Thin infrared imaging systems through multichannel sampling

The size of infrared camera systems can be reduced by collecting low-resolution images in parallel with multiple narrow-aperture lenses rather than collecting a single high-resolution image with one wide-aperture lens. We describe an infrared imaging system that uses a three-by-three lenslet array with an optical system length of 2.3 mm and achieves Rayleigh criteria resolution comparable with a conventional single-lens system with an optical system length of 26 mm. The high-resolution final image generated by this system is reconstructed from the low-resolution images gathered by each lenslet. This is accomplished using superresolution reconstruction algorithms based on linear and nonlinear interpolation algorithms. Two implementations of the ultrathin camera are demonstrated and their performances are compared with that of a conventional infrared camera.     

Image preprocessing for improving computational efficiency in implementation of restoration and superresolution algorithms

Computational complexity is a major impediment to the real-time implementation of image restoration and superresolution algorithms in many applications. Although powerful restoration algorithms have been developed within the past few years utilizing sophisticated mathematical machinery (based on statistical optimization and convex set theory), these algorithms are typically iterative in nature and require a sufficient number of iterations to be executed to achieve the desired resolution improvement that may be needed to meaningfully perform postprocessing image exploitation tasks in practice. Additionally, recent technological breakthroughs have facilitated novel sensor designs (focal plane arrays, for instance) that make it possible to capture megapixel imagery data at video frame rates. A major challenge in the processing of these large-format images is to complete the execution of the image processing steps within the frame capture times and to keep up with the output rate of the sensor so that all data captured by the sensor can be efficiently utilized. Consequently, development of novel methods that facilitate real-time implementation of image restoration and superresolution algorithms is of significant practical interest and is the primary focus of this study. The key to designing computationally efficient processing schemes lies in strategically introducing appropriate preprocessing steps together with the superresolution iterations to tailor optimized overall processing sequences for imagery data of specific formats. For substantiating this assertion, three distinct methods for tailoring a preprocessing filter and integrating it with the superresolution processing steps are outlined. These methods consist of a region-of-interest extraction scheme, a background-detail separation procedure, and a scene-derived information extraction step for implementing a set-theoretic restoration of the image that is less demanding in computation compared with the superresolution iterations. A quantitative evaluation of the performance of these algorithms for restoring and superresolving various imagery data captured by diffraction-limited sensing operations are also presented.     

Super-Resolution Reconstruction of Images Based on Microarray Camera

In the field of images and imaging, super-resolution (SR) reconstruction of images is a technique that converts one or more low-resolution (LR) images into a highresolution (HR) image. The classical two types of SR methods are mainly based on applying a single image or multiple images captured by a single camera. Microarray camera has the characteristics of small size, multi views, and the possibility of applying to portable devices. It has become a research hotspot in image processing. In this paper, we propose a SR reconstruction of images based on a microarray camera for sharpening and registration processing of array images. The array images are interpolated to obtain a HR image initially followed by a convolution neural network (CNN) procedure for enhancement. The convolution layers of our convolution neural network are 3×3 or 1×1 layers, of which the 1×1 layers are used to improve the network performance particularly. A bottleneck structure is applied to reduce the parameter numbers of the nonlinear mapping and to improve the nonlinear capability of the whole network. Finally, we use a 3×3 deconvolution layer to significantly reduce the number of parameters compared to the deconvolution layer of FSRCNN-s. The experiments show that the proposed method can not only ameliorate effectively the texture quality of the target image based on the array images information, but also further enhance the quality of the initial high resolution image by the improved CNN.     

Superresolution image reconstruction from a sequence of aliased imagery

We present a superresolution image reconstruction from a sequence of aliased imagery. The subpixel shifts (displacement) among the images are unknown due to the uncontrolled natural jitter of the imager. A correlation method is utilized to estimate subpixel shifts between each low-resolution aliased image with respect to a reference image. An error-energy reduction algorithm is derived to reconstruct the high-resolution alias-free output image. The main feature of this proposed error-energy reduction algorithm is that we treat the spatial samples from low-resolution images that possess unknown and irregular (uncontrolled) subpixel shifts as a set of constraints to populate an oversampled (sampled above the desired output bandwidth) processing array. The estimated subpixel locations of these samples and their values constitute a spatial domain constraint. Furthermore, the bandwidth of the alias-free image (or the sensor imposed bandwidth) is the criterion used as a spatial frequency domain constraint on the oversampled processing array. The results of testing the proposed algorithm on the simulated low- resolution forward-looking infrared (FLIR) images, real-world FLIR images, and visible images are provided. A comparison of the proposed algorithm with a standard interpolation algorithm for processing the simulated low-resolution FLIR images is also provided.     

Face recognition performance with superresolution

With the prevalence of surveillance systems, face recognition is crucial to aiding the law enforcement community and homeland security in identifying suspects and suspicious individuals on watch lists. However, face recognition performance is severely affected by the low face resolution of individuals in typical surveillance footage, oftentimes due to the distance of individuals from the cameras as well as the small pixel count of low-cost surveillance systems. Superresolution image reconstruction has the potential to improve face recognition performance by using a sequence of low-resolution images of an individual's face in the same pose to reconstruct a more detailed high-resolution facial image. This work conducts an extensive performance evaluation of superresolution for a face recognition algorithm using a methodology and experimental setup consistent with real world settings at multiple subject-to-camera distances. Results show that superresolution image reconstruction improves face recognition performance considerably at the examined midrange and close range.     

Infrared image registration and high-resolution reconstructionusing multiple translationally shifted aliased video frames

Forward looking infrared (FLIR) detector arrays generally produce spatially undersampled images because the FLIR arrays cannot be made dense enough to yield a sufficiently high spatial sampling frequency. Multi-frame techniques, such as microscanning, are an effective means of reducing aliasing and increasing resolution in images produced by staring imaging systems. These techniques involve interlacing a set of image frames that have been shifted with respect to each other during acquisition. The FLIR system is mounted on a moving platform, such as an aircraft, and the vibrations associated with the platform are used to generate the shifts. Since a fixed number of image frames is required, and the shifts are random, the acquired frames will not fall on a uniformly spaced grid. Furthermore, some of the acquired frames may have almost similar shifts thus making them unusable for high-resolution image reconstruction. In this paper, we utilize a gradient-based registration algorithm to estimate the shifts between the acquired frames and then use a weighted nearest-neighbor approach for placing the frames onto a uniform grid to form a final high-resolution image. Blurring by the detector and optics of the imaging system limits the increase in image resolution when microscanning is attempted at sub-pixel movements of less than half the detector width. We resolve this difficulty by the application of the Wiener filter, designed using the modulation transfer function (MTF) of the imaging system, to the high-resolution image. Simulation and experimental results are presented to verify the effectiveness of the proposed technique. The techniques proposed herein are significantly faster than alternate techniques, and are found to be especially suitable for real-time applications     

Adaptive framework for robust high-resolution image reconstruction in multiplexed computational imaging architectures

In multiplexed computational imaging schemes, high-resolution images are reconstructed by fusing the information in multiple low-resolution images detected by a two-dimensional array of low-resolution image sensors. The reconstruction procedure assumes a mathematical model for the imaging process that could have generated the low-resolution observations from an unknown high-resolution image. In practical settings, the parameters of the mathematical imaging model are known only approximately and are typically estimated before the reconstruction procedure takes place. Violations to the assumed model, such as inaccurate knowledge of the field of view of the imagers, erroneous estimation of the model parameters, and/or accidental scene or environmental changes can be detrimental to the reconstruction quality, even if they are small in number. We present an adaptive algorithm for robust reconstruction of high-resolution images in multiplexed computational imaging architectures. Using robust M-estimators and incorporating a similarity measure, the proposed scheme adopts an adaptive estimation strategy that effectively deals with violations to the assumed imaging model. Comparisons with nonadaptive reconstruction techniques demonstrate the superior performance of the proposed algorithm in terms of reconstruction quality and robustness.     

Superresolution optical system with two fixed generalized Damman gratings

We present a system that exceeds the Rayleigh limit of resolution, by placing two fixed gratings in predetermined positions. Lukosz suggested a setup that managed to exceed the Rayleigh limit of resolution [J. Opt. Soc. Am. 56, 1463 (1966); 57, 932 (1967)]. However, Lukosz's technique had some drawbacks that the new suggested system attempts to resolve. Similar to Lukosz's technique, the proposed system works without any moving elements and with no time- or wavelength-restricting conditions. It is suitable for coherent or incoherent two-dimensional imaging. However, the new system contains some important modifications. Although the system uses only two gratings, it is capable of producing superresolution without using an additional imaging lens at the output plane. The generalized Damman gratings allow for obtaining undistorted spectral restoration of information. To achieve superresolution, the input object is duplicated. The trade-off for higher resolution is a smaller field of view. Experimental results validate the theoretical analysis.     

MAP fusion method for superresolution of images with locally varying pixel quality

Superresolution is a procedure that produces a high‐resolution image from a set of low‐resolution images. Many of superresolution techniques are designed for optical cameras, which produce pixel values of well‐defined uncertainty, while there are still various imaging modalities for which the uncertainty of the images is difficult to control. To construct a superresolution image from low‐resolution images with varying uncertainty, one needs to keep track of the uncertainty values in addition to the pixel values. In this paper, we develop a probabilistic approach to superresolution to address the problem of varying uncertainty. As direct computation of the analytic solution for the superresolution problem is difficult, we suggest a novel algorithm for computing the approximate solution. As this algorithm is a noniterative method based on Kalman filter‐like recursion relations, there is a potential for real‐time implementation of the algorithm. To show the efficiency of our method, we apply this algorithm to a video sequence acquired by a forward looking sonar system. © 2008 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 18, 242–250, 2008; Published online in Wiley InterScience (www.interscience.wiley.com).     

Mathematical extrapolation of image spectrum for constraint-set design and set-theoretic superresolution

Several powerful iterative algorithms are being developed for the restoration and superresolution of diffraction-limited imagery data by use of diverse mathematical techniques. Notwithstanding the mathematical sophistication of the approaches used in their development and the potential for resolution enhancement possible with their implementation, the spectrum extrapolation that is central to superresolution comes in these algorithms only as a by-product and needs to be checked only after the completion of the processing steps to ensure that an expansion of the image bandwidth has indeed occurred. To overcome this limitation, a new approach of mathematically extrapolating the image spectrum and employing it to design constraint sets for implementing set-theoretic estimation procedures is described. Performance evaluation of a specific projection-onto-convex-sets algorithm by using this approach for the restoration and superresolution of degraded images is outlined. The primary goal of the method presented is to expand the power spectrum of the input image beyond the range of the sensor that captured the image.     

A Magnifying Glass for Virtual Imaging of Subwavelength Resolution by Transformation Optics

Traditional magnifying glasses can give magnified virtual images with diffraction‐limited resolution, that is, detailed information is lost. Here, a novel magnifying glass by transformation optics, referred to as a “superresolution magnifying glass” (SMG) is designed, which can produce magnified virtual images with a predetermined magnification factor and resolve subwavelength details (i.e., light sources with subwavelength distances can be resolved). Based on theoretical calculations and reductions, a metallic plate structure to produce the reduced SMG in microwave frequencies, which gives good performance verified by both numerical simulations and experimental results, is proposed and realized. The function of SMG is to create a superresolution virtual image, unlike traditional superresolution imaging devices that create real images. The proposed SMG will create a new branch of superresolution imaging technology.     

Synthetic aperture-based beam compression for intravascular ultrasound imaging

In this paper, intravascular ultrasound (IVUS) images acquired with a 64-element array transducer using a multistatic acquisition scheme are presented. The images are reconstructed from a collection of pulse-echo measurements using a synthetic aperture array imaging technique. The main limitations of IVUS imaging are a poor lateral resolution and elevated grating lobes caused by the imaging geometry. We propose a Synthetic Aperture Focusing Technique (SAFT), which uses a limited number of A-scan signals. The focusing process, which is performed in the Fourier domain, requires far less computation time than conventional delay-and-sum methods. Two different reconstruction kernel functions have been derived and are compared for the processing of experimental data     

Superresolution readout system with electrical equalization for optical disks

An electrical equalizer for a superresolution readout system with an optical apodizer is proposed and verified experimentally. This superresolution readout system uses a five-tap transversal filter as the electrical equalizer instead of additional optics to suppress enlarged sidelobes, and it achieves higher resolution than the diffraction-limited system. The transfer function of the electrical equalizer is also derived theoretically. This approach allows fabrication of a readout system with a good signal-to-noise ratio and a compact head.     

Relationship between microscanned image quality and fill factor of detectors

Microscanning is an important technique in high-resolution electro-optical imaging. It can increase the resolution and improve the performance of imaging systems. For optimum design of a staring imaging system with microscanning modes it is necessary to choose the optimum microscanning mode according to the fill factor of the detector. Hence it is important to study the effect of the fill factor on the microscanning image quality. With some assumptions, we introduce the sampling-averaging modulation transfer function of a detector array at the spatial Nyquist frequency with which to study quantitatively the improvement in image quality of various microscanning modes for selected fill factors (1, 2/3, and 1/2). Analytical results show that the amount of improvement is closely associated with the fill factor. Finally, typical sampling imaging of focal plane arrays with these fill factors are simulated. Experimental results qualitatively describe the effect of the fill factor on the microscanning image and show good agreement with theoretical analysis.     

Geometrically superresolved lensless imaging using a spatial light modulator

In this paper we introduce an imaging system based on a reflective phase-only spatial light modulator (SLM) in order to perform imaging with improved geometric resolution. By using the SLM, we combine the realization of two main abilities: a lens with a tunable focus and a phase function that, after proper free-space propagation, is projected as an amplitude distribution on top of the inspected object. The first ability is related to the realization of a lens function combined with a tunable prism that yields a microscanning of the inspected object. This by itself improves the spatial sampling density. The second ability is related to a projection of a phase function that is computed using an iterative beam-shaping Gerchberg-Saxton algorithm. After the free-space propagation from the SLM toward the inspected object, an amplitude pattern is generated on top of the object. This projected pattern and a set of low-resolution images with relative shift are interlaced and, after applying the proper regularization method, a geometrically superresolved image is reconstructed.     

Reconstruction of a high-resolution image on a compound-eye image-capturing system

The authors have proposed an architecture for a compact image-capturing system called TOMBO (thin observation module by bound optics), which uses compound-eye imaging for a compact hardware configuration [Appl. Opt. 40, 1806 (2001)]. The captured compound image is decomposed into a set of unit images, then the pixels in the unit images are processed with digital processing to retrieve the target image. A new method for high-resolution image reconstruction, called a pixel rearrange method, is proposed. The relation between the target object and the captured signals is estimated and utilized to rearrange the original pixel information. Experimental results show the effectiveness of the proposed method. In the experimental TOMBO system, the resolution obtained is four times higher than that of the unit image that did not undergo reconstruction processing.     

Reconstruction of complex signals using minimum Rényi information

An information divergence, such as Shannon mutual information, measures the distance between two probability-density functions (or images). A wide class of such measures, called α divergences, with desirable properties such as convexity over all space, was defined by Amari. Rényi's information Dα is an α divergence. Because of its convexity property, the minimum of Dα is easily attained. Minimization accomplishes minimum distance (maximum resemblance) between an unknown image and a known reference image. Such a biasing effect permits complex images, such as occur in inverse syntheticaperture- radar imaging, to be well reconstructed. The algorithm permits complex amplitudes to replace the probabilities in the Rényi form. The bias image may be constructed as a smooth version of the linear, Fourier reconstruction of the data. Examples on simulated complex image data with and without noise indicate that the Rényi reconstruction approach permits superresolution in low-noise cases and higher fidelity than ordinary, linear reconstructions in higher-noise cases.     

Three-dimensional image sensing and reconstruction with time-division multiplexed computational integral imaging

A method to compute high-resolution three-dimensional images based on integral imaging is presented. A sequence of integral images (IIs) is captured by means of time-division multiplexing with a moving lenslet array technique. For the acquisition of each II, the location of the lenslet array is shifted periodically within the lenslet pitch in a plane perpendicular to the optical axis. The II sequence obtained by the detector array is processed digitally with superresolution reconstruction algorithms to obtain a reconstructed image, appropriate to a viewing direction, which has a spatial resolution beyond the optical limitation.     

Use of modulated excitation signals in medical ultrasound. Part II: Design and performance for medical imaging applications

In the first paper, the superiority of linear FM signals was shown in terms of signal-to-noise ratio and robustness to tissue attenuation. This second paper in the series of three papers on the application of coded excitation signals in medical ultrasound presents design methods of linear FM signals and mismatched filters, in order to meet the higher demands on resolution in ultrasound imaging. It is shown that for the small time-bandwidth (TB) products available in ultrasound, the rectangular spectrum approximation is not valid, which reduces the effectiveness of weighting. Additionally, the distant range sidelobes are associated with the ripples of the spectrum amplitude and, thus, cannot be removed by weighting. Ripple reduction is achieved through amplitude or phase predistortion of the transmitted signals. Mismatched filters are designed to efficiently use the available bandwidth and at the same time to be insensitive to the transducer's impulse response. With these techniques, temporal sidelobes are kept below 60 to 100 dB, image contrast is improved by reducing the energy within the sidelobe region, and axial resolution is preserved. The method is evaluated first for resolution performance and axial sidelobes through simulations with the program Field II. A coded excitation ultrasound imaging system based on a commercial scanner and a 4 MHz probe driven by coded sequences is presented and used for the clinical evaluation of the coded excitation/compression scheme. The clinical images show a significant improvement in penetration depth and contrast, while they preserve both axial and lateral resolution. At the maximum acquisition depth of 15 cm, there is an improvement of more than 10 dB in the signal-to-noise ratio of the images. The paper also presents acquired images, using complementary Golay codes, that show the deleterious effects of attenuation on binary codes when processed with a matched filter, also confirmed by presented simulated images.     

小波多辨率CT成像及处理算法

分析了小波变换进行低分辨率快速图像轮廓重构和局部区域精确重构的算法。在这种算法中，滤波器与小波有关，从而可由反投影得到各种小波图像。通过小波多尺度分析和小波系数控制，提出一种简单算法进行图像增强和噪声去除。与标准的算法相比，该算法提高了重建速度和图像精度。