Photos captured in the shortwave infrared (SWIR) spectrum are interesting in military applications because they are independent of what time of day the pic- ture is captured because the sun, moon, stars and night glow illuminate the earth with short-wave infrared radiation constantly. A major problem with today’s SWIR cameras is that they are very expensive to produce and hence not broadly available either within the military or to civilians. Using a relatively new tech- nology called compressive sensing (CS), enables a new type of camera with only a single pixel sensor in the sensor (a SPC). This new type of camera only needs a fraction of measurements relative to the number of pixels to be reconstructed and reduces the cost of a short-wave infrared camera with a factor of 20. The camera uses a micromirror array (DMD) to select which mirrors (pixels) to be measured in the scene, thus creating an underdetermined linear equation system that can be solved using the techniques described in CS to reconstruct the im- age. Given the new technology, it is in the Swedish Defence Research Agency (FOI) interest to evaluate the potential of a single pixel camera. With a SPC ar- chitecture developed by FOI, the goal of this thesis was to develop methods for sampling, reconstructing images and evaluating their quality. This thesis shows that structured random matrices and fast transforms have to be used to enable high resolution images and speed up the process of reconstructing images signifi- cantly. The evaluation of the images could be done with standard measurements associated with camera evaluation and showed that the camera can reproduce high resolution images with relative high image quality in daylight.

To explore spectral opportunities in wideband regime for cognitive radio we need a wideband spectrum sensing receiver. Current wideband receiver architectures need wideband analog to digital converter (ADC) to sample wideband signal. As current state-of-art ADC has limitation in terms of power and sampling rate, we need to explore some alternative solutions. Compressive sampling (CS) data acquisition method is one of the solutions. Cognitive Radio signal, which is sparse in frequency domain can be sampled at Sub-Nyquist rate using low rate ADC. To relax the receiver complexity in terms of performance requirement we can use Modulated Wideband Converter (MWC) architecture, a Sub-Nyquist sampling method. In this thesis circuit design of this architecture covers signal within a frequency range of 500 MHz to 2.1 GHz, with a channel bandwidth of 1600 MHz. By using 8 parallel lines with channel trading factor of 11, effective sampling rate of 550 MHz is achieved for successful support recovery of multi-band input signal of size N=12.

Fast detection and identification of unknown substances is an area of interest for many parties. Raman spectroscopy is a laser-based method allowing for long range no contact investigation of substances. A Coded Aperture Snapshot Spectral Imaging (CASSI) system allows for fast and efficient measurements of hyperspectral images of a scene, containing a mixture of the spatial and spectral data. To analyze the scene and the unknown substances within it, it is required that the spectra in each spatial position are known. Utilizing the theory of compressed sensing allows for reconstruction of hyperspectral images of a scene given their CASSI measurements by assuming a sparsity prior. These reconstructions can then be utilized by a human operator to deduce and classify the unknown substances and their spatial locations in the scene. Such classifications are then applicable as decision support in various areas, for example in the judicial system. Reconstruction of hyperspectral images given CASSI-measurements is an ill-posed inverse problem typically solved by utilizing regularization techniques such as total variation (TV). These TV-based reconstruction methods are time consuming relative to the time needed to acquire the CASSI measurements, which is in the order of seconds. This leads to a reduced number of areas where the technology is applicable. In this thesis, a Generative Adversarial Network (GAN) based reconstruction method is proposed. A GAN is trained using simulated training data consisting of hyperspectral images and their respective CASSI measurements. The GAN provides a learned prior, and is used in an iterative optimization algorithm seeking to find an optimal set of latent variables such that the reconstruction error is minimized. The results of the developed GAN based reconstruction method are compared with a traditional TV method and a different machine learning based reconstruction method.  The results show that the reconstruction method developed in this thesis performs better than the compared methods in terms of reconstruction quality in short time spans.

No description available.

In large antenna systems, traditional channel estimation is costly and infeasible in some situations. Compressive sensing was proposed to estimate the channel with fewer measurements. Most of the previous work uses a predefined discrete Fourier transform matrix or overcomplete Fourier transform matrix to approximate the channel. Then, a learned dictionary trained by K-singular value decomposition (K-SVD) was proposed and was proved superiority using orthogonal matching pursuit (OMP) to reconstruct the sparse channel. However, with the development of compressive sensing, there are plenty of dictionary learning algorithms and sparse recovery algorithms. It is important to identify the effect and the performance of different algorithms when transforming the high dimensional channel vectors to low dimensional representations. In this thesis, we use a spatial channel model to generate channel vectors. Dictionaries are trained by K-SVD and method of optimal directions (MOD). Several sparse recovery algorithms are used to find the sparse approximation of the channel like OMP and gradient descent with sparsification (GraDeS). We present simulation results and discuss the performance of the various algorithms in terms of accuracy, sparsity, and complexity. We find that predefined dictionaries works with most of the algorithms in sparse recovery but learned dictionaries only work with pursuit algorithms, and only show superiority when the algorithm coincides with the algorithm in the sparse coding stage. / I stora antennsystem är traditionell kanaluppskattning kostsam och omöjlig i vissa situationer. Kompressionsavkänning föreslogs för att uppskatta kanalen med färre mätningar. Det mesta av det tidigare arbetet använder en fördefinierad diskret Fourier transformmatris eller överkompletterad Fourier -transformmatris för att approximera kanalen. Därefter föreslogs en inlärd ordbok som utbildats av K-SVD och bevisades överlägsen med hjälp av OMP för att rekonstruera den glesa kanalen. Men med utvecklingen av komprimerad avkänning finns det gott om algoritmer för inlärning av ordlistor och glesa återställningsalgoritmer. Det är viktigt att identifiera effekten och prestandan hos olika algoritmer när de högdimensionella kanalvektorerna omvandlas till lågdimensionella representationer. I denna avhandling använder vi en rumslig kanalmodell för att generera kanalvektorer. Ordböcker tränas av K-SVD och MOD. Flera glesa återställningsalgoritmer används för att hitta den glesa approximationen av kanalen som OMP och GraDeS. Vi presenterar simuleringsresultat och diskuterar prestanda för de olika algoritmerna när det gäller noggrannhet, sparsamhet och komplexitet. Vi finner att fördefinierade ordböcker fungerar med de flesta algoritmerna i gles återhämtning, men inlärda ordböcker fungerar bara med jaktalgoritmer och visar bara överlägsenhet när algoritmen sammanfaller med algoritmen i det glesa kodningsstadiet.

<p>The compressive sensing (CS) paradigm provides an efficient image acquisition technique through simultaneous sensing and compression. Since the imaging philosophy in CS imagers is different from conventional imaging systems, new physical structures are required to design cameras suitable for CS imaging.</p> <p>While this work is focused on the hardware implementation of CS encoding for CMOS sensors, the image reconstruction problem of CS is also studied. The energy compaction properties of the image in different domains are exploited to modify conventional reconstruction problems. Experimental results show that the modified methods outperform the 1-norm and TV (total variation) reconstruction algorithms by up to 2.5dB in PSNR.</p> <p>Also, we have designed, fabricated and measured the performance of two real-time and area-efficient implementations of the CS encoding for CMOS imagers. In the first implementation, the idea of active pixel sensor (APS) with an integrator and in-pixel current switches are used to develop a compact, current-mode implementation of CS encoding in analog domain. In another implementation, the conventional three-transistor APS structure and switched capacitor (SC) circuits are exploited to develop the analog, voltage-mode implementation of the CS encoding. With the analog and block-based implementation, the sensing and encoding are performed in the same time interval, thus making a real-time encoding process. The proposed structures are designed and fabricated in 130nm technology. The experimental results confirm the scalability, the functionality of the block read-out, and the validity of the design in making monotonic and appropriate CS measurements.</p> <p>This work also discusses the CS-CMOS sensors for high frame rate CS video coding. The method of multiple-camera with coded exposure video coding is discussed and a new pixel and array structure for hardware implementation of the method is presented.</p> / Doctor of Philosophy (PhD)

Exploiting signal sparsity has recently received considerable attention in a variety of areas including signal and image processing, compressive sensing, machine learning and so on. Many of these applications involve optimization models that are regularized by certain sparsity-promoting metrics. Two most popular regularizers are based on the l1 norm that approximates sparsity of vectorized signals and the total variation (TV) norm that serves as a measure of gradient sparsity of an image. Nevertheless, the l1 and TV terms are merely two representative measures of sparsity. To explore the matter of sparsity further, in this thesis we investigate relaxations of the regularizers to nonconvex terms such as lp and TVp "norms" with 0 <= p < 1. The contributions of the thesis are two-fold. First, several methods to approach globally optimal solutions of related nonconvex problems for improved signal/image reconstruction quality have been proposed. Most algorithms studied in the thesis fall into the category of iterative reweighting schemes for which nonconvex problems are reduced to a series of convex sub-problems. In this regard, the second main contribution of this thesis has to do with complexity improvement of the l1/TV-regularized methodology for which accelerated algorithms are developed. Along with these investigations, new techniques are proposed to address practical implementation issues. These include the development of an lp-related solver that is easily parallelizable, and a matrix-based analysis that facilitates implementation for TV-related optimizations. Computer simulations are presented to demonstrate merits of the proposed models and algorithms as well as their applications for solving general linear inverse problems in the area of signal and image denoising, signal sparse representation, compressive sensing, and compressive imaging. / Graduate