Open Access
Add to BibSonomy
Abstract
In this study, we propose a parallel processing method for analyzing video-image radiation-response signals and suppressing radiation noise. We studied the linear-representation law of various image-information components on the radiation dose rate. Subsequently, the simulation images were used to examine the response-signal extract and radiation-noise suppression. The results indicate that the majority of response signals in the global image comprise forward superposition. The peak signal-to-noise ratio of the red channel was significantly improved when the noise signal-substitution algorithm and median filter were applied successively. Real-time radiation dose-rate measurements and clear images under irradiation can be obtained simultaneously.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
Corrections
23 January 2023: A typographical correction was made to the author listing.
1. Introduction
In this COVID-19 era, nuclear energy and nuclear technology have strengthened their strategic standing, and many nuclear energy technologies are also advancing rapidly [1]. Nuclear safety and security are the most complex technical aspects of national security. Moreover, advancements in nuclear safety and nuclear emergency technology are a global concern. Nuclear radiation-detection technology based on video images is a novel method for detecting radiation in unknown and complex environments, such as nuclear accidents, emergencies, and terrorist attacks. Moreover, in an unknown and complex radiation environment, it is desirable to comprehend the environmental information using clear video images. However, nuclear radiation contains a significant amount of nuclear noise in video images, making it difficult to apply advanced intelligent algorithms to nuclear-radiation environment monitoring.
Monolithic active pixel sensors (MAPS) are solid-state image sensor used for visible light acquisition, and they are widely used in a variety of fields, such as digital photography, safety and security, medicine, and astronomy [2]. In the past decade, the application of the MAPS has expanded tremendously, particularly in the detection of nuclear radiation, and they have garnered considerable attention from researchers. Studies have demonstrated that MAPS can be used for video monitoring in radioactive environments and that the effects of ionizing radiation can be reflected in imaging signals [3–5]. The single-event transient (SET) effect of ionized particles generates radiation-response events in frames with parameters that reflect the radioactivity level of the radiation environment. The count and value of the “bright spot” produced by the SET can reflect the radiation dose rate [6,7], and the radiation field can be directly detected without using a scintillation crystal as the conversion layer [8]. This has a wide range of radiation-dose-rate response intervals, high upper detection limit, excellent radiation resistance, and strong adaptability to complex environments [9].
MAPS that ionize radiation detection are commonly used in interventional radiology. In recent years, they has been used to monitor the personal doses of medical personnel and patients [10–12], detect charged particle tracks [13], and detect nuclear radiation using mobile phone cameras [14]. Research on the application of radiation detection has garnered considerable interest [15–19]. Experimentation demonstrates that MAPS are sensitive to individual ionized particles and generate radiation-responsive events in pixel arrays [20–22,10]. Knowing whether the algorithm can effectively preserve the image edge information while reducing the image noise is crucial. Common traditional image-denoising algorithms, such as median filtering and mean filtering, frequently result in the loss of some image details, which reduces image-edge pixel density and image quality. Many improved composite image denoising methods can significantly reduce noise while preserving edges [23,24]. However, these methods primarily focus on luminance information, and the color component of the image is deemed insufficient [25,26]. Radiation-noise reduction necessitates additional research, but there are numerous types of particle damage effects on image sensors. Ionizing radiation also reacts with various substances in the environment to generate interference information, such as fluorescence, leading to interference noise with varying characteristics in the collected video images. Therefore, there are still many questions to be answered regarding the characteristics of radiation noise, and algorithms for the efficient and robust reduction of nuclear radiation noise have great research and application value.
In this study, we first investigate response characteristics under various radiation doses and then discuss the effect of the response signal on the background information of various video images and the representation form. Horizontal, vertical, and diagonal detail images under different radiation doses were obtained through tow-dimensional (2D) wavelet packet decomposition (WPD), and their histograms were analyzed to determine the best image detail to characterize the change in the radiation level. The effect of texture information characteristics of the best image detail on the representation of radioactivity was studied by rotating the images. In conclusion, the simulation image was used to extract the response signal and study the radiation-noise suppression, and the optimal radiation-noise suppression method was identified, enabling parallel processing of the radiation detection and radiation-noise suppression of video images.
2. Experiment
2.1 Camera samples
The experiments were conducted using a Sony IMX 222LQJ-C MAPS (Sony Corporation, Tokyo, Japan) [27]. The effective number of pixels was 2.43 million. A 0.18 m CMOS process with four transistors and a pinned photodiode was used to design the pixels. The pixel pitch was 2.8 µm, and the image dimensions were 6.4 mm × 5.8 mm. Analog, digital, and interface voltages of the chips were 2.7, 1.2, and 1.8 V, respectively. A sensor was integrated onto an Ambarella system sensor board (A5s ARM, Ambarella, Santa Clara, CA, USA) [27] to generate readout signals and digital signal processing. During the experiments, the aperture, shutter, gain, and white-balance control systems were manually set, and the noise reduction and exposure compensation functions were disabled. The integration time of the sensor was fixed at 40 ms.
2.2 Experimental setup
Radiation experiments were conducted to examine the response of various cameras to gamma rays. Figure 1 shows a system diagram of the experimental setup. The experiment utilized a cylindrical 60Co -ray radiation source with energy of 1.17 MeV and 1.33 MeV and activity of 3.33 × 1014 Bq. A tungsten shielding structure was used to prevent radiation damage to camera board components other than the photosensitive chip, and only the sensors were irradiated by X-rays. All experiments were performed at the China Institute of Atomic Energy, all data were transmitted to a nonradiation area via network transmission, and data were simultaneously stored and processed.
The irradiated dose rate of the camera module was altered by adjusting the distance between the sample and the radiation source. The camera samples were positioned above a slide rail. The total ionizing dose was measured with a radio chromic film dosimeter, and the dose rate was calculated by dividing the total ionizing dose by the irradiation time obtained with a measurement error of < 5%. The experimental temperature was maintained at 21 °C. The data were imported using MATLAB R2019a (MathWorks Inc., Natick, MA, USA), and they were then split into individual frames. The camera's ability to capture color images was evaluated using a video test card and a light-emitting diode.
During the dark-image radiation-response experiments, dark images were captured using a layer of opaque plastic material on the front of the sensors to insulate the sensor from contamination due to the surrounding visible light. The opaque plastic material had a thickness of 0.11 mm and size of 12.8 mm × 11.6 mm.
3. Experimental results and discussion
Figure 2 depicts a three-dimensional (3D) bar graph of the dark-image response signal in a 200 × 200-pixel area under various radiation dose-rate conditions, with the vertical axis representing the pixel value. As shown in the figure, discrete random response signals appeared in the pixel region when the radiation dose rate reached 94.99 Gy/h. The count of the response signals increased dramatically as the radiation dose rate increased, but the peak value remained unchanged. When the radiation dose rate exceeded 355.01 Gy/h, the response signal encompassed approximately all of the region's pixels when the radiation dose rate increased.
Fig. 2. Three-dimensional histogram of the response signal of 200 × 200 dark images at various dose rates.
The chance of photons hitting multiple pixels simultaneously should be low given the radiation levels under investigation. After a high-energy or low-energy photon hits a pixel, it can deposit energy in the epitaxial layer and generate electron-hole pairs. A good fraction of the charge shall be collected by this pixel, but the remaining amount of the charge will be collected by its neighboring pixels through diffusion. This explains the phenomenon of multiple pixels fired simultaneously. The gray value of a pixel is determined by the number of ions collected by the space-charge region of the photodiode [28]. This explains why the intensity dots in Fig. 2 are relatively small.
Figure 3 shows a color image (1) and its 3D bar chart (2) under irradiation at a 479.24 Gy/h irradiation dose rate. As shown in the figure, different color regions of the color image have different gray background values, and the response signal still exhibits a random peak at the background level. The response signals were randomly distributed in the color–image space plane. All color backgrounds could reflect the radiation response; however, the variance in grayscale values between regions drowned out some weak response signals. As reported in Ref. [29], areas with a high-background gray value are unsuitable for radiation-dose-rate characterization; the lower the background gray value, the more accurate the dose rate characterization. In other words, pixels with a lower background grayscale had more space to store and transfer the charges generated by the γ-ray ionizing irradiation, but the grayscale value of pixels with a higher background could easily reach saturation.
The influence of radiation-response signals on the image information of the color image's various color backgrounds is depicted in Fig. 4, which shows a color image comprising different color backgrounds and its corresponding bar chart before and during irradiation. The background color's RGB values, from left to right, are (246, 161, 8), (69, 93, 191), (244, 66, 147), (161, 11, 180), (241, 233, 96), and (253, 217, 81). As depicted in the figure, the image pixel values in each background color region before irradiation were relatively uniform and exhibited small variations. During irradiation, the pixel values in each background-color region of the image fluctuated significantly. Before and during irradiation, the fluctuations in pixel values in different background-color regions were distinct. Figure 4 (2) shows that the pixel value decreased during irradiation, near the peak area of the pixel value, indicating that the pixel value decreased significantly in the background-color area with RGB values of (246, 161, 8), (69, 93, 191), (244, 66, 147), and (246, 161, 180). In the area of the background color with RGB values of (241, 233, 96) and (253, 217, 81), the pixel value decrease significantly. For each background area, the amplitude of the increasing pixel value was greater than that of the decreasing pixel value. In the background-color areas with RGB values of (69, 93, 191) and (161, 11, 180), irradiation had a significant impact on the pixel value, which fluctuated by more than 120 before and after irradiation. In the background-color area with RGB values of (241, 233, 96) and (253, 217, 81), the pixel value was least affected by irradiation, with an amplitude of less than 50. When the RGB value was between (246, 161, 8) and (244, 66, 147), the pixel value fluctuation amplitude of the background-color area was between 75 and 100. This can help us conclude that there was a weak radiation response in images with a light background color but a significant effect in those with a dark background color. Dense irradiation-response signals can be observed in the RGB values (69, 93, 191) and background-color area (161, 11, 180) of Fig. 3(2), whereas weak response signals can be observed in the RGB values (241, 233, 96) and background-color area (253, 217, 81). Owing to the high background pixel values of the R, G, and B grayscale channels, radiation-induced pixel-value fluctuations are small, and the radiation response signal is not readily apparent in this region. Removing regions with low background pixel values that are easily perturbed by background noise and regions with high background pixel values can improve the accuracy of the radiation dose-rate representation.
Fig. 4. Histogram illustrating the effect of radiation response signals on the image data of various color backgrounds in color images.
A 2D four-band filter bank for subband image coding is shown in Fig. 5, the separable filters are first applied in one dimension (e.g., vertically) and then in the other (e.g., horizontally). Moreover, downsampling is performed in two stages: first, before the second filtering operation to reduce the overall number of computations. The resulting filtered outputs, denoted dV(m, n), dH(m, n), and dD(m, n) are called the vertical detail, horizontal detail, and diagonal detail subbands of the input image, respectively. These subbands can be split into three smaller subbands, which can be split again, and so on [30].
Figure 6 illustrates the histogram of the 2D WPD detail frame of a color image under various dose-rate irradiation conditions. As depicted in the figure, the histogram curves of the horizontal, vertical, and diagonal detail frames of the color image are correlated with the radiation dose rate within the pixel value range of 10–100. The greater the radiation dose rate, the larger the number of signals within these three details. The image information histograms of the three frame details under irradiation at various dose rates were compared. This reveals that the histogram curves of each detail are smooth between 10 and 70, and that the relationship with the dose rate is evident. The best detail is the diagonal detail.
Figure 7 depicts the statistical results of the data with pixel values ranging from 10 to 100 shown in the histogram in Fig. 6, which shows the linear relationship between the three details of the 2D WPD of the color image and the radiation dose rate under varying dose-rate irradiations. As shown in Fig. 7, the information content of each detail increased linearly as the radiation dose rate increased. The diagonal detail contains the least amount of radiation-response information. When the dose rate was below 250 Gy/h, the horizontal detail contained more information than the vertical detail. When the dose rate exceeded 250 Gy/h, the horizontal detail contained less information than the vertical detail. As shown in Table 1, the linear fitting parameters of various color-image details have distinct slopes on the radiation dose-rate representation curve. Nonetheless, the linear relationship between the diagonal detail frame and the change in the radiation dose rate had the highest linearity of 0.99641%. Therefore, the most accurate representation of the radiation dose-rate change is the diagonal detail.
Fig. 6. Histograms of high-frequency details decomposed by a 2D wavelet packet of color images irradiated at various dose rates.
Fig. 7. Linear relationship diagram of the radiation dose rate represented by a frame detail decomposed by a 2D wavelet packet of color images under various dose-rate irradiation conditions.
Table 1. Linear fitting parameters for various color image components.
As mentioned in Fig. 5, the separable filters of 2D wavelet packet processing are related to the row and column of images. We want to know whether the dose rate measurement will be affected by rotating images. Figure 8 and 9 depict the histograms of the diagonal detail of the color images with different rotation angles and dose-rate radiation conditions, as well as their corresponding linear-representation relationship graphs. As depicted in Fig. 8, when the rotation angle was set to 0°, there was a clear correlation between the histogram curve and dose rate. When the shooting angle of the MAPS camera was altered, the captured image was rotated, the smoothness of the diagonal detail frame histogram curve in the region with gray values greater than 50 and the gradient relationship with the dose rate deteriorated. Upon image rotation, the number of pixels decreased if it was initially greater than 50. This implies that the information of the diagonal detail obtained by the 2D WPD varies as the angle of the image changes. Combined with the image characteristics depicted in Fig. 3 (1), this may be associated with the direction of the contour of the image formed by the different colors. Nonetheless, as shown by the linear relationship in Fig. 9, the image information content of the diagonal detail has a good linear relationship with the radiation dose rate at any rotation angle. When the angle of rotation was set to 45°, the diagonal detail contained the most image data. The diagonal detail had the lowest value when the angle of rotation was zero. Table 2 outlines the linear fitting parameters for the image data of the diagonal details at various rotation angles. The linearity of each fitted curve was greater than 0.989, indicating that the image information of the diagonal detail exhibited a linear relationship with the radiation dose rate at each rotation angle. The rotation angle has a minimal effect on this linear relationship.
Fig. 8. Histogram of the diagonal detail frame of a color image captured by changing the shooting angle ofan MAPS camera under different dose-rate radiation conditions.
Fig. 9. Linear relationship between the diagonal detail of the image at various rotation angles and the dose rate.
Table 2. Linear fitting parameters for diagonal detail image data under varying rotation angles.
It can be concluded that for any image, the change in the dose rate can be directly characterized by extracting the response signal of the diagonal detail frame obtained by 2D WPD at any rotation angle.
4. Simulation and test
According to the preceding analysis, the response signal generated by the ionizing radiation on the image primarily consists of a forward superposition signal. Therefore, a simulation image in a real radiation environment can be obtained by superimposing the real image and the image of the radiation-response signal by subtracting the background value. Figure 10 shows a schematic of the image synthesis. Pixels with a value less than 15 were removed from the radiation-response image, and the simulation image could be obtained by overlaying the processed radiation-response image on a real image. The simulation image can be used to simulate the image in a real radiation environment, which is sufficient for the research of radiation-noise-suppression algorithms.
Figure 11 shows the pixel’s gray distribution of various image details under irradiation at different dose rates. Figure 11 (1), 11 (2), and 11 (3) depicts the horizontal, vertical, and diagonal high-frequency details of 2D WPD, respectively. The histogram of each detail of the simulation image is extremely sensitive to changes in the dose rate, and the rule is quite clear. The number of pixels increases as the dose rate increases for the horizontal, vertical, and diagonal details of the histogram. However, in the region with pixel values greater than 180, the smoothness of the histogram curve of these details deteriorated, whereas the count in the region with pixel values below 25 was not significantly affected by the dose rate. Therefore, the statistical data in regions with pixel values below 25 and over 180 are unsuitable for describing dose rate changes.
Fig. 11. Pixel’s gray distribution of various image details under varying dose-rate irradiation conditions.
Figure 12 depicts the statistical results of the data with pixel values ranging from 25 to 180 shown in the histogram of the horizontal, vertical, and diagonal details in Fig. 11, which illustrates the linear relationship between the frame detail and the radiation dose rate in the simulation images. The data for each detail indicated that the number of response signals increased linearly with the radiation dose rate. Combining the parameters of the fitted curves in Table 3, it is evident that the linearity of the diagonal detail was the best, with R2 = 0.998. This conclusion is consistent with what has been stated thus far, and information on the diagonal high-frequency detail can accurately reflect the variation in the radiation dose rate.
Fig. 12. Linear relationship between the frame detail and the radiation dose rate in the simulation images.
Table 3. Linear fitting parameters for diagonal-detail image data under varying rotation angles.
The peak signal-to-noise ratio (PSNR) is the most prevalent evaluation metric for image denoising. The better the denoising effect, the greater the PSNR value. The formula for this calculation is as follows:
where, MSE is the mean square error, which is the average of the square error of the gray value of each pixel. Its calculation formula is as follows:In this study, six noise-reduction techniques were used to suppress radiation noise, and the PSNR of each color channel was calculated and compared under various irradiation doses. Using median filtering, Gaussian low-pass filtering, and Gaussian fuzzy filtering, as well as two noise-reduction techniques based on the characteristics of radiation noise. Radiated noise in video images is distinguished by the fact that it appears infrequently in multiple consecutive frames at the same pixel position. Figure 13 shows the logic block diagrams of the two proposed noise-suppression algorithms based on this characteristic.
Fig. 13. Schematic block diagrams of two noise-reduction algorithms at the threshold and noise replacement.
Threshold denoising with median filtering: First, the image from frame i to frame i + j was collected for noise suppression through threshold denoising combined with median filtering. The upper limits of the filtering threshold Tth_1 and lower threshold Tth_2 were set according to the noise characteristics. When a certain pixel value Fi(m,n) is greater than the upper threshold Tth_1 or less than the lower threshold Tth_2, it is replaced with the pixel value from the adjacent frame Fi ± j(m,n). However, when the pixel value of pixel Fi(m,n) is less than the upper threshold Tth_1 and greater than the lower threshold Tth_2, the original pixel value is maintained. The new denoised image matrix Gi(m,n) was obtained by replacing the pixel values. The denoised image was then obtained by applying median filtering.
Noise-signal substitution with median filtering: The image from frame i to frame i + n is first extracted. Each image is then converted into an R-, G-, and B-channel gray matrix. The gray matrices of each color channel in N frames were compared, and a new noise-reduction image matrix S(m,n) was obtained by selecting the minimum pixel value at each position. Subsequently, median filtering was performed to obtain a denoised image.
Figure 14 depicts the PSNR contrast scatterplots for each color channel of the simulation image after various noise-reduction processes under a stable light source. As shown in Fig. 14, each noise-reduction algorithm can enhance the PSNR of each color channel of a video image. However, the Gaussian smoothing filter had the least impact on the image quality, whereas the noise-signal substitution algorithm had the greatest impact on noise suppression in the green and blue channels. When the noise-signal substitution algorithm and median filter were employed in succession, the PSNR of the red channel was significantly enhanced.
Fig. 14. Scatter plot of peak signal ratio of each color channel of a simulation image processed by different noise-suppression algorithms.
Figure 15 depicts the optimized result of the simulation image processed using the noise suppression algorithm depicted in Fig. 14. As shown in Fig. 14, the noise signal substitution algorithm can achieve an effect similar to that of the non-irradiated video image in Fig. 9. The three noise-reduction algorithms with the worst effects are median filtering, Gaussian low-pass filtering, and Gaussian fuzzy filtering. Figure 15 (1) depicts the denoised image produced by the noise-signal substitution algorithm, whereas Fig. 15 (2) depicts the denoised image produced by the noise-signal substitution algorithm with a median filter. Comparing the two figures, it is clear that the noise-suppression method employing the noise-signal-substitution algorithm first, followed by median filtering can achieve the best results. Under the same radiation dose-rate irradiation condition, the probability that a single pixel in the video-color image data stream will be affected by radiation in every frame is extremely low. The adjacent multiframe image data were then analyzed, and the gray value of each pixel in the adjacent multiframe image was compared. A pixel whose image value deviates from the average image value is affected by radiation noise and should be replaced with the normal image data from the adjacent frame. The core of the noise-signal substitution algorithm is the assignment of the average image value to radiation-affected pixels. After the noise-signal substitution algorithm, median filtering is the most effective method for enhancing image quality.
Fig. 15. Optimization effect of simulation-image processing using various noise-suppression algorithms.
Through 2D WPD, the count of pixels from 10 to 70 in the diagonal detail frame can be used to extract the radiation information. Using a noise-signal-substitution algorithm in conjunction with median filtering, noise suppression can simultaneously remove irradiation noise from video images more effectively. A logical flowchart of the video-image-processing method was established to achieve parallel processing of radiation detection and radiation noise suppression. Figure 16 shows a logic diagram of the proposed method. Initially, the video images captured by the MAPS camera at various radiation dose rates were decomposed using a 2D wavelet packet, and diagonal details were selected. The total number of pixels in the diagonal detail with image values between 10 and 70 was determined, and a linear relationship between the statistical values and radiation dose rates was fitted. Simultaneously, the noise-signal substitution and median-filter denoising algorithms were utilized to reduce the noise in the video images. Thus, we can obtain a clear image of the radiation site along with information regarding the radiation dose rate.
Fig. 16. Parallel-processing logic block diagram of radiation-response signal extraction and radiation-noise suppression.
5. Conclusion
Based on an analysis of the radiation response characteristics of the MAPS camera, 2D WPD was used in this study to investigate the linear response laws of various details to the variation in radiation dose rate. Simultaneously, the noise-signal substitution and median-filter denoising algorithms were used to reduce the noise in the video images captured at the radiation site. This study demonstrated that the diagonal detail can accurately represent the radiation dose rate. The informational characteristics of image texture have little impact on the characterization of radiation dose-rate changes. A noise-reduction algorithm that employs the noise-signal substitution algorithm in conjunction with median filtering can effectively remove radiation noise in video images, and it possesses simple logic, high computational efficiency, and excellent robustness. On the video images captured by MAPS cameras in a radiation environment, 2D WPD was performed, the diagonal detail was selected, and the total number of pixels from 10 to 70 was counted to characterize the radiation field dose rate. Simultaneously, a noise-signal substitution algorithm combined with median filtering was used to reduce the noise in the collected video images under a stable light source. Based on this, the rate of radiation dose can be measured in real time, and clear images can be taken simultaneously.
Funding
National Natural Science Foundation of China (11905102); Natural Science Foundation of Hunan Province (2020JJ5499).
Acknowledgments
The authors would like to express their sincere gratitude to the China Institute of Atomic Energy for providing the 60Co γ source and nuclear radiation detector.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. Y. Zhong, Q. Huang, X. Gong, Q. Tan, P. Yu, G. Yang, P. Zheng, L. Yin, T. Yang, Z. Wang, and M. Xu, “Electron cyclotron current drive under neutral beam injection on HL-2M,” Nucl. Fusion 62(12), 126027 (2022). [CrossRef]
2. A. Drăgulinescu, “Applications of CMOS image sensors: state-of-the-art,” Advanced Topics in Optoelectronics, Microelectronics, and Nanotechnologies 8411, 331–340 (2012). [CrossRef]
3. Q. Chen, CMOS-based Gamma and X-Ray Radiation Detection (Henan Normal University, 2018).
4. Q. Wei, Z. Wang, T. Dai, and G. Yu, “Nuclear radiation detection based on un-covered CMOS camera under StaticScene,” Atomic Energy Science and Technology 51(1), 175–179 (2017). [CrossRef]
5. A. Aguilar-Arevalo, X. Bertou, C. Canet, M. A. Cruz-Pérez, A. Deisting, A. Dias, J. C. D’Olivo, F. Favela-Pérez, E. A. Garcés, A. González Muñoz, J. O. Guerra-Pulido, J. Mancera-Alejandrez, D. J. Marín-Lámbarri, M. Martinez Montero, J. Monroe, S. Paling, S. J. M. Peeters, P. Scovell, C. Türkoğlu, E. Vázquez-Jáuregui, and J. Walding, “Dosimetry and calorimetry performance of a scientific CMOS camera for environmental monitoring,” Sensors 20(20), 5746 (2020). [CrossRef]
6. H. Wang, J. Feng, and Y. Li, “Noise of CIS effect on imaging resolution of camera under high radiation environment,” Atomic Energy Science and Technology 56(4), 775 (2022).
7. M. S. Haro, F. A. Bessia, M. Perez, J. J. Blostein, D. F. Balmaceda, M. G. Berisso, and J. Lipovetzky, “Soft X-rays spectroscopy with a commercial CMOS image sensor at room temperature,” Radiat. Phys. Chem. 167, 108354 (2020). [CrossRef]
8. S. Xu, Z. Qin, Y. Han, H. Dong, Y. Huang, F. Zhao, Y. Zou, Y. Liu, D. Gong, L. Xu, S. Zou, X. Yu, and H. Yang, “Ultrawide-range radiation detection based on dynamic identification and analysis of the response of a monolithic active pixel sensor,” Opt. Express 30(9), 14134–14145 (2022). [CrossRef]
9. S. Xu, S. Zou, Y. Huang, Y. Kuang, and Z. Guo, “Analysis of photon radiation response characteristics of 4T-PPD-APS,” Acta Opt. Sin. 38(5), 0504002 (2018). [CrossRef]
10. L. Servoli, F. Baldaccini, M. Biasini, B. Checcucci, S. Chiocchini, R. Cicioni, E. Conti, R. Di Lorenzo, A.C. Dipilato, A. Esposito, L. Fanó, M. Paolucci, D. Passeri, A. Pentiricci, and P. Placidi, “Active pixel as dosimetric device for interventional radiology,” Nucl. Inst Methods Phys. Res. A 720(1), 26–30 (2013). [CrossRef]
11. D. Magalotti, P. Placidi, M. Paolucci, A. Scorzoni, and L. Servoli, “Experimental characterization of a wireless personal sensor node for the dosimetry during interventional radiology procedures,” IEEE Trans. Instrum. Meas. 65(5), 1070–1078 (2016). [CrossRef]
12. D. Magalotti, P. Placidi, M. Dionigi, A. Scorzoni, and L. Servoli, “Experimental characterization of a personal wireless sensor network for the medical X-ray dosimetry,” IEEE Trans. Instrum. Meas. 65(9), 2002–2011 (2016). [CrossRef]
13. M. Fu, Study on Key Parameters of CMOS MAPS for Charged Particle Detection (Dalian University of Technology, 2012).
14. L. Y. Pang, “Can smartphones measure radiation exposures?” Pacific Historical Review 8(1), 387–388 (2015). [CrossRef]
15. S. Xu, S. Zou, Y. Huang, and L. Song, “Availability evaluation and improvement plan of image sensors in accident of nuclear power plant,” Nucl. Electron. Detect. Technol. 36(6), 651–655 (2016).
16. F. Wang, M. Wang, C. Ma, J. Zhou, R. Zhang, and Y. Liu, “Direct application of CMOS in radiation distribution and exposure rate detection,” Nucl. Electron. Detect. Technol. 10, 1278–1281 (2013).
17. F. Wang, M. Wang, Y. Liu, C. Ma, and L. Chang, “Obtaining low energy γ dose with CMOS sensors,” Nucl. Sci. Tech. 25(6), 60401 (2014). [CrossRef]
18. F. J. Spang, I. Rosenberg, E. Hedin, and G. Royle, “Photon small-field measurements with a CMOS active pixel sensor,” Phys. Med. Biol. 60(11), 4383–4398 (2015). [CrossRef]
19. Z. Wang, W. Ma, S. Huang, Z. Yao, M. Liu, B. He, J. Liu, J. Sheng, and Y. Xue, “Characterization of total ionizing dose damage in COTS pinned photodiode CMOS image sensors,” AIP Adv. 6(3), 035205 (2016). [CrossRef]
20. L. M. Hirvonen and K. Suhling, “Photon counting imaging with an electron-bombarded pixel image sensor,” Sensors 16(5), 617 (2016). [CrossRef]
21. S. Xu, S. Zou, Y. Han, and Y. Qu, “Study on the availability of 4T-APS as a video monitor and radiation detector in nuclear accidents,” Sustainability 10(7), 2172 (2018). [CrossRef]
22. M. Pérez, J. Lipovetzky, M. S. Haro, I. Sidelnik, J. J. Blostein, F. A. Bessia, and M. G. Berisso, “Particle detection and classification using commercial off the shelf CMOS image sensors,” Nucl. Instrum. Methods Phys. Res. 827, 171–180 (2016). [CrossRef]
23. H. Wang, R. Sang, H. Zhang, and X. Xie, “A new image denoising method for monitoring in intense radioactive environment,” Transducer Microsyst. Technol. 30, 59–61 (2011).
24. B. Yang, L. Zhao, and Q. Deng, “A novel anti-nuclear radiation image restoration algorithm based on inpainting technology,” J. Nanhua Univ. Sci. Technol. 30(4), 56–61 (2016).
25. M. Chen, Research on Nuclear Radiation Contaminated Image Enhancement Based on Total Variation and Sparsity Representation, (Southwest University of Science and Technology, 2020).
26. X. Sun, H. Luo, G. Liu, C. Chen, and F. Xu, “Lightweight image restoration network for strong noise removal in nuclear radiation scenes,” Sensors 21(5), 1810 (2021). [CrossRef]
27. IMX222LQJ Datasheet, SONY. Available online:https://html.alldatasheetcn.com/html-pdf/1132800/SONY/IMX222LQJ/233/2/IMX222LQJ.html (accessed on 10 February 2022).
28. S. Xu, S. Zou, and Y. Huang, “γ-ray Detection using commercial off-the-shelf CMOS and CCD images sensors,” IEEE Sens. J. 17(20), 6599–6604 (2017). [CrossRef]
29. Z. Qin, S. Xu, H. Dong, and Y. Han, “Research on calculation method of radiation response eigenvalue of a single-chip active pixel sensor,” Sensors 22(13), 4815 (2022). [CrossRef]
30. P. Habercker, Digital Image Processing (Springer Netherlands, 1983), p. 495.
