Fig 9 - uploaded by Leonidas Spinoulas
Content may be subject to copyright.
Source publication
Abstract. Passive millimeter-wave (PMMW) imagers using a single radiometer, called single pixel imagers, employ raster scanning to produce images. A serious drawback of such a single pixel imaging system is the long acquisition time needed to produce a high-fidelity image, arising from two factors: (a) the time to scan the whole scene pixel by pixe...
Context in source publication
Context 1
... of fully scanning the extended Hadamard mask, one may sample the mask randomly or sequentially every n'th pixel in the horizontal and vertical directions. Figure 9 gives a flowchart of data acquisition and image reconstruction steps as the data acquisition proceeds. The measured data with the Hadamard matrices are in the Hadamard transform space. ...
Similar publications
Recording digital holograms without wave interference simplifies the optical systems, increases their power efficiency and avoids complicated aligning procedures. We propose and demonstrate a new technique of digital hologram acquisition without two-wave interference. Incoherent light emitted from an object propagates through a random-like coded ph...
Citations
... Systems of this sort frequently employ fast rotating disks to impart spatial modulation to the illumination 34,35 . To enhance pattern deployment from continuous rotation, much emphasis in pattern design and reconstruction has been placed on the use of cyclic matrices [36][37][38][39][40] , which allow for the deployment of a new masking pattern from a translational shift of only one pattern pixel 41 . The latest development in this approach reported a pattern projection rate of 2.4 MHz, which transferred to a non-CS recovery of a frame size of 101×103 pixels at 72 fps 22 . ...
Single-pixel imaging (SPI) has emerged as a powerful technique that uses coded wide-field illumination with sampling by a single-point detector. Most SPI systems are limited by the refresh rates of digital micromirror devices (DMDs) and time-consuming iterations in compressed-sensing (CS)-based reconstruction. Recent efforts in overcoming the speed limit in SPI, such as the use of fast-moving mechanical masks, suffer from low reconfigurability and/or reduced accuracy. To address these challenges, we develop SPI accelerated via swept aggregate patterns (SPI-ASAP) that combines a DMD with laser scanning hardware to achieve pattern projection rates of up to 14.1 MHz and tunable frame sizes of up to 101×103 pixels. Meanwhile, leveraging the structural properties of S-cyclic matrices, a lightweight CS reconstruction algorithm, fully compatible with parallel computing, is developed for real-time video streaming at 100 frames per second (fps). SPI-ASAP allows reconfigurable imaging in both transmission and reflection modes, dynamic imaging under strong ambient light, and offline ultrahigh-speed imaging at speeds of up to 12,000 fps. The authors present single-pixel imaging accelerated via swept aggregate patterns (SPI-ASAP), which combines a digital micromirror device with laser scanning for fast and reconfigurable pattern projection, and a lightweight reconstruction algorithm. They demonstrate real-time video streaming at 100 fps, and up to 12,000 fps offline.
... The Hadamard matrix is a self-adjoint, orthogonal matrix. There are various forms of the Hadamard transform for CS 45,46 . Here, we use the Paley ordered Hadamard matrix, which is defined recursively by 6 where the matrix R n is unitary, R 0 = 1 and ⊗ is the Kronecker product. ...
Compressive sensing (CS) is a sub-Nyquist sampling framework that has been employed to improve the performance of numerous imaging applications during the last 15 years. Yet, its application for large and high-resolution imaging remains challenging in terms of the computation and acquisition effort involved. Often, low-resolution imaging is sufficient for most of the considered tasks and only a fraction of cases demand high resolution, but the problem is that the user does not know in advance when high-resolution acquisition is required. To address this, we propose a multiscale progressive CS method for the high-resolution imaging. The progressive sampling refines the resolution of the image, while incorporating the already sampled low-resolution information, making the process highly efficient. Moreover, the multiscale property of the progressively sensed samples is capitalized for a fast, deep learning (DL) reconstruction, otherwise infeasible due to practical limitations of training on high-resolution images. The progressive CS and the multiscale reconstruction method are analyzed numerically and demonstrated experimentally with a single pixel camera imaging system. We demonstrate 4-megapixel size progressive compressive imaging with about half the overall number of samples, more than an order of magnitude faster reconstruction, and improved reconstruction quality compared to alternative conventional CS approaches.
... After Nyquist-Shannon sampling theorem, Donoho and Candè s proposed a new sensing modality Compressed Sensing (CS [1,2,3]), which compresses the signal being acquired at the time of sensing. Since then, CS has been used in many applications such as fast millimeter-wave wireless image transmission [4] and millimeter-wave imaging [5,6,7,8]. CS not only breaks the sampling rate limitation of the Nyquist-Shannon sampling theorem, but also alleviates the storage demand. For sparse signals, CS can accurately reconstruct the original signals with a much smaller number of samples than Nyquist-Shannon theorem. ...
28×28 binary images are used for all experiments, which contain patches and sparse points that are common in object classification of automotive sensors like millimeter-wave radars.
... With the development of optical digital modulating devices (DMDs) and millimeterwave metasurface materials, the single-pixel imaging (SPI) technique, which can be used to develop an imaging system, has attracted the attention of many scholars in recent years [11][12][13][14][15][16]. The SPI technique uses different modulated patterns to illuminate the target and recovers the target image by a computational method. ...
The single-pixel imaging (SPI) technique enables the tracking of moving targets at a high frame rate. However, when extended to the problem of multi-target tracking, there is no effective solution using SPI yet. Thus, a multi-target tracking method using windowed Fourier single-pixel imaging (WFSI) is proposed in this paper. The WFSI technique uses a series of windowed Fourier basis patterns to illuminate the target. This method can estimate the displacements of K independently moving targets by implementing 6K measurements and calculating 2K windowed Fourier coefficients, which is a measurement method with low redundancy. To enhance the capability of the proposed method, we propose a joint estimation approach for multi-target displacement, which solves the problem where different targets in close proximity cannot be distinguished. Using the independent and joint estimation approaches, multi-target tracking can be implemented with WFSI. The accuracy of the proposed multi-target tracking method is verified by numerical simulation to be less than 2 pixels. The tracking effectiveness is analyzed by a video experiment. This method provides, for the first time, an effective idea of multi-target tracking using SPI.
... One emerging image compression technique is Compressed Sensing (CS), which is an effective technique to increase the image transmission speed through exploring the sparse property of the images [14]. The CS technique can be implemented in both software and hardware, and has already been widely used in optics [15], [16], millimeter wave and terahertz imaging [2], [17], [18], [19], [20], [21], [22], as well as wireless communication [23], [24], [25], [26]. What's more, since the total number of significant sparsifying basis of the images database determines the compression ratio of the CS technique, the facial image basis has to be known to achieve the optimal CS image transmission speed. ...
In FaceID era, large number of facial images could be used to breach the FaceID system, which demands effective FaceID privacy protection of the facial images for widespread adoption of FaceID technique. In this paper, to our best knowledge, we take the first step to systematically study such important FaceID privacy issue, under the framework of Compressed Sensing (CS) for fast facial image transmission. Specifically, we develop the Face-IDentification Privacy (FaceIDP) approach to protect the facial images from being used by the adversary to breach some FaceID system. First, a Dictionary Learning neural Network (DLNet) has been developed and trained with facial images database, to learn the common dictionary basis of the facial image database. Then, the encoding coefficients of the facial images are obtained. After that, the sanitizing noise is added to the encoding coefficients, which obfuscates the FaceID feature vector that is used to identify the FaceID. We have also proved that the FaceIDP is $\varepsilon$-differentially private. More importantly, optimal noise scale parameters have been obtained via the Lagrange Multiplier (LM) method to achieve better data utility for a given privacy budget $\varepsilon$. Finally, substantial experiments have been conducted to validate the efficiency of the FaceIDP with two real-life facial image databases, i.e., the LFW (Labeled Faces in the Wild) database and the PubFig database, and the results show that it outperforms other commonly used Differential Privacy (DP) approaches.
... it. An alternative way is to use a structural matrix (e.g., DCT or Hadamard matrix [68], [69]) with some permutations [70] or some designed matrices [71]. This will help the computation since fast transformations can be used. ...
Capturing high-dimensional (HD) data is a long-term challenge in signal processing and related fields. Snapshot compressive imaging (SCI) uses a two-dimensional (2D) detector to capture HD ($\ge3$D) data in a {\em snapshot} measurement. Via novel optical designs, the 2D detector samples the HD data in a {\em compressive} manner; following this, algorithms are employed to reconstruct the desired HD data-cube. SCI has been used in hyperspectral imaging, video, holography, tomography, focal depth imaging, polarization imaging, microscopy, \etc.~Though the hardware has been investigated for more than a decade, the theoretical guarantees have only recently been derived. Inspired by deep learning, various deep neural networks have also been developed to reconstruct the HD data-cube in spectral SCI and video SCI. This article reviews recent advances in SCI hardware, theory and algorithms, including both optimization-based and deep-learning-based algorithms. Diverse applications and the outlook of SCI are also discussed.
... For data acquisition, DMD 1 displays complete sets of binary masking patterns generated by a cyclic S-matrix [35,36], which is known for enhancing the signal-to-noise ratio in SPI reconstruction [37,38]. Generated by the twin-prime construction [39][40][41], the cyclic S-matrix of size MN × MN can be obtained for each pair of twin primes M and N = M + 2. Denoting the elements of such a matrix by S = [s ji ], where j, i = 0, . . . ...
We report camera-free three-dimensional (3D) dual photography. Inspired by the linkage between fringe projection profilometry (FPP) and dual photography, we propose to implement coordinate mapping to simultaneously sense the direct component of the light transport matrix and the surface profiles of 3D objects. By exploiting Helmholtz reciprocity, dual photography and scene relighting can thus be performed on 3D images. To verify the proposed imaging method, we have developed a single-pixel imaging system based on two digital micromirror devices (DMDs). Binary cyclic S-matrix patterns and binary sinusoidal fringe patterns are loaded on each DMD for scene encoding and virtual fringe projection, respectively. Using this system, we have demonstrated viewing and relighting 3D images at user-selectable perspectives. Our work extends the conceptual scope and the imaging capability of dual photography.
... In addition, Compressed Sensing (CS) is another effective way to increase the data rate by up to 10 folds [12]- [16]. Compared to the software compression, CS measurement matrix operations can be implemented on the hardware level through massive MIMO antenna arrays, which mean faster speed. ...
... The working principle and the functional blocks of the basic CS image transmission with the Hadamard mask [12], [14]- [16] as the measurement matrix is shown on the top of Fig. 1: ...
Artificial Intelligence (AI) has been used to jointly optimize a mmWave Compressed Sensing (CS) for high-speed 5G/6G image transmission. Specifically, we have developed a Dictionary Learning Compressed Sensing neural Network (DL-CSNet) to realize three key functionalities: 1) to learn the dictionary basis of the images for transmission; 2) to optimize the Hadamard measurement matrix; and 3) to reconstruct the lossless images with the learned dictionary basis. A 94-GHz prototype has been built and up to one order of image transmission speed increase has been realized for letters ``A" to ``Z".
... Additionally, aperture synthesis techniques can substantially reduce the required number of array elements and associated hardware complexity [19]. In this work, we propose further reduction of the number of signal chains through a procedure fully analogous to the computational methods pioneered in coded aperture imaging schemes [28]. The coded aperture passive imaging concept is illustrated in Fig. 1(b). ...
Passive microwave imaging of incoherent sources is often approached in a lensless configuration through array-based interferometric processing. We present an alternative route in the form of a coded aperture realized using a dynamic metasurface. We demonstrate that this device can achieve an estimate of the spectral source distribution from a series of single-port spectral magnitude measurements and complex characterization of the modulation patterns. The image estimation problem is formulated in this case as compressive inversion of a set of standard linear matrix equations. In addition, we demonstrate that a dispersive metasurface design can achieve spectral encoding directly, offering the potential for spectral imaging from frequency-integrated, multiplexed measurements. The microwave dynamic metasurface aperture as an encoding structure is shown to comprise a substantially simplified hardware architecture than that employed in common passive microwave imaging systems. Our proposed technique can facilitate large scale microwave imaging applications that exploit pervasive ambient sources, while similar principles can readily be applied at terahertz, infrared, and optical frequencies.
... In [16,17], the CS was used to get super-resolved images. CS has received great appreciation for its usage in the advanced fields such as terrestrial applications [18], passive radar perception [19], and circular SAR applications [20]. Furthermore, some military GB-SAR implementations such as through-wall radar detection, concealed target detection, and terrestrial imaging have become emerging technologies in recent years [21][22][23][24]. ...
... Furthermore, some military GB-SAR implementations such as through-wall radar detection, concealed target detection, and terrestrial imaging have become emerging technologies in recent years [21][22][23][24]. However, the studies presented elsewhere [16][17][18][19][20][21][22][23][24] were implemented by choosing random samples from whole spatial-frequency SAR data. Since these types of sampling strategies require many sampling probes that make the application impractical, alternative sampling strategies have been proposed recently [25]. ...
Millimeter-wave synthetic aperture radar (SAR) implementations need many sampling points that means huge data
collection and long processing time. Owing to compressed sensing, SAR data can be easily reconstructed with fewer random
samples than Shannon/Nyquist criteria. The most used sampling strategy is to select samples from all spatial-frequency SAR
data. Since this type of sampling strategy requires many sampling probes that makes application impractical, sparse aperture
sampling strategies have been proposed recently. However, a high noise ratio in spatial-sampled (SS) images is still a disturbing
problem to be solved. In this study, a novel Hue-domain filtering technique (HFT) is proposed to remove unwanted noises from
SS-SAR images. For this purpose, reconstructed SAR images are transformed to the red, green, and blue, and hue, saturation,
and value formats, respectively. Then the best threshold ratio to distinguish the noises from the targets is determined by means
of the Otsu method. Thus, the mask containing only the target information is formed and applied to the SAR image. Thanks to
the proposed technique, whole noise signs are removed without any target loss from the reconstructed image. The accuracy
and effectiveness of the proposed HFT are verified by means of both visual comparison of the results and integrated side-lobe
ratios of the real measurements.
