Hardware-transmitter to receiver modules and antennas setup

Figure 1 illustrates a transmitter and receiver module configuration having transmitting and receiving antennas for wireless image capture and transfer.

Figure 1. Arrangement of image transferring using TS832 transmitter and RC832 receiver module.

The transmitter and the receiver modules with their respective antennas are placed at a fixed distance D from each other. This handheld transmitter module with battery can move anywhere from the receiver module (with antenna) while varying the distance D up to 5 m with a step size of 0.5 m in different directions. A camera connected to the transmitter module captures real-time images, which are transmitted at a frequency of 5.8 GHz from the transmitter module. The signals are received by the receiver module and processed to give two analog output ports. To de-embed the images in any laptop or desktop computer, an audio/video to USB converter module was used. A stand-alone GUI application is developed to take the picture or image; Figure 2 depicts the operational flow chart. The BRISQUE method is created in Python code to assess the quality of the acquired picture [Reference Mittal and Bovik37]. This BRISQUE method maps the combined data into a numerical value, which is known as the image quality score (IQS) and its value varies from 0 (very good quality) to 100 (very bad quality).

Figure 2. Flowchart of graphical user interface (GUI) for image capturing.

The experiment setup is realized by using transmitter model TS832 audio/video, receiver model RC832 audio/video, a pair of standard dipole antennas, camera model Run Cam Nano 4, and batteries for 12 V DC. These models can transmit/receive the frequency-modulated carrier frequencies in the frequency range (FR)1 band of 5G communication.

Use of BRISQUE method

To evaluate image quality, most of the assessment techniques require a reference image and first, these techniques analyze image structure and identify patterns among its features. However, the BRISQUE method is considered particularly effective as it uses image pixel information to calculate these features [Reference Mittal and Bovik37, Reference Chawdhary, Kumari, Bhavsar and Verma38]. Along with the pairwise product coefficients, the design of the BRISQUE-based model takes into account the NSS of locally normalized brightness coefficients in the spatial domain. The process model, shown in Fig. 3, is developed taking these considerations into account.

Figure 3. Process model of the BRISQUE method used in this work.

The initial step involves subtracting the local mean from the image to obtain locally normalized luminescence, which is then divided by the local deviation. To prevent division by zero, a constant is included. Before computing the mean subtracted contrast normalized (MSCN) coefficients for a grayscale image, it is necessary to first calculate the local mean. The intensity value of the pixel I (i, j) is given by Eq. (1) as

(1)\begin{equation}{{\skew6\hat I}}\left( {{\text{i,}}\;{\text{j}}} \right){\text{ }} = \frac{{{\text{I}}\left( {{\text{i}},{\text{j}}} \right) - {{\mu }}\left( {{\text{i}},{\text{j}}} \right)}}{{{{\sigma }}\left( {{\text{i}},{\text{j}}} \right) + {\text{C}}}}\end{equation}

where i ∈ 1, 2 … M, j ∈ 1, 2 … N are spatial indices, M and N represent the height and width of the image, respectively. As the denominator gets closer to zero, a constant value of C = 1 is used to prevent any instability problems. The terms µ(i, j) and σ(i, j) present the local mean of the pixel’s neighborhood and the local standard deviation of the same neighborhood, respectively, and obtained using Eqs. (2) and (3) as follows.

(2)\begin{equation}{{\mu }}\left( {{\text{i,}}\;{\text{j}}} \right) = \mathop \sum \limits_{{\text{k}} = - {\text{K}}}^{\text{K}} \mathop \sum \limits_{{\text{l}} = - {\text{L}}}^{\text{L}} {{\text{w}}_{{\text{k}},{\text{l}}}}{{\text{I}}_{{\text{k}},{\text{l}}}}\left( {{\text{i}},{\text{j}}} \right)\end{equation}

(3)\begin{equation}\sigma \left( {{\text{i,}}\;{\text{j}}} \right){\text{ }} = \sqrt {\mathop \sum \limits_{{\text{k}} = - {\text{K}}}^{\text{K}} \mathop \sum \limits_{{\text{l}} = - {\text{L}}}^{\text{L}} {{\text{w}}_{{\text{k}},{\text{l}}}}{{\text{I}}_{{\text{k}},{\text{l}}}}\left( {{\text{i}},{\text{j}}} \right)} \end{equation}

where $w = {\text{ }}\left\{ {w{\text{ }}k,l,k = {\text{ }} - K,{\text{ }}.{\text{ }}.{\text{ }}.,{\text{ }}K,{\text{ }}l = {\text{ }} - L,{\text{ }}.{\text{ }}.{\text{ }}.{\text{ }}L} \right\}$ is the implementation employs a circularly symmetric 2D Gaussian weighting function, sampled up to 3 standard deviations and adjusted to have a unit volume. In this case, K and L, which stand for the dimensions of the square window centered on the pixel of interest, are set to 3. The window size is an adjustable parameter that can be set based on the specific use case and the properties of the image.

The pixel’s contrast about its immediate neighborhood is represented by the resulting MSCN value, where positive values denote a higher contrast than the mean and negative values represent a lower contrast than the mean. The existence of distortion can affect the distinctive statistical characteristics of MSCN coefficients. By analyzing the changes in the statistical characteristics of an image, one can predict the effect of distortion on the image and its perceived quality. The MSCN coefficients are distributed as a GGD. The density function of the GGD is,

(4)\begin{equation}f\left( {x;a,{{{\sigma }}^2}} \right) = \frac{a}{{2\beta \Gamma \left( {\frac{1}{a}} \right)}}\exp \left( { - {{\left( {\frac{{\left| x \right|}}{\beta }} \right)}^a}} \right)\end{equation}

(5)\begin{equation}\beta = \sigma \sqrt {\frac{{\Gamma \left( {1/a} \right)}}{{\Gamma \left( {3/a} \right)}}} \end{equation}

where β is the parameter that affects the shape of the distribution and Г is the gamma function. The parameter α controls the form and σ ² is the variance. Pairwise products of neighboring MSCN coefficients along four directions (1) horizontal H, (2) vertical V, (3) main-diagonal D1, and (4) secondary-diagonal D2 are considered and obtained using Eq. (6)–Eq. (9) as

(6)\begin{equation}H\left( {i,j} \right) = {{\skew6\hat I}}\left( {i,j} \right){{\skew6\hat I}}\left( {i,j + 1} \right)\end{equation}

(7)\begin{equation}V\left( {i,j} \right) = {{\skew6\hat I}}\left( {i,j} \right){{\skew6\hat I}}\left( {i + 1,j} \right)\end{equation}

(8)\begin{equation}D2\left( {i,j} \right) = {{\skew6\hat I}}\left( {i,j} \right)\;{{\skew6\hat I}}\left( {i + 1,j - 1} \right)\end{equation}

(9)\begin{equation}D2\left( {i,j} \right) = {{\skew6\hat I}}\left( {i,j} \right)\;{{\skew6\hat I}}\left( {i + 1,j - 1} \right)\end{equation}

Since GGD does not fit the empirical histograms of coefficient products well. Therefore, the model is fitted to asymmetric generalized Gaussian distribution. Its density function is given as,

(10)\begin{equation}f\left( {x;a,{{\sigma }}_l^2{{\sigma }}_r^2} \right) = \left\{ {\begin{array}{*{20}{c}} {\frac{v}{{\left( {{\beta _l} + {\beta _r}} \right)\Gamma \left( {\frac{1}{v}} \right)}}\exp \left( { - {{\left( {\frac{{ - x}}{{{\beta _l}}}} \right)}^v}} \right),\,\,x \lt 0} \\ {\frac{v}{{\left( {{\beta _l} + {\beta _r}} \right)\Gamma \left( {\frac{1}{v}} \right)}}\exp \left( { - {{\left( {\frac{x}{{{\beta _l}}}} \right)}^v}} \right),\,\,x \geq 0} \end{array}} \right.\end{equation}

As the BRISQUE approach performs well in real time, is very accurate, and does not require reference images, it is the method preferred for assessing image quality. Equations (1–10) are used to develop a Python code.

Antennas used in the experiments

Four different antennas are employed as seen in Fig. 4, to assess the best antenna preferred for the intended image transfer application. Table 1 lists each antenna’s attributes.

Figure 4. (a) Dipole antenna, (b) front and (c) back side H-SRR antenna, (d) front and (e) backside C-SRR antenna, (f) front and (g) backside metamaterial (double negative index) Tx antenna, (h) front and (i) backside metamaterial (epsilon near zero) Rx antenna.

Table 1. Different antennas and their features

C-SRR and H-SRR structures are a Mu-negative type of metamaterial and these structures act like an LC resonance circuit. When such a structure is used in place of a patch in a microstrip antenna [Reference Sehgal and Patel49], it radiates the same resonance frequency with a moderate gain and bandwidth. The resonance or radiated frequency is defined by the physical dimensions of SRRs and the gain between them [Reference Sehgal and Patel50]. A three-layer metamaterial Tx antenna consists of the complementary SRR on the top patch layer, microstrip feed on the middle layer, and grid patterns on the bottom layer and is a type of a double-negative metamaterial which gives wideband and high gain [Reference Yılmaz and Yaman51], whereas a three-layer metamaterial Rx antenna is made up of three layers: a middle layer of air, a top layer with circular patches etched from the Cu layer as the epsilon near zero layer, and a bottom layer with a rectangular patch of inset-fed microstrip with parasitic elements. It is designed for very high gain with a narrow beam at 5.8 GHz.

Experimental setup for image capturing

Four setups are considered in this study to perform the image transfer as shown in Fig. 4. In these setups, the receiver with antenna is placed in a laboratory on the ground floor of the building, whereas the transmitter with antenna is placed at different locations. In Setup 1, the transmitter is placed horizontally inside the same lab with a varying distance from 0 to 5 m with a step size of 0.5 m (Fig. 5(a)), while in Setup 2 (Fig. 5(b)), the transmitter is placed vertically inside the same lab with a varying 0 to 5 m with step size 0.5 m. In other setups, the transmitter is placed in different locations on the ground floor in the same building (Setup 3) which is represented by flag icons in Fig. 5(c), and the transmitter is placed in different locations on the first floor in the building (Setup 4) represented by flag icons in Fig. 5(d).

Figure 5. Image transfer using dipole antennas placed in (a) horizontal setup (Setup1), (b) vertical setup (Setup2), (c) ground floor (Setup3), and (d) first floor (Setup4).

First, a pair of dipole antennas are used in all mentioned four setups as shown in Fig. 5(a) and (b), and images are captured, transferred and processed for IQA. In the same manner, a pair of H-SRR antennas and C-SRR antennas and metamaterial Tx and Rx (MMTR) antennas are used to transfer the images using Setup1 and Setup 3, which are illustrated in Fig. 6.

Figure 6. Setup 1 using (a) H-SRR antennas, (b) C-SRR antennas, and (c) metamaterial Tx and Rx antennas.

Image capture GUI app

Python programing has been used to create a GUI, as seen in Fig. 7. This application is capable of capturing images. It is completely a stand-alone application that doesn’t require any Python-oriented platform for its functioning. Given the way the developed programing, the user can only select steps to be performed by tapping the buttons on the application. First, double-click on the app icon to start, and a new graphical window will open if the receiver end is already connected to the laptop port, we can see the visuals in the webcam feed sub-window, for capturing the image by pressing the capture button and specify the path by browsing option and we can check the image saved or not using image preview sub window by browsing image.

Figure 7. Graphical user interface (GUI) for image capturing.

Images transferred using a pair of dipole antennas

The image taken in Setup 2 at a distanc of 1 m is shown in Fig. 8, and it has the lowest IQS score of 37.97 out of the four images and is of the highest quality. However, Setups 3 and 4’s IQSs of 41.68 and 40.90 demonstrate that the proposed wireless image transfer scheme works regardless of obstructions and orientations.

Figure 8. IQS of images transferred using a pair of dipole antennas for different setups.

Comparing the effectiveness of different antennas for image transfer

Experiments are carried out using three additional antennas for each of the two setups, Setup 1 and Setup 3, to determine which antenna set is best for the suggested application. For these configurations, the images with their IQS acquired are displayed in Figs. 9 and 10, respectively.

Figure 9. Comparison of IQSs of images for Setup 1.

Figure 10. Comparison of IQSs of images for Setup 3.

On observing Fig. 9, it is noted that Setup1 with dipole antennas offers the best IQS value of 46.390 than other antennas up to 1 m of distance. For larger distances of 2 and 3 m, the use of MMTR antennas provides IQS values of 42.065 and 40.231, respectively. Figure 10 shows the images and their IQS values obtained in Setup 3. The lowest value of IQS obtained using MMTR antennas confirms that the higher gain antennas with narrow beamwidth (Table 1) are best for wireless image transfer. The higher gain of both three-layer metamaterial antennas with good bandwidth and narrow beam width led to more efficient radiated power transmission and focused reception, which improved the signal transmission/reception than the C-SRR and H-SRR antennas. So, the use of MMTR antennas offered good quality image transfer and a longer range. Additionally, IQS values are thoroughly assessed for every image utilizing each of the four antenna sets, and the results are shown in Fig. 11. Using Eq. (11) to estimate the variance in IQS values, performances can be evaluated efficiently against the standard dipole.

(11)\begin{align} {\rm IQS\ deviation} \left( {\Delta} \right) & = {\rm IQS\ of\ different\ antennas}\nonumber \\ & \qquad - {\rm IQS\ of\ Dipole\ Antenna}\end{align}

Figure 11. Comparison of image quality score for Setup 1: (a) dipole and C-SRR antennas, (b) dipole and H-SRR antennas, (c) dipole and metamaterial antennas, and (d) IQS deviation with respect to dipole antenna.

According to Fig. 11(a), the IQS of the dipole antenna is lower than the IQS of the C-SRR antenna for distances up to 1.5 m and then becomes higher. So, the quality of images obtained using C-SRR antennas is better after these distances. Notably, the IQS of the C-SRR antenna remained relatively constant as the distance between the transmitter and receiver was further increased. Given that Table 1’s antenna characteristics are the same for both C-SRR and H-SRR antennas, it can be observed in Fig. 11(b) that the IQS of the H-SRR antenna is comparable to those generated from C-SRR antennas. Again, as in Fig. 11(c), the IQS of the dipole antenna is about 48 and much better than that of the metamaterial (MMTR) antennas for distances up to 1.5 m. After this range, the metamaterial antennas exhibited further lower values of IQS, which shows that the quality of images gradually enhanced with longer distances while the performance of C-SRR and H-SRR is nearly constant at the same distance. Figure 11(d) shows the IQS deviation (Δ) of different antennas with respect to the dipole antenna. The IQS deviation of the C-SRR and H-SRR antenna is negative for most of the distance, it implies that C-SRR and H-SRR are better than dipole antenna. The IQS deviation of metamaterial antennas is found to be more positive as compared to the other two SRR antennas up to 1.5 m, which implies that the use of metamaterial antennas is unsuitable for very small or close distances. However, after 2 m, the deviation becomes negative and further negative with distance increases, which confirms the image quality is becoming better using these antennas for longer distances.

To further validate the comparison, images taken at the same locations using four antennas in Setup3 are analyzed and IQS values are shown location-wise of Setup 3 in Fig. 12. Figure 12 confirms that the dipole and C-SRR antennas are providing higher values of IQS in the same floor of building while H-SRR and metamaterial antennas offer better IQS values. In addition, the IQS values obtained using H-SRR and metamaterial antennas remain almost constant and independent of the direction or distance of the transmitter from the receiver position, even though the metamaterial antennas are more directional. The results show that presented metamaterial antennas are best for the proposed application of wireless image transfer.

Figure 12. Comparison of IQS for Setup 3.

To show the novelty in the current research work a comparison with the previously published work is shown in Table 2. In comparison, most of the previous techniques involve complex processing and encryption algorithms to improve the quality of transmitted or encrypted images, whereas the proposed technique of wireless image transfer is a simple, cheaper, and faster technique that can be implemented with a laptop and a little signal processing. Most of the reported techniques are verified by simulation in various scenarios and modulations, the proposed technique is implemented physically and the quality of received or transferred images is assessed in terms of IQS.

Table 2. Comparison of present work with previously published literature

PSNR = peak signal-to-noise ratio; MSE = mean square error; SSIM = structure similarity index measure; IQS = image quality score; NPCR = number of changing pixel rate; UACI = unified averaged changed intensity.

Limitation and future scope

The real-time image quality can be improved further by optimization of antenna designs, including metamaterial antennas, and the use of advanced signal processing methods. These advanced signal processing algorithms are required to enhance image quality by noise reduction, and compression techniques, to minimize data loss during transmission. Also, the optimal deployment and positioning of antennas would maximize signal coverage and minimize signal attenuation, by considering factors such as antenna height, orientation, and spatial arrangement in the surveillance area. Also, in this work, a camera model RunCam Nano 4 is used to keep the proposed application cheaper and handy. By replacing it with a good quality camera, the quality of images can be improved directly. The limitation of the present work is the use of bulky and large-size MMTR antennas for better image quality. In addition to improving the image quality as mentioned in the response to previous comment 7, the new antenna design should focus on the compact high gain and narrow beam width antennas. Also, to increase the range of transmission/reception beyond 100 m, a similar analysis should be performed at a lower frequency like 3.5 GHz in the 5G FR1 band.