Open Access
Add to BibSonomy
Abstract
As one of the simplest methods to construct snapshot spectral imagers, multispectral filter array (MSFA) has been applied to commercial miniatured spectral imagers. While most of them have fixed configurations of spectral channels, lacking flexibility and replaceability. Moreover, conventional MSFA only comprises filtering channels but lacks the panchromatic channel which is essential in detecting dim and indistinct objects. Here, we propose a modular assembly method for snapshot imager which can simultaneously acquire the object’s multispectral and panchromatic information based on a customized filter array. The multispectral-panchromatic filter array is batch fabricated and integrated with the imaging senor through a modular mode. Five-band spectral images and a broadband intensity image can be efficiently acquired in a single snapshot photographing. The efficacy and accuracy of the imager are experimentally verified in imaging and spectral measurements. Owing to the modular architecture, our proposed assembly method owns the advantages of compactness, simple assembling, rapid replacement, and customized designing, which overcomes the expensiveness and complexity of scientific-level snapshot spectral imaging systems.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Spectral imaging has served as one of the most powerful and precise analytic tools for a variety of applications, such as scientific research [1,2], remote sensing [3,4], and medical diagnosis [5–7]. Apart from the benchtop scientific-level spectral imaging instruments which offer precise resolution and analysis, the demands for mobile systems and consumer applications have triggered the continuous miniaturization of spectral imagers [8]. Among the various miniatured spectrometers or spectral imagers, the techniques for light splitting mainly involve dispersive optics, narrowband filters, Fourier transformation, and computing reconstruction [9]. As one of the most common and simplest methods to construct snapshot spectral imagers, multispectral filter array (MSFA) has been proved to be a powerful and excellent strategy [10]. Over the past few decades, several snapshot spectral imagers based on pixel-level or tiled MSFAs have become commercially available, represented by the IMEC [11,12]. These miniatured imagers not only own small volumes but also possess the capacity of rapid and dynamic detection owning to their snapshot mechanism, opening the new fields to portable, integrated, and handheld applications.
For specific applications, the required number of spectral bands, parameters of filtering channels, and dimensions of MSFA chips are unique and diverse. Especially for multispectral imaging, which is compromising but effective and cost-efficient, the selection of spectral bands is extremely important. However, the monolithically integrated MSFAs have fixed the configuration and parameters of spectral channels, lacking the flexibility considering various applications. To overcome this, innovative approaches have emerged for the high-efficiency fabrication of MSFAs in recent years, such as gray-scale lithography [10] and stencil deposition [13]. In order to address the customizable wafer-level fabrication of MSFA, we have proposed a manufacturing method via the optimization of photolithography and deposition processes [14]. Different from the conventional etching techniques [15–18] which belong to subtractive manufacturing, our method utilizes personalized film deposition to define both the position and thickness of different layers.
As for the assembly of MSFA with the imaging sensor, precisely integrating the MSFA to the imaging sensor [19–22] or directly depositing filtering structures onto the pixels [11,23] is complicated and only suitable for the large-scale industrial production. Therefore, developing a simple mode for modular and instant assembly and replacement is of significance for wider applications. In our previous work [14], we developed a compact assembling method and verified its effectiveness. Whereas the imaging system was a preliminary prototype which was lack of the microlens array to realize snapshot imaging [24,25]. Moreover, in the majority of presently reported MSFA-based spectral imagers, all the channels are for spectral filtering and lack the panchromatic channel to acquire the object’s original intensity information. In some cases, such as detecting dim and indistinct targets with no obvious spectral characteristics, the broadband information with a higher signal-to-noise ratio becomes particularly valuable.
In this work, we propose a snapshot imaging method which can simultaneously acquire the object’s multispectral and panchromatic information. The multispectral-panchromatic imager (MSPI) is composed of a filtering & imaging module and a microlens module, which can be simply and compactly assembled. The former module is based on a customized multispectral-panchromatic filter array (MSPFA) which owns both the spectral and all-band-pass channels in one filtering chip, fabricated by combining the photolithography and electron beam (E-beam) evaporation in wafer level. Through experimental verification, the MSPI has proved to be effective and accurate in imaging and spectral measurement. The modular architecture overcomes the expensiveness and complexity of scientific-level snapshot spectral imaging systems, which provides an alternative method in simple, cheap, and flexible manufacturing and assembly, opening a new route for rapid target identification, portable devices, and cost-effective optoelectronic applications.
2. Results and discussions
2.1 Principle of multispectral-panchromatic imager (MSPI)
Our proposed modular MSPI, schematically shown in Fig. 1(a), comprises three core functioning components—a microlens array (MLA), a MSPFA, and a CMOS imaging sensor. The MLA replicates the same scene and relays the imaging object to the filter array. In order to simultaneously acquire both the multispectral and panchromatic information, all the spectral channels and an all-band-pass channel are periodically arranged in the MSPFA. Accordingly, the imaging sensor is partitioned into several square regions which are closely arranged and equal in sizes. After filtered by the MSPFA, the object can be imaged on each region of the sensor, containing both the spectral information and the original intensity information.
Fig. 1. Principle of the snapshot multispectral-panchromatic imager (MSPI). (a) Schematic diagram of the snapshot imager, which is composed of a microlens array, a multispectral-panchromatic filter array (MSPFA), and the imaging sensor. The MSPFA consists of five spectral channels and an all-band-pass channel. (b) The spectral filtering channels are composed of three layers—an Al2O3 dielectric layer sandwiched between the top and bottom Ag layers. The thicknesses of Al2O3 layers (h) range from 75 nm to 205 nm. The simulated electric-field profiles (c) and absorbed power distributions (d) for the FP filter at different wavelengths, taking the 110 nm-thick dielectric layer as an example. (e) Experimentally measured results in the CIE 1931 chromaticity coordinates for the five spectral channels. The white dashed lines with arrows represent the evolution trend for the colors when h increases.
In this work, five spectral channels plus an all-band-pass channel are defined in the MSPFA, which are arranged in a 2 × 3 configuration to form a structural cell. Each channel has a size of 600 µm × 600 µm, covering around 320 × 320 pixels. The MSPFA is matched with the upper MLA, which is an off-the-shelf product. At the bottom, a monochromatic CMOS imaging sensor (Sony IMX226) with 12 million pixels (1.85 µm × 1.85 µm) is chosen here.
As illustrated in Fig. 1(b), inside each spectral channel, it incorporates a Fabry-Perot (FP) filter that is composed of an aluminum oxide (Al2O3) dielectric layer sandwiched between the top and bottom silver (Ag) layers, all deposited on a glass substrate. The top and bottom Ag layers are 40 nm in thickness to realize a relatively narrow full width at half maximum (FWHM). The thicknesses of the middle Al2O3 layers (h) are varied from 75 nm to 205 nm to acquire the central wavelengths covering the visible band. All the filtering channels and films are fabricated by an optimized set of photolithography and E-beam evaporation. In our former research [14], the FP filters are easy to be oxidated after a period of time, while the films are also easy to be injured during the fabrication especially in ultrasonic cleaning. In order to improve the filter’s stability and reliability, as the modification to the previous process, two 5 nm Al2O3 layers are added here to act as the anti-oxidation layer (on the top) and the adhesion layer (between the bottom Ag layer and substrate), respectively (see the Supplement 1 for details).
To investigate the mechanism of the FP filters, the distribution of electric field and absorbed power are simulated by the finite-different time-domain (FDTD) method. As plotted in Fig. 1(c), taking the FP filter which is configured with 40 nm Ag/110 nm Al2O3/40 nm Ag as an example. When the wavelength of incident light (λ) varies from 400 nm to 800 nm, the electric-field distribution changes and gets a resonant peak at around 530 nm. It indicates that the standing waves are formed here due to the constructive interference of the incoming and reflected lights. As for the absorbed power (shown in Fig. 1(d)), absorption is approximately only produced at the metal/dielectric interfaces at 530 nm. The Al2O3 layer does not absorb or consume the incident power due to its dielectric merit of low loss. Furthermore, the transmissive colors of five fabricated spectral filters are converted as the discrete points in the CIE 1931 chromaticity coordinates, plotted in Fig. 1(e). The positions of colors generate a circle, demonstrating the capability of the FP filters for widely ranged color tunability.
2.2 Fabrication of multispectral-panchromatic filter array (MSPFA)
In order to realize the large-scale fabrication of multispectral filter array, we have invented a batch fabrication method [14]. However, there are few reports that add the panchromatic channel in the MSFA. Here, we propose the concept of MSPFA and fabricate it by the wafer-level processing technique.
Through iteratively performing the photolithography and E-beam evaporation, five spectral channels with different dielectric layers and the panchromatic channel with no films can be efficiently manufactured. Firstly, as illustrated in Fig. 2(a), a photolithography process was conducted to define the deposition position of the spectral channels and protect the all-band-pass channel from being filled. Then the Al2O3 adhesion layer, the bottom Ag layer, and an original 75 nm Al2O3 layer were evaporated on the glass substrate. After this, photolithography and E-beam evaporation were alternately conducted for three times to build five resonant cavities with different heights, as depicted in Figs. 2(b) to 2(d). Finally, the top 40 nm Ag film was deposited upon the five spectral channels (Fig. 2(e)), followed by an anti-oxidation layer deposition (Fig. 2(f)). As the scanning electron microscope (SEM) image of the MSPFA shown in Fig. 2(g), the all-band-pass channel is transparent with no film on, while the spectral channels are structured with varied layers and periodically arranged. Through this way, the regions, thicknesses, and materials to be deposited can be flexibly designed. Therefore, both the spectral characteristics and spatial arrangements of each filtering channel can be customized, showing the method’s advantages of feasibility and versatility.
Fig. 2. Schematic batch fabrication process for the MSPFA. (a) The adhesion layer, bottom Ag layer, and Al2O3 layer depositions by E-beam evaporation after the 1st photolithography. (b)-(d) Alternative processes of photolithography and E-beam evaporation to produce dielectric layers with a customized definition of regions, materials, and thicknesses. The asterisks represent the regions for Al2O3 evaporation. (e) The top Ag film deposition, with the numbers showing the thicknesses of Al2O3 resonant cavities (nm). (f) The Al2O3 anti-oxidation layer deposition. (g) The SEM image of the MSPFA.
Apart from the fabrication procedure of the MSPFA, another important task is to select suitable spatial arrangement of these filtering channels, in order to reduce the crosstalk and produce high-quality imaging reconstruction. After setting the spectral bands, the design of an MSFA involves two issues: the selection of tessellation mechanisms and the arrangement of different spectral bands [26–28]. As for our five-band spectral filtering and one-band panchromatic imaging mode, we chose three types of spatial arrangement according to the commonly used methods for MSFA—2n type, 3n type, and uniformly arranged. As illustrated in Fig. 3, the numbers ‘1’ to ‘5’ represent the spectral channels with the increasing central wavelength, while the number ‘6’ indicates the all-band-pass channel. We simultaneously manufactured all these three types of MSPFAs on an intact wafer.
As illustrated in Fig. 4(a), three types of MSPFAs were fabricated and photographed by a microscope both in transmission and reflection modes. In the transmission mode, the five spectral channels display different colors with high saturability, demonstrating their capabilities of frequency-selecting and filtering. Owning to the batch fabrication method, the MSPFAs with different dicing sizes and channel configurations were manufactured on a four-inch wafer, as shown in Fig. 4(b). The 2n type, 3n type, and uniformly-arranged type MSPFAs with 600 µm × 600 µm channels are illustrated in Fig. 4(c), each one tiled on an 8.2 mm × 5.4 mm chip. The transmittance of the MSPFAs were measured using a spectrometer (NOVA-EX) coupled to a microscope system (Olympus-BX53). A white light source of the wavelength from 400 nm to 800 nm was focused by an objective onto the sample surface and the transmitted light was collected by another common-path objective. As the measuring results shown in Figs. 4(d) and 4(e), the five spectral bands of the MSPFA exhibit the central wavelengths (λc) from 456 nm to 725 nm with the increasement of Al2O3 layer’s thickness, while all the FWHMs are around 30 nm to 40 nm. The transmittance of the panchromatic channel is flat across the whole visible band, which can retain the imaging object’s complete intensity information.
Fig. 4. Fabrication and measurement of the MSPFA. (a) Transmission and reflection microscopy images of the MSPFA. (b) Different types of MSPFA were batch fabricated on a four-inch wafer. (c) The MSPFA chips after dicing. (d) Measured transmission spectra of the six channels, with the transmittance and FWHM shown in (e).
2.3 Integration and assembly of MSPI
In order to compactly integrate the MSPFA with the imaging sensor, the MSPFA chip was inserted into a customized holder and then assembled upon the protective glass of the imaging sensor to constitute the filtering & imaging module, as illustrated in Fig. 5(a). The chip was mounted upside down to minimize the effect of substrate thickness. As shown in Fig. 5(b), the MLA was also mounted in a customized holder and assembled into the XY two-axis positioner, forming the microlens module. Then the above two modules can be connected by standard threads with other optical components.
Fig. 5. Integration and assembly of the MSPI. (a) The MSPFA chip is compactly assembled onto the imaging sensor by a customized holder to form the filtering & imaging module. (b) The microlens array is inserted on a two-axis positioner, constructing the microlens module and connected in front of the MSPFA. (c) Optical architecture of the modular MSPI.
Figure 5(c) demonstrates the main architecture of the MSPI. A commercially available zoom objective lens is mounted in front of the adjustable optical aperture at a distance of 17.5 mm (the standard flange focal distance for C mount). Then the objective lens forms an intermediate image at the center of the optical aperture. The focal length (f) of the MLA is 15 mm with a period of 500 µm (D). The distance from the optical aperture to the MLA (objective distance u for the MLA) and the distance from the MLA to the CMOS sensor (image distance v) follow the Gaussian imaging formula. The length and period (L) of the channels in MSPFA are both 600 µm. To match the MLA with the MSPFA, the above parameters can be calculated from the following equations
Here, u is calculated to be 90 mm and v is 18 mm.
In order to shorten the system’s length and improve the imaging quality, other additional optics such as the convergent lens and collimating lens can be added in front of the MLA. Benefiting from the modular and convenient assembly method, our proposed imager owns the advantage for easy, instant, and customized replacement of the MSPFA, MLA, imaging sensor, and objective lens according to specific applications. Apart from the MSPFA and its holder, other components are all off-the-shelf optics, which provides a cheap and convenient manner for most researchers to construct their own MSPI.
2.4 Imaging verification of MSPI
In order to verify the effectiveness and performance of the developed MSPI, imaging and spectral measurement experiments were conducted. Firstly, the MSPI was used for simultaneous snapshot multispectral imaging and panchromatic imaging. As presented in Fig. 6(a), a magic cube was selected as the imaging target. Owning to the co-working by the MLA and MSPFA, the same scene is duplicated and tiled simultaneously on the imaging sensor’s focal plane. Here, we select a 7 × 7-channel region, numbered with a to g in the horizontal direction and A to G vertically. The numbers on the top left corners represent the MSPFAs’ channels, in the same order with the sub-images from the bottom to top in Fig. 2(d). The nine color blocks in the magic cube present varying intensities under different spectral channels (number 1 to 5), demonstrating that the images which are corresponding to the five spectral channels reveal the spectral characteristics of the imaging object. Meanwhile, the all-band-pass channels do not affect the object, reserving its original intensity information. Therefore, they can help to record and recognize the object’s original characteristics, especially for the objects which are indistinct or in weak brightness. Besides, the panchromatic images can be used as reference substances to compare with the spectral images. Further, five spectral channels and an all-band-pass channel are extracted from Fig. 6(a) (number Dc/d/e and Ec/d/e) and arranged according to the increasement of central wavelength, as shown in Fig. 6(b). It can be observed that, compared with the all-band-pass image, the brightness of different color blocks changes under different spectral channels. However, the images look a little fuzzy, which is caused by the relatively few pixels in each sub-image (around 320 × 320 pixels). In the following research, we will optimize the designing of the MSPFA and MLA, as well as the image processing methods, to improve the imaging quality.
Fig. 6. Snapshot multispectral-panchromatic imaging. (a) A magic cube after filtered by the MSPFA is duplicated and tiled simultaneously on the imaging sensor’s focal plane, selecting a 7 × 7-channel region here. (b) Five spectral channels and an all-band-pass channel extracted from (a). Photographing of a building in our university without (c) and with (d) the collimating lens.
As demonstrated in the Fig. 5(c), the other optical component like the collimating lens may be necessary to construct the MSPI to improve the imaging performance. Here, the comparative experiment was performed to verify the effect of the collimating lens. As depicted in Fig. 6(c), when photographing the building by our imager, the parallax is obvious between the sub-images. After putting the collimating lens in front of the MLA, the parallax tremendously decreases, as demonstrated in Fig. 6(d).
2.5 Spectral measurement of MSPI
After the imaging verification, through the white calibration, the gray levels of the objects in each sub-image can be used to characterize their spectral responses [14]. As illustrated in Fig. 7(a), five different blocks in the magic cube are chosen to demonstrate the MSPI’s capability for spectral measurement. The measuring blocks are indicated by the dashed lines and labeled with (i) and (ii), respectively. As shown in Figs. 7(b) and 7(c), which respectively corresponds to regions (i) and (ii), the results are compared with a commercial hyperspectral imaging (HSI) system (GaiaField-V10, 400∼1000 nm with the spectral resolution of 3.5 nm). The dashed lines with data markers are the fitting curves of our MSPI’s results, which present the similar trends, peaks, and peak positions as the mature HSI system.
Fig. 7. Spectral measurement by the MSPI. (a) The magic cube taking by a color camera. (b) The spectral measurement results by our MSPI (dashed lines with marks) compared with a commercial HSI system (solid lines), corresponding to (a)-i. (c) The spectral measurement results corresponding to (a)-ii. (d) Two groups of standard PANTONE color cards. Each group is the same in color but different in materials. The asterisks represent the measuring points. (e) (f) The spectra of the color cards measured by our MSPI and the HSI system.
Based on the measuring verification of targets with distinctly different spectral characteristics, the MSPI was then used to test its capability of distinguishing similar objects. As illustrated in Fig. 7(d), two groups of standard PANTONE color cards were selected as the experimental subjects. The cards in each group are identical in colors but different in materials, respectively. The color of the left group is ‘Banana’ which is numbered with PANTONE 13-0947. The bottom color card is made of cotton while the top cards are in coated paper. As for the nearly red purple cards, from the left to right (indicated by the asterisks), the color numbers of them are 18-1757 TN, 18-1754 TCX, and 18-1754 TPG, respectively. The materials of the cards are nylon, cotton, and coated paper, respectively. As illustrated in Fig. 7(e), the red solid and dashed curves are the spectral responses of the coated paper cards, measured by the commercial HSI system and our MSPI, respectively. The gray lines represent the nylon and cotton cards. As the measuring results present, compared with the nylon and cotton cards, the spectrum of the coated paper is slightly different in intensity between 450 nm and 575 nm, but nearly coincides with the other two cards in the remaining visible bands. On the other side, as shown in Fig. 7(f), the spectra of the ‘Banana’ cards exhibit minute intensity distinction from 550 nm to 650 nm and remain high similarity in other spectral range. Consequently, the above measurements demonstrate that the objects which are in the same colors but different materials may own different visible spectra, but the differences are insignificant, which is called metamerism in some studies. The results measured by our MSPI are identical to those obtained by the HSI system, proving its effectiveness and accuracy in recognizing small spectral differences. In comparison with the scanning HSI system, our proposed modular imager exhibits its superiorities in easy assembling, low cost, and snapshot working mode which avoids time-consuming scanning. Moreover, the MSPI is able to capture the panchromatic images of the object simultaneously, which seems impossible for the commercial HSI system.
3. Conclusions and outlook
In summary, we have developed a modular assembly method for a snapshot imager which can simultaneously acquire the object’s multispectral and panchromatic information based on the customized MSPFA. By iteratively utilizing the photolithography and E-beam evaporation, we have realized the batch fabrication of MSPFA chips. Through the compact integration of MSPFA with an imaging sensor and MLA with the two-axis positioner, two main functioning modules are constructed and assembled together to realize the spectral-panchromatic imaging and spectral measurement. Five-band spectral images and a broadband intensity image can be efficiently acquired in a single snapshot photographing. The spectral measurements are experimentally verified, demonstrating the efficacy and accuracy of our MSPI.
As the outlook, our modular architecture owns the advantages of compactness, simple assembling, rapid replacement, and customized design, which is prospective in mobile applications, rapid food detection, and more consumer fields. For traditional industrial manufacturing of optical instruments, it regularly employs some optics with special design, especially for the lens and optical-mechanical structures. Different from that, our MSPI mainly utilize commercially available products except for the core MSPFA chip. It provides a feasible design philosophy, that is seeking a balance between the customization and versatility through the optimized iteration. Moreover, owing to the simultaneous capturing of the broadband intensity information, our MSPI is superior than conventional spectral imagers in detecting and recognizing weak targets or the objects lacking obvious spectral characteristics. Nevertheless, it can be found that there exists obvious parallax between the sub-images, which is inevitable due to the MLA but is not certainly negative. In fact, the light-field camera works utilizing the parallax caused by MLA to record the angle information. Therefore, our MSPI is also a potential light-field camera to realize refocusing and depth estimation, which will be explored in the next-stage research.
Funding
Science, Technology and Innovation Commission of Shenzhen Municipality (JCYJ20180508151936092, YFJGJS1.0); Science and Development Program of Local Lead by Central Government, Shenzhen Science and Technology Innovation Committee (2021Szvup112); National Natural Science Foundation of China (51975483, 51735011); Natural Science Foundation of Ningbo (202003N4033); Key Research Projects of Shaanxi Province (2020ZDLGY01-03).
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.
Supplemental document
See Supplement 1 for supporting content.
References
1. A. M. Valm, S. Cohen, W. R. Legant, J. Melunis, U. Hershberg, E. Wait, A. R. Cohen, M. W. Davidson, E. Betzig, and J. Lippincott-Schwartz, “Applying systems-level spectral imaging and analysis to reveal the organelle interactome,” Nature 546(7656), 162–167 (2017). [CrossRef]
2. K. Zhou, X. Wang, X. Jing, F. Wang, Q. Zhang, F. Chen, J. Hao, C. Deng, J. Zhou, and Y. Yu, “Large-aperture, widely and linearly tunable, electromagnetically actuated MEMS Fabry-Perot filtering chips for longwave infrared spectral imaging,” Opt. Express 30(23), 42541–42552 (2022). [CrossRef]
3. E. L. Bullock, C. E. Woodcock, and P. Olofsson, “Monitoring tropical forest degradation using spectral unmixing and Landsat time series analysis,” Remote Sens. Environ. 238, 110968 (2020). [CrossRef]
4. Y. Q. Xie, C. Y. Liu, S. Liu, and X. H. Fan, “Optical design of imaging spectrometer based on linear variable filter for nighttime light remote sensing,” Sensors 21(13), 4313 (2021). [CrossRef]
5. Y. Wang, B. Soetikno, J. Furst, D. Raicu, and A. A. Fawzi, “Drusen diagnosis comparison between hyper-spectral and color retinal mages,” Biomed. Opt. Express 10(2), 914–931 (2019). [CrossRef]
6. F. Yesilkoy, E. R. Arvelo, Y. Jahani, M. K. Liu, A. Tittl, V. Cevher, Y. Kivshar, and H. Altug, “Ultrasensitive hyperspectral imaging and biodetection enabled by dielectric metasurfaces,” Nat. Photonics 13(6), 390–396 (2019). [CrossRef]
7. J. Xiong, X. S. Cai, K. Y. Cui, Y. D. Huang, J. W. Yang, H. B. Zhu, W. Z. Li, B. Hong, S. J. Rao, Z. K. Zheng, S. Xu, Y. H. He, F. Liu, X. Feng, and W. Zhang, “Dynamic brain spectrum acquired by a real-time ultraspectral imaging chip with reconfigurable metasurfaces,” Optica 9(5), 461–468 (2022). [CrossRef]
8. S. M. Park, M. Onufrak, M. M. Haque, M. C. Were, V. Naanyu, M. K. Hasan, and Y. L. Kim, “mHealth spectroscopy of blood hemoglobin with spectral super-resolution,” Optica 7(6), 563–573 (2020). [CrossRef]
9. Z. Yang, T. Albrow-Owen, W. Cai, and T. Hasan, “Miniaturization of optical spectrometers,” Science 371(6528), eabe0722 (2021). [CrossRef]
10. C. Williams, G. S. D. Gordon, T. D. Wilkinson, and S. E. Bohndiek, “Grayscale-to-color: scalable fabrication of custom multispectral filter arrays,” ACS Photonics 6(12), 3132–3141 (2019). [CrossRef]
11. B. Geelen, C. Blanch, P. Gonzalez, N. Tack, A. Lambrechts, G. von Freymann, W. V. Schoenfeld, R. C. Rumpf, and H. Helvajian, “A tiny VIS-NIR snapshot multispectral camera,” Proc. SPIE 9374, 937414 (2015). [CrossRef]
12. P. Gonzalez, K. Tack, B. Geelen, B. Masschelein, W. Charle, B. Vereecke, A. Lambrechts, M. A. Druy, and R. A. Crocombe, “A novel CMOS-compatible, monolithically integrated line-scan hyperspectral imager covering the VIS-NIR range,” Proc. SPIE 9855, 98550N (2015). [CrossRef]
13. X. Li, Z. J. Tan, and N. X. Fang, “Grayscale stencil lithography for patterning multispectral color filters,” Optica 7(9), 1154–1161 (2020). [CrossRef]
14. X. Yu, Y. Su, X. Song, F. Wang, B. Gao, and Y. Yu, “Batch fabrication and compact integration of customized multispectral filter arrays towards snapshot imaging,” Opt. Express 29(19), 30655–30665 (2021). [CrossRef]
15. S. Wang, X. Chen, W. Lu, L. Wang, Y. Wu, and Z. Wang, “Integrated optical filter arrays fabricated by using the combinatorial etching technique,” Opt. Lett. 31(3), 332–334 (2006). [CrossRef]
16. S. W. Wang, C. Xia, X. Chen, W. Lu, M. Li, H. Wang, W. Zheng, and T. Zhang, “Concept of a high-resolution miniature spectrometer using an integrated filter array,” Opt. Lett. 32(6), 632–634 (2007). [CrossRef]
17. Z. Y. Xuan, Y. Zhi, S. W. Wang, Y. Zhang, Y. L. Zheng, J. Y. Li, F. Liu, Y. Q. Xie, Y. W. Chen, W. Z. Shi, F. Yi, and W. Lu, “Rapid and precise wavelength determination approach based on visually patterned integrated narrow bandpass filters,” IEEE Photonics J. 11(1), 1–7 (2019). [CrossRef]
18. Z. Xuan, Q. Liu, Z. Cui, S. Huang, B. Yang, C. Li, S. Wang, and W. Lu, “On-chip short-wave infrared multispectral detector based on integrated Fabry-Perot microcavities array,” Chin. Opt. Lett. 20(6), 061302 (2022). [CrossRef]
19. X. He, Y. J. Liu, K. Ganesan, A. Ahnood, P. Beckett, F. Eftekhari, D. Smith, M. H. Uddin, E. Skafidas, A. Nirmalathas, and R. R. Unnithan, “A single sensor based multispectral imaging camera using a narrow spectral band color mosaic integrated on the monochrome CMOS image sensor,” APL Photonics 5(4), 046104 (2020). [CrossRef]
20. S. P. Burgos, S. Yokogawa, and H. A. Atwater, “Color imaging via nearest neighbor hole coupling in plasmonic color filters integrated onto a complementary metal-oxide semiconductor image sensor,” ACS Nano 7(11), 10038–10047 (2013). [CrossRef]
21. Y. Horie, S. Han, J. Lee, J. Kim, Y. Kim, A. Arbabi, C. Shin, L. Shi, E. Arbabi, S. M. Kamali, H. Lee, S. Hwang, and A. Faraon, “Visible wavelength color filters using dielectric subwavelength gratings for backside-illuminated cmos image sensor technologies,” Nano Lett. 17(5), 3159–3164 (2017). [CrossRef]
22. K. Nikolaidou, H. M. Oliveira, S. Cardoso, P. P. Freitas, V. Chu, and J. P. Conde, “Monolithic integration of multi-spectral optical interference filter array on thin film amorphous silicon photodiodes,” IEEE Sens. J. 22(6), 5636–5643 (2022). [CrossRef]
23. B. Geelen, N. Tack, and A. Lambrechts, “A compact snapshot multispectral imager with a monolithically integrated per-pixel filter mosaic,” Proc. SPIE 8974, 89740L (2014). [CrossRef]
24. J. Tanida, T. Kumagai, K. Yamada, S. Miyatake, K. Ishida, T. Morimoto, N. Kondou, D. Miyazaki, and Y. Ichioka, “Thin observation module by bound optics (TOMBO): concept and experimental verification,” Appl. Opt. 40(11), 1806–1813 (2001). [CrossRef]
25. R. Shogenji, Y. Kitamura, K. Yamada, S. Miyatake, and J. Tanida, “Multispectral imaging using compact compound optics,” Opt. Express 12(8), 1643–1655 (2004). [CrossRef]
26. L. Miao and H. Qi, “The design and evaluation of a generic method for generating mosaicked multispectral filter arrays,” IEEE Trans. on Image Process. 15(9), 2780–2791 (2006). [CrossRef]
27. R. Shrestha, J. Y. Hardeberg, and R. Khan, “Spatial arrangement of color filter array for multispectral image acquisition,” Proc. SPIE 7875, 787503 (2014). [CrossRef]
28. Y. Q. Li, A. Majumder, H. Zhang, and M. Gopi, “Optimized multi-spectral filter array based imaging of natural scenes,” Sensors 18(4), 1172 (2018). [CrossRef]
