Abstract
Optical synthetic aperture imaging system has grown out the quest for higher angular resolution in astronomy, which combines the radiation from several small sub-apertures to obtain a resolution equivalent to that of a single filled aperture. Due to the discrete distribution of the sub-apertures, pupil function is no longer a connected domain, which further leads to the attenuation or loss of the mid-frequency modulation transfer function (MTF). The mid-frequency MTF compensation is therefore a key focus. In this paper, a complete mid-frequency compensation algorithm is proposed, which can extract and fuse the frequency of different synthetic aperture systems and monolithic aperture systems according to their special MTF characteristics. The dimensions of the monolithic aperture and optical synthetic aperture system are derived, and the longest baseline of the monolithic aperture is much smaller than that of the optical synthetic aperture system. Then the separated spatial frequency information is extracted and synthesized according to the spatial frequency equivalence point. Finally, the full-frequency enhanced image is recovered by using improved Wiener-Helstrom filter, which adopts specific parameters based on different sub-aperture arrangements. The mid-frequency MTF of Golay-3 increases from 0.12 to 0.16 and that of Golay-6 increases from 0.06 to 0.18. Both the simulation and experiment prove that the proposed method not only realizes the spatial resolution determined by the longest baseline of the optical synthetic aperture system, but also successfully compensates its mid-frequency MTF.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Due to the humans’ ambitions to explore the universe, optical synthetic aperture imaging is proposed to enhance the spatial resolution by increasing the effective aperture diameter. Meanwhile, it is also desirable to minimize the total image collection area because of the volume and weight of large monolithic mirror. The optical synthetic aperture system has successfully solved the manufacturing, installation and launching problems caused by the fully filled large aperture system [1–3]. The system consists of several in-phase circular sub-mirrors. The synthetic aperture system combines the radiation from these small sub-mirrors to realize a resolution equivalent to that of a monolithic aperture, whose area is greater than the sum of the individual sub-mirrors [3]. However, due to the discrete distribution of the sub-apertures, the pupil function is no longer a connected domain, which further leads to the attenuation or loss of the mid-frequency modulation transfer function (MTF) [4–5]. It means that the mid-band spatial frequency contains lower energy, therefore the image information at the corresponding spatial frequency will be suppressed or lost. Thus, the mid-frequency MTF compensation is a key focus in designing a good synthetic aperture system.
Fortunately, there are two main ways to compensate the mid-frequency MTF [5]. The first method is to improve mid-frequency by image restoration method. In general, a high-speed and efficient image restoration method employs deconvolution algorithms, which include Wiener-Helstrom filter, linear least-squares filter, or wavelet transformation, and so on [6–8]. The second method is to optimize the aperture configuration. By using optimization algorithm, some aperture configurations with excellent performance are obtained, which maximize the spatial cutoff frequency while simultaneously maintain a minimum MTF level over all frequencies up to the cutoff [9–11]. However, in both two methods, the decrease of the mid-frequency MTF caused by the dispersion and sparsity is still inevitable. That is, the deteriorated or lost part of the mid-frequency information cannot be recovered.
For a monolithic aperture, the MTF decreases approximately linearly before the cutoff frequency, which means that the MTF approximates 0.5 at half the cutoff frequency [12]. But for optical synthetic aperture system, the MTF quickly drops below 0.5 at low frequency and slowly decays at middle frequency until it becomes zero at cutoff frequency. Therefore, Chenghao Zhou and Zhile Wang proposed an image fusion algorithm to combine the images of the optical synthetic aperture system and the monolithic aperture system [5]. However, in their method, the images of the two systems are added directly, which leads to many special features of the system not being utilized. In this paper, based on the above analysis, a complete mid-frequency MTF compensation algorithm based on MTF characteristics of the synthetic aperture system and monolithic aperture system is proposed. The spatial frequency information according to the MTF is extracted and fused. Then the whole spatial frequency information is recovered to full-frequency enhanced image. The proposed method can not only expand the spatial frequency cutoff, but also compensate the attenuation of MTF in the middle frequency and maintain an appropriate MTF level at the full frequency. Both simulation and experiment prove that compensating the mid-frequency MTF by synthesizing two systems is effective.
The rest of paper is organized as follows. Section 2 derives the spatial frequency equivalence point and presents the process of frequency extraction and synthesis. The improved Wiener algorithm is also introduced in the section. Section 3 analyzes the performance of the proposed method. Section 4 details the experiment results. Finally, section 5 summarizes the conclusions.
2. Method
In this section, in order to compensate the mid-frequency MTF of optical synthetic aperture, another monolithic aperture system is introduced. The size of the monolithic aperture is designed and the spatial frequency equivalence point is derived. Then the spatial frequency information is extracted and synthesized. The Wiener-Helstrom filter is improved to adapt to the specific case.
2.1. MTF analysis
As the concept of the baseline of astronomical telescope, the baseline of the optical synthetic aperture system refers to the lines that can be obtained over the image collection area [13]. As shown in Fig. 1, a short baseline A1A2 corresponds to the low-frequency of the acquired information, which is usually the contour of the image. The long baseline C1C2 is responsible for the high frequency information, which is the detail of the image. The longest baseline Dspan determines the cutoff frequency, which corresponds to the resolution of the system. The cutoff frequency ρc can be expressed as:
However, due to the reduction of the image collecting area, the sampling of information at different frequencies is not completely compared with the monolithic aperture. In general, the smaller the image collecting area is, the more information will be lost. It can be quantified by the fill factor α which is the ratio of the area of the sub-apertures to the area of a single filled aperture with equivalent image resolution, given by:
The MTF is the modulus of the optical transfer function (OTF), which is defined as the normalized Fourier transform of the point spread function (PSF) intensity. The PSF intensity can be obtained by taking the magnitude squared of the amplitude point spread function (ASF). The ASF is the Fourier transform of the system’s pupil function P(x, y):
After the ASF is obtained, the PSF intensity of the optical synthetic aperture system can be calculated by:
The configurations and their corresponding PSF of two typical optical synthetic aperture systems are shown in Fig. 2.
The optical transfer function is defined as the normalized Fourier transform of the PSF intensity:
From Eq. (9), it is proved that the MTF of a fully filled circular aperture has unity modulation at zero spatial frequency, and the MTF decreases almost linearly until it reduces to zero at the cutoff frequency. Therefore, the MTF of a monolithic aperture is simplified to a linear function. When the MTF of the Golay-N system with a spatial frequency cutoff ρc1 is equal to that of the fully filled aperture system with a spatial frequency cutoff ρc2, the spatial frequency can be calculated by:
For example, the spatial frequency equivalence point of Golay-3 is calculated as:
Thus, the diameter of the monolithic aperture satisfies:
Then, Eq. (10) can be rewritten as:
Figure 4 illustrates the spatial frequency equivalence point ρeq of the Golay-3 and Golay-6.
After the spatial frequency equivalence point is obtained, the spatial frequency information of this two systems is extracted and fused to compensate the mid-frequency.
2.2. Frequency synthesis method
The combination of different object spatial frequencies information is the imaging process, which can be simultaneous or not. According to Fourier optics, for one dimensional continuous image, the imaging process of a diffraction-limited incoherent imaging system can be modeled by:
where A(ρ) denotes the amplitude of the original image and θρ denotes the phase, respectively [5]. The amplitude and phase can be obtained by Fourier transform of images taken by the system. In the same way, two dimensional images are similar to one dimensional images, which are the Fourier transform of the captured two dimensional images. This integral can be divided into two terms, given by:Therefore, the image whose whole spatial frequencies are high can be obtained by fusing the images from a monolithic aperture and optical synthetic aperture system. The fusion process can be expressed as:
Equation (17) can be realized by a low pass filter and a band pass filter. Then, the fusion process can be rewritten as:
According to the imaging model of the optical system, Eq. (18) can be summarized as: the images collected by the monolithic aperture system and the optical synthetic aperture system are first processed by Fourier transform. Then the information of separate frequencies is extracted by using filters in the frequency space, and finally the fused image information on the frequency space is obtained. The detailed process of the algorithm is shown in Fig. 5. The separate frequency information with higher MTF level is extracted and synthesized, and the result is shown in Fig. 6.
2.3. Image restoration
Post-processing is necessary to restore the synthesized images, which can make the reconstructional images sharp and improve the contrast. In theory, the post-processing method is to take the deconvolution of the image collected with the known PSF of the imaging system. In frequency space, the reciprocal of the OTF is therefore a rudimentary example of a restoration filter. However, the finite OTF size and the noise existence make the rudimentary reciprocal OTF restoration filter problematic. The improved Wiener-Helstrom filter is a typical post-processing method for optical synthetic aperture system because of its excellent performance and good anti-noise ability [15]. For the Wiener deconvolution, the transfer function of the filter can be expressed as:
where (fx, fy) represents the coordinate of the frequency space, and OTF*(fx, fy) is the complex conjugate of OTF(fx, fy), and the K value is given by: where Sn(fx, fy) and Sf(fx, fy) are the power spectrum of the noise and the original image respectively. Therefore, K value is reciprocal of the signal-to-noise ratio (SNR). Generally, the priori information about Sn(fx, fy) and Sf(fx, fy) is unknown, therefore K is generally estimated based on experience in practice.According to Eq. (18), the OTF(fx, fy) of fused image of the monolithic aperture system and the optical synthetic aperture system can be calculated as:
After the OTFFused of the system is obtained, the SNR of the system needs to be estimated, so that the Wiener filter can have a good performance. The relationship between the SNR, MTF, exposure time, and fill factor for an optical synthetic aperture system has been analyzed, which can be expressed as [4]:
Where ϕo and ϕb are the number of image photons and bias photos per unit area per unit time respectively. α is the fill factor. ηt is the total efficiency of the system, which includes system throughput and the detector quantum efficiency. T is the exposure time. M is the size of the image collected. μ(fx,fy) is the normalized scene Fourier transform. γdc and σr are the dark current and the RMS readout noise respectively.
This expression shows that SNR is proportional to MTF. Since the maximum MTF level of the sidelobes of the Golay-N array is 1/N, the SNR of the sidelobes is 1/N of the center part. Thus the K value can be rescaled by filters. The transfer function of the Wiener filter can be rewritten as:
Then K can finally get the optimal value according to peak signal-to-noise ratio [15].
3. Simulation results
In this section, the effect of the proposed method is demonstrated. The effect of the image restoration algorithm is analyzed under different SNR. And the performance of the mid-frequency MTF compensation is evaluated quantitatively.
To demonstrate the performance in real situations, where the optical synthetic aperture systems are used to capture images of celestial objects through ground-based or space telescopes, the images used in the simulation are based on Hubble space telescope [16]. A total of 60 images are used for tested. Part of the test set is shown in Fig. 7.
In simulation, the PSF of the monolithic aperture system and optical synthetic aperture system are both acquired in ZEMAX. In practice, the phase differences between the sub-mirrors can be reduced to a small fraction of the wavelength by using interferometry. Thus, the sub-mirrors are co-phased in the simulation. The Gaussian white noise is added because it is proved that the total noise of all noise mentioned in Eq. (24) can be modeled as Gaussian white noise [17,18]. But real systems will have many multiplicative noise which depends on the state of the systems. They generally do not obey the Gaussian distribution and may affect image quality. In order to evaluate the noise resistance of the proposed method, the signal to noise ratio is set to 30 dB and 40 dB respectively. The peak signal to noise ratio (PSNR) and the structural similarity index (SSIM) are selected to evaluate quantitatively, which are widely applied to the analysis of image restoration of optical synthetic aperture system [19]. The PSNR can be defined as:
The SSIM can be calculated as:
The performance is compared with the image recovery results based on single system, which is illustrated in Fig. 8 and Fig. 9. Figure 8(a) and Fig. 9(a) are the original clear images, which are taken as the ideal benchmarks. And the PSNR and SSIM of other images are calculated from these benchmarks. The degraded images of column 2 and column 3 in Fig. 8 and Fig. 9 are the results of the original images being processed by different systems. And the degraded images contain different levels of noise. In order to make the comparison fair, the optimal value of Wiener parameter K of other method is designed based on the PSNR according to the method in Ref. [15]. Although, for example, the narrow galaxies in Fig. 8(d) and Fig. 8(e) are still fuzzy. Meanwhile, the Fig. 8(f) obtained by synthesizing two systems is visually close to the original image. Besides, the maximum PSNR and SSIM values are both achieved by Fig. 8(f).
Apart from the PSNR and SSIM indices, several other matrices are adopted, which make the evaluation more comprehensive and objective. These matrices include information fidelity criterion (IFC), and visual information fidelity (VIF), weighted peak signal-to noise ratio (WPSNR), and multi-scale structure similarity index (MSSSIM) [20]. They are all commonly used in the field of image quality assessment. And the closer the restored image is to the original image, the larger the values of all these evaluation indexes are. The final results are shown in Table 2. All these qualitative evaluations get the highest value in the proposed method, which demonstrates that restoring by using two systems can successfully improve the image restoration effect of an individual system.
In order to directly analyze the effect of the mid-frequency MTF compensation of optical synthetic aperture system, an edge method proposed in Ref. [21] is applied to assert the MTF. The edge spread function of the system can be obtained from an angled edge image. The line spread function is determined by the edge spread function. Thus the MTF is estimated because it is the zero frequency normalized modulus of the Fourier Transform of the line spread function. The results are shown in Fig. 10. It can be seen that the MTF curves of the proposed method are similar to that of monolithic aperture system in the middle and low frequencies, which improves the performance of a separate synthetic aperture at the middle and low frequencies. It means that the proposed approach can better maintain the contour and texture of the image. And the cutoff frequency of the MTF curve is still the same as that of synthetic aperture. Thus the results of two systems preserve the high frequency information of the optical synthetic aperture system. Because the MTF curve of the Golay-6 decreases more rapidly than that of the Golay-3, the mid-frequency MTF compensation of the Golay-6 aperture system is more obvious.
The compensation effect of the mid-frequency MTF is quantitatively evaluated. The middle spatial frequency can be represented by a single value MTFmid freq, which is the mean MTF value over spatial frequencies from ρd to ρc, where ρd denotes the spatial cutoff frequency corresponding to individual sub-aperture [7,22]. Here, ρd is approximately 0.2. The MTFmid freq can be expressed as the following equation:
According to the Fig. 10, in the range from ρd to ρc, the ratio of the area enclosed by the MTF curve and the coordinate axis to the difference between the two frequencies is the value of MTFmid freq. It can be calculated that the MTFmid freq of Golay-3 increases from 0.12 to 0.16. And the MTFmid freq of Golay-6 increases from 0.06 to 0.18. Thus, the proposed method not only realizes the cutoff frequency determined by the longest baseline of the optical synthetic aperture system, but also successfully compensates its mid-frequency MTF.
4. Experimental results
In this section, a reflective synthetic aperture system is established to validate the proposed approach. The experimental platform is shown in Fig. 11. Different images are collected from the detector by adjusting the sources. Specifically, the helium-neon laser and the pinhole are used to obtain the PSF of the system, where the size of the pinhole is 40 µm. And the LED and the resolution board are used to get the degraded images, where the resolution board (GCG-020101), which consists of different groups of line pairs, has a measuring capacity of 200 cycles per 1 mm. Then the pellicle beam splitter is applied to control the input of different sources. The beams of different sources both pass through the collimator, which is used to produce the parallel beams. Then the parallel beams are reflected by a spherical reflector with a pupil mask. The pupil mask is designed in the form of Golay-3 and a single circle to simulate the Golay-3 synthetic aperture system and the monolithic aperture system respectively. For the pupil mask of Golay-3, the diameter of sub-aperture is 18 mm and that of the ex-circle is 35 mm. For the monolithic aperture, the diameter is 25 mm. Finally, the beams pass through another pellicle beam splitter and are captured by a CCD camera (Point Grey GS2-GE-20S4M), where the pixel size is 4.4 µm × 4.4 µm. Because the focal length of the reflector is 20 mm. Thus the sum of the distance from the reflector to the beam splitter and the beam splitter to the detector is 20 mm.
The actual degraded images and the PSF are obtained by the experiment, as shown in Fig. 12. By adjusting the pupil mask, the degraded image and the PSF of the Golay-3 synthetic aperture system and the monolithic aperture system are captured respectively. In order to better display the image changes, the collected images are partially enlarged. As shown in Fig. 12(a1) and Fig. 12(a2), the collected images are visually fuzzy and low contrast. And because the Golay-3 system's MTF decays faster at low and middle frequencies, the collected image of this system has lower contrast than that of the monolithic aperture system. The result is consistent with the simulation above. The restored result in Fig. 12(a3) is obviously sharper and has a higher contrast than these collected images. A comparison of two groups of line pairs of the resolution board is analyzed, as shown in Fig. 12(c1) and Fig. 12(c2). Here, the oblique line pairs are group1, and the vertical line pairs are group2. For the line traces of the restored result of the proposed method, the gaps between the peaks and valleys are widest, which means it has the highest contrast. Therefore, image restoration by synthesizing two systems can effectively improve the image quality.
5. Conclusion
For a single optical synthetic aperture system, the attenuation of the mid-frequency MTF caused by the dispersion and sparsity is inevitable. In this paper, another monolithic aperture system is combined to compensate the mid-frequency MTF. The dimension of the monolithic aperture is designed, which is much smaller than that of the optical synthetic aperture system as we expected. The MTF of the Golay system is analyzed and the spatial frequency equivalence point of the two systems is determined. Then the separated spatial frequency information is extracted and synthesized by using a low pass filter and a band pass filter. Finally an improved Wiener-Helstrom filter is proposed, which adopts specific parameters based on different sub-aperture arrangements. Both simulation and practical experiment are established to verify the proposed method. The final MTF approximates to that of monolithic aperture system in the middle and low frequencies, which better maintains the contour and texture of the image. Thus the proposed method improves the performance of a separate optical synthetic aperture system at the middle and low frequencies. At the same time, the MTF still preserves the cutoff frequency of the synthetic aperture. And the mid-frequency MTF of Golay-3 increases from 0.12 to 0.16 and that of Golay-6 increases from 0.06 to 0.18. Therefore, synthesizing both two systems not only realizes the spatial resolution determined by the longest baseline of the optical synthetic aperture system, but also successfully compensates its mid-frequency MTF. In addition, in order to be applied to the more complex reality, the proposed method can be further improved. The combination of MTF characteristics of non-aligned optical synthetic aperture system and that of monolithic aperture system is worth further study. We believe that the proposed mid-frequency MTF compensation method will be useful to other researchers in real-world situations.
Funding
National Natural Science Foundation of China (61475018).
Disclosures
The authors declare no conflicts of interest.
References
1. A. B. Meinel, “Aperture Synthesis Using Independent Telescopes,” Appl. Opt. 9(11), 2501 (1970). [CrossRef]
2. G. R. Lemaitre, “Optical design and active optics methods in astronomy,” Opt. Rev. 20(2), 103–117 (2013). [CrossRef]
3. J. S. Fender, “Synthetic apertures: An Overview,” Proc. SPIE 0440, 2–7 (1984). [CrossRef]
4. J. R. Fienup, “MTF and integration time versus fill factor for sparse-aperture imaging systems,” Proc. SPIE 4091, 43–47 (2000). [CrossRef]
5. C. Zhou and Z. Wang, “Mid-frequency MTF compensation of optical sparse aperture system,” Opt. Express 26(6), 6973–6992 (2018). [CrossRef]
6. J. R. Fienup and J. J. Miller, “Comparison of reconstruction algorithms for images from sparse-aperture systems,” Proc. SPIE 4792, 1–8 (2002). [CrossRef]
7. A. J. Stokes, B. D. Duncan, and M. P. Dierking, “Improving mid-frequency contrast in sparse aperture optical imaging systems based upon the Golay-9 array,” Opt. Express 18(5), 4417–4427 (2010). [CrossRef]
8. Z. Xie, H. Ma, B. Qi, G. Ren, X. He, L. Dong, and Y. Tan, “Active sparse aperture imaging using independent transmitter modulation with improved incoherent Fourier ptychographic algorithm,” Opt. Express 25(17), 20541–20556 (2017). [CrossRef]
9. L. M. Mugnier, G. Rousset, and F. Cassaing, “Aperture configuration optimality criterion for phased arrays of optical telescopes,” J. Opt. Soc. Am. A 13(12), 2367–2374 (1996). [CrossRef]
10. F. Cassaing and L. M. Mugnier, “Optimal sparse apertures for phased-array imaging,” Opt. Lett. 43(19), 4655–4658 (2018). [CrossRef]
11. S. Kiehbadroudinezhad and N. K. Noordin, “Optimization of a large correlator array configuration to maximize the coverage of υ-ν plane and minimize the sidelobes,” Astronomical Review 10(1080), 2167–2857 (2013).
12. N. Miller, B. Duncan, and M. P. Dierking, “Resolution enhanced sparse aperture imaging,” Aerospace Conference. IEEE, (2006).
13. R. C. Walker, “Very long baseline interferometry,” Radio Sci. 21(4), 665–679 (2016).
14. M. J. E. Golay, “Point arrays having compact nonredundant autocorrelations,” J. Opt. Soc. Am. 61(2), 272–273 (1971). [CrossRef]
15. H. Chen, Z. Cen, C. Wang, S. Lan, and X. Li, “Image Restoration via Improved Wiener Filter Applied to Optical Sparse Aperture Systems,” Optik 147, 350–359 (2017). [CrossRef]
16. D. Geman and C. Yang, “Nonlinear image recovery with half-quadratic regularization,” IEEE Trans. Image Process. 4(7), 932–946 (1995). [CrossRef]
17. L. Li, Y. Jiang, and C. Wang, “Noise Analysis and Image Restoration for Optical Sparse Aperture Systems”, International Workshop on Geoscience and Remote Sensing. (2008).
18. Z. Zhou, D. Wang, and Y. Wang, “Effect of noise on the performance of image restoration in an optical sparse aperture system,” J. Opt. 13(7), 075502 (2011). [CrossRef]
19. W. Zhou, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality assessment: from error visibility to structural similarity,” IEEE Trans. Image Process. 13(4), 600–612 (2004). [CrossRef]
20. C. Y. Yang, C. Ma, and M. H. Yang, “Single-image super-resolution: A benchmark,” in Proc. Eur. Conf.
21. P. B. Greer and T. V. Doorn, “Evaluation of an algorithm for the assessment of the MTF using an edge method,” Med. Phys. 27(9), 2048–2059 (2000). [CrossRef]
22. M. D. Turner, A. J. Stokes, G. W. Kamerman, B. D. Duncan, M. P. Dierking, and N. J. Miller, “Increasing mid-frequency contrast in sparse aperture optical imaging systems,” Laser Radar Technology and Applications XIV, (2009).