means Method

According to Figure 14, four classes were extracted from the classification of K-means
on Frost, Lee and Average filters. Comparing filters with the classification results on
Landsat 8 images showed that the closest result is accurate. Above is the Lee filter with
a 3 × 3 size kernel. The total accuracy obtained is (97.36%), and the kappa coefficient is
(0.96%). Between a classified land cover map and a reference or actual land map, the
producer error is related to the percentage of pixels related to the classes but classified
as other classes. On the other hand, user error is an error that is calculated based on user
accuracy and is equivalent to the percentage of pixels that do not belong to the class in
question but are considered in that classification.

5. Conclusion

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 11 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Figure 14. Classification of (a) K-means Average filter, (b) Lee filter, (c) Frost filter west of Urmia
City.

Nowadays, images with the appropriate spatial and spectral resolution are often used
to monitor the earth for various purposes, such as detecting changes, classifying land
cover, and diagnosing complications. These techniques are especially important when
data is used in sensitive areas such as the environment, natural resources, and agriculture.
Based on the outputs obtained in this study, it has been shown that the HV band has
better evaluation and resolution in an urban environment with artificial features such as
buildings and different metal structures due to its polarization properties and sensitivity
to the phenomenon in the city. In this study, different filters were applied on the images of
both AlosPalsar satellite bands to estimate the degree of noise-free gray in the northwest of
Urmia City according to the averages through the SSI index. The results include filters that
can help reduce spike noise in AlosPalsar satellite radar images, such as the Lee filter with
3 × 3 kernel size.
Additionally, it is essential to note that the output from the filters may vary depending
on the bands of the AlosPalsar satellite. However, after conducting K-means classification
on Lee, Gamma-Map, Frost, and Average filters, the Lee filter with a 3 × 3 kernel size
has been identified to provide the closest results to surface phenomena. Specifically, it
produced a kappa coefficient of 0.96% and an accuracy of 97.36% in the northwest area of
Urmia. Future research could include exploring the impact of different filter parameters
on noise reduction and feature extraction accuracy in urban environments and comparing
the results of different satellite systems in monitoring earth changes in sensitive areas.

Acknowledgments. We, the authors, thank the respected reviewers for reviewing this
article. The research does not receive any funding. Data will be available based on request.

References
Abdollahi, A., H. R. R. Bakhtiari, and M. P. Nejad (2017), Investigation of SVM and Level Set Interactive Methods
for Road Extraction from Google Earth Images, Journal of the Indian Society of Remote Sensing, 46(3), 423–430,
https://doi.org/10.1007/s12524-017-0702-x.

Alenin, A. S., and J. S. Tyo (2014), Generalized channeled polarimetry, Journal of the Optical Society of America A, 31(5),
1013, https://doi.org/10.1364/josaa.31.001013.

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 12 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Alves, W. A. L., C. F. Gobber, D. J. Silva, A. Morimitsu, R. F. Hashimoto, and B. Marcotegui (2020), Image segmentation
based on ultimate levelings: From attribute filters to machine learning strategies, Pattern Recognition Letters, 133,
264–271, https://doi.org/10.1016/j.patrec.2020.03.013.

Attarchi, S., M. Poorrahimi, and V. Isazade (2020), Comparison of spectral indices and object-based classification
for built-up area extraction in different urban areas, Geographical Urban Planning Research (GUPR), 8(1), 23–43,
https://doi.org/10.22059/jurbangeo.2020.299492.1249.

Betbeder, J., S. Rapinel, S. Corgne, E. Pottier, and L. Hubert-Moy (2015), TerraSAR-X dual-pol time-series for mapping
of wetland vegetation, ISPRS Journal of Photogrammetry and Remote Sensing, 107, 90–98, https://doi.org/10.1016/j.
isprsjprs.2015.05.001.

Camargo, F. F., E. E. Sano, C. M. Almeida, J. C. Mura, and T. Almeida (2019), A Comparative Assessment of Machine-
Learning Techniques for Land Use and Land Cover Classification of the Brazilian Tropical Savanna Using ALOS-
2/PALSAR-2 Polarimetric Images, Remote Sensing, 11(13), 1600, https://doi.org/10.3390/rs11131600.

Chang, S. G., B. Yu, and M. Vetterli (2000), Adaptive wavelet thresholding for image denoising and compression, IEEE
Transactions on Image Processing, 9(9), 1532–1546, https://doi.org/10.1109/83.862633.

Chen, J., Y. Chen, W. An, Y. Cui, and J. Yang (2011), Nonlocal Filtering for Polarimetric SAR Data: A Pretest Approach,
IEEE Transactions on Geoscience and Remote Sensing, 49(5), 1744–1754, https://doi.org/10.1109/tgrs.2010.2087763.

Dai, J., T. Zhu, Y. Wang, R. Ma, and X. Fang (2020), Road Extraction From High-Resolution Satellite Images Based on
Multiple Descriptors, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 13, 227–240,
https://doi.org/10.1109/jstars.2019.2955277.

Dass, R., and N. Yadav (2020), Image Quality Assessment Parameters for Despeckling Filters, Procedia Computer Science,
167, 2382–2392, https://doi.org/10.1016/j.procs.2020.03.291.

Fazel, N., R. Berndtsson, C. B. Uvo, K. Madani, and B. Kløve (2017), Regionalization of precipitation characteristics in
Iran’s Lake Urmia basin, Theoretical and Applied Climatology, 132(1-2), 363–373, https://doi.org/10.1007/s00704-017
-2090-0.

Frost, V. S., J. A. Stiles, K. S. Shanmugan, and J. C. Holtzman (1982), A Model for Radar Images and Its Application
to Adaptive Digital Filtering of Multiplicative Noise, IEEE Transactions on Pattern Analysis and Machine Intelligence,
PAMI-4(2), 157–166, https://doi.org/10.1109/tpami.1982.4767223.

Ghanbari, A., V. Isazade, and Z. Alibeigy (2022), Analysis of Spatial Justice in Dealing with Urban Facilities Using
Quantitative, Spatial and Moran Correlation Models (Study Area: Isfahan), Journal Research and Urban Planning, 13(50),
240–254.

Ghorbanian, A., M. Kakooei, M. Amani, S. Mahdavi, A. Mohammadzadeh, and M. Hasanlou (2020), Improved land
cover map of Iran using Sentinel imagery within Google Earth Engine and a novel automatic workflow for land cover
classification using migrated training samples, ISPRS Journal of Photogrammetry and Remote Sensing, 167, 276–288,
https://doi.org/10.1016/j.isprsjprs.2020.07.013.

Golilarz, N. A., H. Gao, and H. Demirel (2019), Satellite Image De-Noising With Harris Hawks Meta Heuristic Opti-
mization Algorithm and Improved Adaptive Generalized Gaussian Distribution Threshold Function, IEEE Access, 7,
57,459–57,468, https://doi.org/10.1109/access.2019.2914101.

Hu, T., J. Yang, X. Li, and P. Gong (2016), Mapping Urban Land Use by Using Landsat Images and Open Social Data,
Remote Sensing, 8(2), 151, https://doi.org/10.3390/rs8020151.

Isazade, V., A. Ghanbari, and K. Valizadeh (2021), Evaluation of spectral indices and extraction of constructed and non-
constructed urban features and its comparison with ground surface temperature using Landsat 7 and 8 satellite images
(Study area, Tehran), Journal of Spatial Planning, 11(4), 23–39, https://doi.org/10.30488/GPS.2021.243118.3283.

Isazade, V., A. B. Qasimi, and E. Isazade (2022), Environmental dust effect phenomenon on the sustainability of urban
areas using remote sensing data in GEE, Safety in Extreme Environments, 5(1), 59–67, https://doi.org/10.1007/s42797
-022-00067-z.

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 13 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Kaushik, K., and D. Kumar (2020), Forest Disturbance Mapping with Microwave Remote Sensing, Journal of Landscape
Ecology, 13(2), 95–128, https://doi.org/10.2478/jlecol-2020-0012.

Kraus, K. (2003), Photogrammetry, Remote Sensing and Spatial Information Sciences for heritage documentation, in
Proceedings of the XIXth International Symposium, CIPA 2003: new perspectives to save cultural heritage, pp. 17–24, CIPA
2003 Organising Committee, Antalya (Turkey).

LaCasse, C. F., R. A. Chipman, and J. S. Tyo (2011), Band limited data reconstruction in modulated polarimeters, Optics
Express, 19(16), 14,976, https://doi.org/10.1364/oe.19.014976.

Lang, F., J. Yang, D. Li, L. Shi, and J. Wei (2014), Mean-Shift-Based Speckle Filtering of Polarimetric SAR Data, IEEE
Transactions on Geoscience and Remote Sensing, 52(7), 4440–4454, https://doi.org/10.1109/tgrs.2013.2282036.

Lang, F., J. Yang, and D. Li (2015), Adaptive-Window Polarimetric SAR Image Speckle Filtering Based on a Homogeneity
Measurement, IEEE Transactions on Geoscience and Remote Sensing, 53(10), 5435–5446, https://doi.org/10.1109/tgrs.20
15.2422737.

Lapini, A., S. Pettinato, E. Santi, S. Paloscia, G. Fontanelli, and A. Garzelli (2020), Comparison of Machine Learning
Methods Applied to SAR Images for Forest Classification in Mediterranean Areas, Remote Sensing, 12(3), 369, https:
//doi.org/10.3390/rs12030369.

Lee, J.-S. (1986), Speckle Suppression And Analysis For Synthetic Aperture Radar Images, Optical Engineering, 25(5),
255,636, https://doi.org/10.1117/12.7973877.

Lee, J.-S., M. R. Grunes, and G. de Grandi (1999), Polarimetric SAR speckle filtering and its implication for classification,
IEEE Transactions on Geoscience and Remote Sensing, 37(5), 2363–2373, https://doi.org/10.1109/36.789635.

Lee, J.-S., M. R. Grunes, D. L. Schuler, E. Pottier, and L. Ferro-Famil (2006), Scattering-model-based speckle filtering of
polarimetric SAR data, IEEE Transactions on Geoscience and Remote Sensing, 44(1), 176–187, https://doi.org/10.1109/
tgrs.2005.859338.

Lee, J.-S., T. L. Ainsworth, Y. Wang, and K.-S. Chen (2015), Polarimetric SAR Speckle Filtering and the Extended Sigma
Filter, IEEE Transactions on Geoscience and Remote Sensing, 53(3), 1150–1160, https://doi.org/10.1109/tgrs.2014.23351
14.

Leiva-Murillo, J. M., L. Gomez-Chova, and G. Camps-Valls (2013), Multitask Remote Sensing Data Classification, IEEE
Transactions on Geoscience and Remote Sensing, 51(1), 151–161, https://doi.org/10.1109/TGRS.2012.2200043.

Li, Y., S. Wang, Q. Zhao, and G. Wang (2020), A new SAR image filter for preserving speckle statistical distribution, Signal
Processing, 176, 107,706, https://doi.org/10.1016/j.sigpro.2020.107706.

Liu, N., Z. Cao, Z. Cui, Y. Pi, and S. Dang (2019), Multi-Scale Proposal Generation for Ship Detection in SAR Images,
Remote Sensing, 11(5), 526, https://doi.org/10.3390/rs11050526.

Mohan, M., and A. Kandya (2015), Impact of urbanization and land-use/land-cover change on diurnal temperature
range: A case study of tropical urban airshed of India using remote sensing data, Science of The Total Environment,
506-507, 453–465, https://doi.org/10.1016/j.scitotenv.2014.11.006.

Mun, J., Y. Jang, Y. Nam, and J. Kim (2019), Edge-enhancing bi-histogram equalisation using guided image filter, Journal
of Visual Communication and Image Representation, 58, 688–700, https://doi.org/10.1016/j.jvcir.2018.12.037.

NISAR (2018), NASA-ISRO SAR (NISAR) Mission Science Users’ Handbook, NASA Jet Propulsion Laboratory.

Nouri, H., R. J. Mason, and N. Moradi (2017), Land suitability evaluation for changing spatial organization in Urmia
County towards conservation of Urmia Lake, Applied Geography, 81, 1–12, https://doi.org/10.1016/j.apgeog.2017.02.
006.

Osgouei, P. E., S. Kaya, E. Sertel, and U. Alganci (2019), Separating Built-Up Areas from Bare Land in Mediterranean
Cities Using Sentinel-2A Imagery, Remote Sensing, 11(3), 345, https://doi.org/10.3390/rs11030345.

Podest, E. (2018), SAR for Mapping Land Cover, National Aeronautics and Space Administration.

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 14 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Qasimi, A. B., V. Isazade, G. Kaplan, and Z. Nadry (2022), Spatiotemporal and multi-sensor analysis of surface tempera-
ture, NDVI, and precipitation using google earth engine cloud computing platform, Russian Journal of Earth Sciences,
22(6), 1–12, https://doi.org/10.2205/2022ES000812.

Rasche, C. (2018), Rapid contour detection for image classification, IET Image Processing, 12(4), 532–538, https://doi.org/
10.1049/iet-ipr.2017.1066.

Routray, S., P. P. Malla, S. K. Sharma, S. K. Panda, and G. Palai (2020), A new image denoising framework using bilateral
filtering based non-subsampled shearlet transform, Optik, 216, 164,903, https://doi.org/10.1016/j.ijleo.2020.164903.

Seo, S. (2018), Subpixel Edge Localization Based on Adaptive Weighting of Gradients, IEEE Transactions on Image
Processing, 27(11), 5501–5513, https://doi.org/10.1109/tip.2018.2860241.

Seo, S. (2019a), Subpixel Line Localization With Normalized Sums of Gradients and Location Linking With Straightness
and Omni-Directionality, IEEE Access, 7, 155–167, https://doi.org/10.1109/access.2019.2959320.

Seo, S. (2019b), Line-Detection Based on the Sum of Gradient Angle Differences, Applied Sciences, 10(1), 254, https:
//doi.org/10.3390/app10010254.

Serwa, A. (2020), Studying the Potentiality of Using Digital Gaussian Pyramids in Multi-spectral Satellites Images
Classification, Journal of the Indian Society of Remote Sensing, 48(12), 1651–1660, https://doi.org/10.1007/s12524-020
-01173-w.

Serwa, A., and S. Elbialy (2021), Enhancement of classification accuracy of multi-spectral satellites’ images using
Laplacian pyramids, The Egyptian Journal of Remote Sensing and Space Science, 24(2), 283–291, https://doi.org/10.1016/
j.ejrs.2020.12.006.

Serwa, A., and M. Saleh (2021), New semi-automatic 3D registration method for terrestrial laser scanning data of bridge
structures based on artificial neural networks, The Egyptian Journal of Remote Sensing and Space Science, 24(3, Part 2),
787–798, https://doi.org/10.1016/j.ejrs.2021.06.003.

Seto, K. C., B. Güneralp, and L. R. Hutyra (2012), Global forecasts of urban expansion to 2030 and direct impacts
on biodiversity and carbon pools, Proceedings of the National Academy of Sciences, 109(40), 16,083–16,088, https:
//doi.org/10.1073/pnas.1211658109.

Shanthi, I., and M. L. Valarmathi (2013), SAR image despeckling using possibilistic fuzzy C-means clustering and edge
detection in bandelet domain, Neural Computing and Applications, 23(S1), 279–291, https://doi.org/10.1007/s00521-0
13-1394-y.

Shi, Y., Q. Li, and X. X. Zhu (2019), Building Footprint Generation Using Improved Generative Adversarial Networks,
IEEE Geoscience and Remote Sensing Letters, 16(4), 603–607, https://doi.org/10.1109/LGRS.2018.2878486.

Shokati, B., and B. Feizizadeh (2018), Sensitivity and uncertainty analysis of agro-ecological modeling for saffron plant
cultivation using GIS spatial decision-making methods, Journal of Environmental Planning and Management, 62(3),
517–533, https://doi.org/10.1080/09640568.2018.1427561.

Sun, H., W. Forsythe, and N. Waters (2007), Modeling Urban Land Use Change and Urban Sprawl: Calgary, Alberta,
Canada, Networks and Spatial Economics, 7(4), 353–376, https://doi.org/10.1007/s11067-007-9030-y.

Vasile, G., E. Trouve, J.-S. Lee, and V. Buzuloiu (2006), Intensity-driven adaptive-neighborhood technique for polarimetric
and interferometric SAR parameters estimation, IEEE Transactions on Geoscience and Remote Sensing, 44(6), 1609–1621,
https://doi.org/10.1109/tgrs.2005.864142.

Wang, Y., J. Yang, and J. Li (2016), Similarity-intensity joint PolSAR speckle filtering, in 2016 CIE International Conference
on Radar (RADAR), pp. 1–5, https://doi.org/10.1109/RADAR.2016.8059213.

Wei, C., and T. Blaschke (2018), Pixel-Wise vs. Object-Based Impervious Surface Analysis from Remote Sensing: Correla-
tions with Land Surface Temperature and Population Density, Urban Science, 2(1), https://doi.org/10.3390/urbansci2
010002.

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 15 of 16

Integrating Passive and Active Remote Sensing Data with Spatial Filters. . . Isazade et al.

Yang, C. H., B. K. Kenduiywo, and U. Soergel (2016), Change detection based on persistent scatterer interferometry
– a new method of monitoring building changes, ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial
Information Sciences, III-7, 243–250, https://doi.org/10.5194/isprs-annals-III-7-243-2016.

Zakeri, H., F. Yamazaki, and W. Liu (2017), Texture Analysis and Land Cover Classification of Tehran Using Polarimetric
Synthetic Aperture Radar Imagery, Applied Sciences, 7(5), 452, https://doi.org/10.3390/app7050452.

Zia, S., S. A. Shirazi, M. N. Bhalli, and S. Kausar (2015), The Impact Of Urbanization On Mean Annual Temperature Of
Lahore Metropolitan Area, Pakistan, Pakistan Journal of Science, 67(3).

Russ. J. Earth. Sci. 2023, 23, ES5012, EDN: TSPARR, https://doi.org/10.2205/2023es000861 16 of 16