Super resolution

We present a novel approach for online shrinkage functions learning in single image super-resolution. The proposed approach leverages the classical Wavelet Shrinkage denoising technique where a set of scalar shrinkage functions is applied to the wavelet coefficients of a noisy image. In the proposed approach, a unique set of learned shrinkage functions is applied to the overcomplete representation coefficients of the interpolated input image. The super-resolution image is reconstructed from the post-shrinkage coefficients. During the learning stage, the low-resolution input image is treated as a reference high-resolution image and a super-resolution reconstruction process is applied to a scaled-down version of it. The shapes of all shrinkage functions are jointly learned by solving a Least Squares optimization problem that minimizes the sum of squared errors between the reference image and its super-resolution approximation. Computer simulations demonstrate superior performance compared to state-of-the-art results.

An adaptive two step paradigm for the super-resolution of optical images is developed in this paper. The procedure locally projects image samples onto a family of kernels that are learned from image data. First, an unsupervised feature extraction is performed on local neighborhood information from a training image. These features are then used to cluster the neighborhoods into disjoint sets for which an optimal mapping relating homologous neighborhoods across scales can be learned in a supervised manner. A superresolved image is obtained through the convolution of a low resolution test image with the established family of kernels. Results demonstrate the effectiveness of the approach.

Background: Super-resolution optical fluctuation imaging (SOFI) achieves 3D super-resolution by computing temporal cumulants or spatio-temporal cross-cumulants of stochastically blinking fluorophores. In contrast to localization microscopy, SOFI is compatible with weakly emitting fluorophores and a wide range of blinking conditions. The main drawback of SOFI is the nonlinear response to brightness and blinking heterogeneities in the sample, which limits the use of higher cumulant orders for improving the resolution. Balanced super-resolution optical fluctuation imaging (bSOFI) analyses several cumulant orders for extracting molecular parameter maps, such as molecular state lifetimes, concentration and brightness distributions of fluorophores within biological samples. Moreover, the estimated blinking statistics are used to balance the image contrast, i.e. linearize the brightness and blinking response and to obtain a resolution improving linearly with the cumulant order. Results: Using a widefield total-internal-reflection (TIR) fluorescence microscope, we acquired image sequences of fluorescently labelled microtubules in fixed HeLa cells. We demonstrate an up to five-fold resolution improvement as compared to the diffraction-limited image, despite low single-frame signal-to-noise ratios. Due to the TIR illumination, the intensity profile in the sample decreases exponentially along the optical axis, which is reported by the estimated spatial distributions of the molecular brightness as well as the blinking on-ratio. Therefore, TIR-bSOFI also encodes depth information through these parameter maps. Conclusions: bSOFI is an extended version of SOFI that cancels the nonlinear response to brightness and blinking heterogeneities. The obtained balanced image contrast significantly enhances the visual perception of super-resolution based on higher-order cumulants and thereby facilitates the access to higher resolutions. Furthermore, bSOFI provides microenvironment-related molecular parameter maps and paves the way for functional super-resolution microscopy based on stochastic switching.

Super-resolution enhancement techniques may be used sequence move with subpixel increments. A Bayesian multito estimate a high-resolution still from several low-resoluframe enhancement algorithm is presented to compute an tion video frames, provided that objects within the image HRVS using the spatial information present within each frame sequence move with subpixel increments . To properly as well as the temporal information present due to object motion incorporate temporal correlations into the multiframe obbetween frames. However, the required subpixel-resolution moservation model, high-quality subpixel motion vectors must tion vectors must be estimated from low-resolution and noisy be estimated between video frames . Methods for video frames, resulting in an inaccurate motion field which can adversely impact the quality of the enhanced image. Several attaining subpixel accuracy generally employ an interpolasubpixel motion estimation techniques are incorporated into tion of the image sequence frames, followed by the applicathe Bayesian multiframe enhancement algorithm to determine tion of a parametric or nonparametric motion estimation their efficacy in the presence of global data transformations scheme. The accuracy of the estimated motion fields has between frames (i.e., camera pan, rotation, tilt, and zoom) and a direct influence on the quality of the high-resolution independent object motion. Visual and quantitative comparivideo still (HRVS) image.

Fast magnetic resonance imaging slice acquisition techniques such as single shot fast spin echo are routinely used in the presence of uncontrollable motion. Current applications involve fetal MRI and MRI of moving subjects and organs. Although high-quality slices are frequently acquired by these techniques, inter-slice motion leads to severe motion artifacts that appear in out-of-plane views. Slice sequential acquisitions do not enable 3D volume representation. In this study, we have developed a novel technique based on a slice acquisition model, which enables the reconstruction of a volumetric image from multiple-scan slice acquisitions. The super-resolution volume reconstruction is formulated as an inverse problem of finding the underlying structure generating the acquired slices. We have developed a robust M-estimation solution which minimizes a robust error norm function between the model-generated slices and the acquired slices. The accuracy and robustness of this novel technique has been quantitatively assessed through simulations with digital brain phantom images as well as high-resolution newborn images. We also report here successful application of our new technique for the reconstruction of volumetric fetal brain MRI from clinically acquired data.

Recently, there has been a great deal of work developing super-resolution algorithms for combining a set of lowquality images to produce a set of higher quality images. Either explicitly or implicitly, such algorithms must perform the joint task of registering and fusing the low-quality image data. While many such algorithms have been proposed, very little work has addressed the performance bounds for such problems. In this paper, we analyze the performance limits from statistical first principles using Cramér-Rao inequalities. Such analysis offers insight into the fundamental super-resolution performance bottlenecks as they relate to the subproblems of image registration, reconstruction, and image restoration.

This paper demonstrates a super-resolution method for improving the resolution in clinical positron emission tomography (PET) scanners. Super-resolution images were obtained by combining four data sets with spatial shifts between consecutive acquisitions and applying an iterative algorithm. Super-resolution attenuation corrected PET scans of a phantom were obtained using the two-dimensional and three-dimensional (3-D) acquisition modes of a clinical PET/computed tomography (CT) scanner (Discovery LS, GEMS). In a patient study, following a standard 18 F FDG PET/CT scan, a super-resolution scan around one small lesion was performed using axial shifts without increasing the patient radiation exposure. In the phantom study, smaller features (3 mm) could be resolved axially with the super-resolution method than without (6 mm). The super-resolution images had better resolution than the original images and provided higher contrast ratios in coronal images and in 3-D acquisition transaxial images. The coronal super-resolution images had superior resolution and contrast ratios compared to images reconstructed by merely interleaving the data to the proper axial location. In the patient study, super-resolution reconstructions displayed a more localized 18 F FDG uptake. A new approach for improving the resolution of PET images using a super-resolution method has been developed and experimentally confirmed, employing a clinical scanner. The improvement in axial resolution requires no changes in hardware.

In this paper we formulate a new time dependent convolutional model for superresolution based on a constrained variational model that uses the total variation of the signal as a regularizing functional. We propose an iterative refinement procedure based on Bregman iteration to improve spatial resolution. The model uses a dataset of low resolution images and incorporates a downsampling operator to relate the high resolution scale to the low resolution one. We present an algorithm for the model and we perform a series of numerical experiments to show evidence of the good behavior of the numerical scheme and quality of the results.

Development of focal plane arrays (FPA) for mm wavelength and THz radiation is presented in this paper. The FPA is based upon inexpensive neon indicator lamp Glow Discharge Detectors (GDDs) that serve as pixels in the FPA. It was shown in previous investigations that inexpensive neon indicator lamps GDDs are quite sensitive to mm wavelength and THz radiation. The diameter of GDD lamps are typically 3-6 mm and thus the FPA can be diffraction limited. Development of an FPA using such devices as detectors is advantageous since the costs of such a lamp is around 30-50 cents per lamp, and it is a room temperature detector sufficiently fast for video frame rates. Recently a new 8X8 GDD FPA VLSI board was designed, constructed, and experimentally tested. First THz images as well as DSP methods using this GDD FPA are demonstrated. Super resolution was achieved by moving the 8X8 pixel board appropriately in the image plane so that 32X32 pixel images are also obtained and shown here, with much improved image quality because of much reduced pixelization distortion.

Deep learning and image processing are two areas of great interest to academics and industry professionals alike. The areas of application of these two disciplines range widely, encompassing fields such as medicine, robotics, and security and surveillance. The aim of this book, 'Deep Learning for Image Processing Applications', is to offer concepts from these two areas in the same platform, and the book brings together the shared ideas of professionals from academia and research about problems and solutions relating to the multifaceted aspects of the two disciplines. The first chapter provides an introduction to deep learning, and serves as the basis for much of what follows in the subsequent chapters, which cover subjects including: the application of deep neural networks for image classification; hand gesture recognition in robotics; deep learning techniques for image retrieval; disease detection using deep learning techniques; and the comparative analysis of deep data and big data. The book will be of interest to all those whose work involves the use of deep learning and image processing techniques

Scope of the book:
This book focusses on the technical concepts of deep learning and its associated branch Neural Networks for the various dimensions of image processing applications. The proposed volume intends to bring together researchers to report the latest results or progress in the development of the above-mentioned areas. Since there is a deficit of books on this specific subject matter, the editors aim to provide a common platform for researchers working in this area to exhibit their novel findings.
Topics of Interest:
This book solicits contributions, which include the fundamentals in the field of Deep Artificial Neural Networks and Image Processing supported by case studies and practical examples. Each chapter is expected to be self-contained and to cover an in-depth analysis of real life applications of neural networks to image analysis.

Super-resolution (SR) is the term used to define the process of estimating a high-resolution (HR) image or a set of HR images from a set of low-resolution (LR) observations. In this paper we propose a class of SR algorithms based on the maximum a posteriori (MAP) framework. These algorithms utilize a new multichannel image prior model, along with the state-of-the-art single channel image prior and observation models. A hierarchical (twolevel) Gaussian nonstationary version of the multichannel prior is also defined and utilized within the same framework. Numerical experiments comparing the proposed algorithms among themselves and with other algorithms in the literature, demonstrate the advantages of the adopted multichannel approach.

Term “super-resolution” is typically used for a high-resolution image produced from several low-resolution noisy observations. In this paper, we consider the problem of high-quality interpolation of a single noise-free image. Several aspects of the corresponding super-resolution algorithm are investigated: choice of regularization term, dependence of the result on initial approximation, convergence speed, and heuristics to facilitate convergence and improve the visual quality of the resulting image.

In the last two decades, many papers have been published, proposing a variety methods of multi-frame resolution enhancement. These methods are usually very sensitive to their assumed model of data and noise, which limits their utility. This paper reviews some of these methods and addresses their shortcomings. We propose a different implementation using L 1 norm minimization and robust regularization to deal with different data and noise models. This computationally inexpensive method is robust to errors in motion and blur estimation, and results in sharp edges. Simulation results confirm the effectiveness of our method and demonstrate its superiority to other robust super-resolution methods.

Accurate maps of the static environment are essential for many advanced driver-assistance systems. In this paper a new method for the fast computation of occupancy grid maps with laser range-finders and radar sensors is proposed. The approach utilizes the Graphics Processing Unit to overcome the limitations of classical occupancy grid computation in automotive environments. It is possible to generate highly accurate grid maps in just a few milliseconds without the loss of sensor precision. Moreover, in the case of a lower resolution radar sensor it is shown that it is suitable to apply super-resolution algorithms to achieve the accuracy of a higher resolution laserscanner. Finally, a novel histogram based approach for road boundary detection with lidar and radar sensors is presented.

This paper presents a non-local kernel regression (NL-KR) method for image and video restoration tasks, which exploits both the non-local self-similarity and local structural regularity in natural images. The non-local self-similarity is based on the observation that image patches tend to repeat themselves in natural images and videos; and the local structural regularity reveals that image patches have regular structures where accurate estimation of pixel values via regression is possible. Explicitly unifying both properties, the ...

This paper presents a new approach to single-image superresolution, based on sparse signal representation. Research on image statistics suggests that image patches can be wellrepresented as a sparse linear combination of elements from an appropriately chosen over-complete dictionary. Inspired by this observation, we seek a sparse representation for each patch of the low-resolution input, and then use the coefficients of this representation to generate the high-resolution output. Theoretical results from compressed sensing suggest that under mild conditions, the sparse representation can be correctly recovered from the downsampled signals. By jointly training two dictionaries for the low-and high-resolution image patches, we can enforce the similarity of sparse representations between the low resolution and high resolution image patch pair with respect to their own dictionaries. Therefore, the sparse representation of a low resolution image patch can be applied with the high resolution image patch dictionary to generate a high resolution image patch. The learned dictionary pair is a more compact representation of the patch pairs, compared to previous approaches, which simply sample a large amount of image patch pairs [1], reducing the computational cost substantially. The effectiveness of such a sparsity prior is demonstrated for both general image superresolution and the special case of face hallucination. In both cases, our algorithm generates high-resolution images that are competitive or even superior in quality to images produced by other similar SR methods. In addition, the local sparse modeling of our approach is naturally robust to noise, and therefore the proposed algorithm can handle super-resolution with noisy inputs in a more unified framework.

In this paper, a super-resolution technique is proposed that uses a combination of bicubic interpolation and wavelet transform. Bicubic interpolation produces a high resolution image but is prone to blurring artifact. So the blurring artifact is reduced in the wavelet domain. The input low-resolution is up-sampled using bicubic interpolation. The edges of the resultant highresolution image are enhanced using stationary wavelet transform (SWT). SWT is applied to the image to produce sub-bands of the image and then these sub-bands are modified by multiplying with a boost value. Then these sub-bands are combined using inverse stationary wavelet transform (ISWT) to produce the final highresolution image. The quantitative and qualitative analysis illustrate that the proposed technique is provides superior results as compared to other existing techniques.

Since the work of Golgi and Cajal, light microscopy has remained a key tool for neuroscientists to observe cellular properties. Ongoing advances have enabled new experimental capabilities using light to inspect the nervous system across multiple spatial scales, including ultrastructural scales finer than the optical diffraction limit. Other progress permits functional imaging at faster speeds, at greater depths in brain tissue, and over larger tissue volumes than previously possible. Portable, miniaturized fluorescence microscopes now allow brain imaging in freely behaving mice. Complementary progress on animal preparations has enabled imaging in head-restrained behaving animals, as well as time-lapse microscopy studies in the brains of live subjects. Mouse genetic approaches permit mosaic and inducible fluorescence-labeling strategies, whereas intrinsic contrast mechanisms allow in vivo imaging of animals and humans without use of exogenous markers. This review surveys such advances and highlights emerging capabilities of particular interest to neuroscientists.

In order to improve signal-to-noise ratio ͑SNR͒ and contrastto-noise ratio, we introduces a novel tunable forward-and-backward ͑TFAB͒ diffusion approach for image restoration and edge enhancement. In the TFAB algorithm, an alternative forward-and-backward ͑FAB͒ diffusion process is presented, where it is possible to better modulate all aspects of the diffusion behavior and it shows better algorithm behavior compared to the existing FAB diffusion approaches. In addition, there is no necessity to laboriously determine the value of the gradient threshold. We believe the TFAB diffusion to be an adaptive mechanism for image restoration and enhancement. Qualitative experiments, based on various general digital images and a magnetic resonance image, show significant improvements when the TFAB diffusion algorithm is used versus the existing anisotropic diffusion and the previous FAB diffusion algorithms for enhancing edge features and improving image contrast. Quantitative analyses, based on peak SNR and the universal image quality index, confirm the superiority of the proposed algorithm.

Driven by key law enforcement and commercial applications, research on face recognition from video sources has intensified in recent years. The ensuing results have demonstrated that videos possess unique properties that allow both humans and automated systems to perform recognition accurately in difficult viewing conditions. However, significant research challenges remain as most video based applications do not allow for controlled recordings.

Realizing a smaller or sharper diffractive center spot is a valuable research aim in soft X-ray focus and other related research applications. Fresnel zone plates (FZP) and photon sieves (PS) are often used to focus the X-rays or other wavelength light at present. Here, we show that combination of a super-resolution phase mask (SPM) and an FZP (or PS) as one diffractive optical element can realize a smaller or sharper diffractive center spot without significantly increasing the fabrication difficulty. All these diffractive phase elements can be applied to beam shaping, mask-less lithography, energy congregation in high power lasers, soft X-rays focus, and any other field that requires a smaller or sharper diffractive center spot.

Wireless multipath channels can often be characterized as sparse, i.e., the number of significant paths is small even when the channel delay spread is large. This can be taken advantage of when estimating the unknown channel frequency response using pilot assisted modulation. Other work has largely focused on the greedy orthogonal matching pursuit (OMP) algorithm, using a dictionary based on an equivalent finite impulse response filter to model the channel. This is not necessarily realistic, as the physical nature of the channel is continuous in time, while the equivalent filter taps are based on baseband sampling. In this paper, we consider sparse channel estimation using a continuous time path-based channel model. This can be linked to the direction finding problem from the array processing literature and solved using the well-known root-MUSIC and ESPRIT algorithms, which have no formal time resolution. In addition, we show that a dictionary with finer time resolution considerably improves the performance of OMP and the related Basis Pursuit (BP) algorithm.

. Super-resolution of a scanned keyboard with a high-resolution scan of one of its keys: original model (top), reconstructed with the detail exemplar of a key scanned at high resolution (middle) and reconstructed with the exemplar key generated by low resolution accumulation (bottom).

Recognition and identification ranges are limited to the quality of the images. Both the received contrast and the spatial resolution determine if objects are recognizable. Several aspects affect the image quality. First of all the sensor itself. The image quality depends on the size of the infrared detector array and the sensitivity. Second, also the intervening atmosphere, in particular over longer ranges, has an impact on the image quality. It degrades the contrast, due to transmission effects, as well as it influences the resolution, due to turbulence blur, of the image. We present studies in the field of infrared image enhancement. Several techniques are described: noise reduction, super resolution, turbulence compensation, contrast enhancement, stabilization. These techniques operate in real-time on COTS/MOTS platforms. They are especially effective in the army theatre, where long horizontal paths, and short line-of-sight limited urban operations are both present. Application of these techniques on observation masts, such as on military camp sites, and on UAVs and moving ground vehicles are discussed. Examples will be presented from several trials in which these techniques were demonstrated, including the presentation of test results.

Stochastic regularized methods are quite advantageous in super-resolution (SR) image reconstruction problems. In the particular techniques, the SR problem is formulated by means of two terms, the datafidelity term and the regularization term. The present work examines the effect of each one of these terms on the SR reconstruction result with respect to the presence or absence of noise in the low-resolution (LR) frames. Experimentation is carried out with the widely employed L 2 , L 1 , Huber and Lorentzian estimators for the data-fidelity term. The Tikhonov and Bilateral (B) Total Variation (TV) techniques are employed for the regularization term. The extracted conclusions can, in practice, help to select an effective SR method for a given sequence of LR frames. Thus, in case that the potential methods present common data-fidelity or regularization term, and frames are noiseless, the method which employs the most robust regularization or data-fidelity term should be used. Otherwise, experimental conclusions regarding performance ranking vary with the presence of noise in frames, the noise model as well as the difference in robustness of efficiency between the rival terms. Estimators employed for the data-fidelity term or regularizations stand for the rival terms.

Spotting the difference between bird species is a challenging task because the minute characteristics cannot be distinguished by the human eye. This identification problem is an example of fine grained classification task. The algorithm in such cases should be strong enough to identify small differences .However; there can be multiple problems such as orientation of bird, background colours, brightness and intensity of image which makes it difficult to correctly classify an image. The proposed paper presents a novel approach towards identification of species by increasing the resolution of image and applying Mask RCNN and custom net for classification. SRCNN with Mask RCNN and Inception V3 achieves 51.73% precison , 49.68% recall and 53.52 F1 Scor

The authors describe the application of modern spectral analysis techniques to synthetic aperture radar data. The purpose is to improve the geometrical resolution of the image with respect to the numerical values related to the compressed coded waveform and the synthetic aperture, so that subsequent classification procedures will have improved performance as well. The classical spectral estimator, i.e. the FFT, produces an image with resolution in azimuth and range bounded by the Rayleigh limits. Super-resolved images are obtained by replacing the FFT with parametric spectral estimators such as those built around an autoregressive model of the dechirped signal. The proposed processing scheme is based on a twodimensional covariance method. The expected improvement in resolution is discussed together with the results of a simulation analysis. The application of the technique to images captured by an airborne SAR resulted in a resolution gain factor of about two, The paper concludes with a perspective on future research and applications.

An approach that uses an electro-optically tunable two dimensional phase grating to enhance the resolution in digital holographic microscopy is proposed. We show that, by means of a flexible hexagonal phase grating, it is possible to increase the numerical aperture of the imaging system, thus improving the spatial resolution of the images in two dimensions. The augment of the numerical aperture of the optical system is obtained by recording spatially multiplexed digital holograms. The grating tuneability allows one to adjust the intensity among the spatially multiplexed holograms maximizing the grating diffraction efficiency. Furthermore we demonstrate that the flexibility of the numerical reconstruction allows one to use selectively the diffraction orders carrying useful information for increasing the spatial resolution. The proposed approach can improve the capabilities of digital holography in threedimensional imaging and microscopy.

The exact molecular mechanisms of ovarian cancer platinum resistance are not well understood, and biomarkers to reliably predict ovarian cancer resistance to platinum and other chemotherapeutic agents are lacking. Biomechanics of cisplatin-treated ovarian cancer cells were measured quantitatively at nanoscale level using atomic force microscopy. We demonstrate that cisplatin modulates the cellular nanomechanics of ovarian cancer cells; sensitive cells show dose-dependent increase in cell stiffness, which is effected by disrupting the Factin polymerization. In contrast, resistant cells show no significant changes in cell stiffness upon cisplatin treatment. Further, stimulated emission depletion, an emerging super-resolution microscopy, shows that at the molecular level, F-actin is indeed remodeled considerably in cisplatin-sensitive and cisplatin-resistant cells. These findings reveal a direct role of the actin remodeling mechanism in cisplatin resistance of ovarian cancer cells, suggesting potential future applications of nanomechanical profiling as a marker for cancer drug sensitivity.

Background: Harmonic Nanoparticles are a new family of exogenous markers for multiphoton imaging exerting optical contrast by second harmonic (SH) generation. In this tutorial, we present the application of Hyper-Rayleigh Scattering (HRS) for a quantitative assessment of the nonlinear optical properties of these particles and discuss the underlying theory and some crucial experimental aspects. Methods: The second harmonic properties of BaTiO 3 , KNbO 3 , KiTiOPO 4 (KTP), LiNbO 3 and ZnO nanocrystals (NCs) are investigated by HRS measurements after careful preparation and characterization of colloidal suspensions. Results: A detailed analysis of the experimental results is presented with emphasis on the theoretical background and on the influence of some experimental parameters including the accurate determination of the nanocrystal size and concentration. The SH generation efficiency and averaged nonlinear optical coefficients are then derived and compared for six different types of NCs. Conclusions: After preparation of colloidal NC suspensions and careful examination of their size, concentration and possible aggregation state, HRS appears as a valuable tool to quantitatively assess the SH efficiency of noncentrosymmetric NCs. All the investigated nanomaterials show high SH conversion efficiencies, demonstrating a good potential for bio-labelling applications.

Mosaicing and super resolution are two ways to combine information from multiple frames in video sequences. Mosaicing displays the information of multiple frames in a single panoramic image. Super-resolution uses regions which appear in multiple frames to improve resolution and reduce noise. Simple combination of the two methods is problematic since the alignment used for mosaicing may not be accurate enough over the entire region for super resolution. In this paper we introduce a 2-step process: First we create a panoramic mosaic from the images. We then align small image regions to the panorama, and apply super resolution resulting in a panoramic image with higher resolution.

Gradient based motion estimation techniques (GM) are considered to be in the heart of stateof-the-art registration algorithms, being able to account for both pixel and subpixel registration and to handle various motion models (translation, rotation, a¢ne, projective). These methods estimate the motion between two images based on the local changes in the image intensities while assuming image smoothness. This paper o¤ers two main contributions: (i) Enhancement of the GM technique by introducing two new bidirectional formulations of the GM. This improves the convergence properties for large motions. (ii) We present an analytical convergence analysis of the GM and its properties. Experimental results demonstrate the applicability of these algorithms to real images.

Modern avionics equipment, such as super resolution direction-finding systems, now require resolutions on the order of 20 to 22 bits. Oversampled analog-todigital converter architectures offer a means of exchanging resolution in time for that in amplitude and represent an attractive approach to implementing precision converters without the need for complex precision analog circuits. Using oversampling techniques based on sigma-delta (EA) modulation, a convenient tradeoff exists between sampling rate and resolution. One of the major advantages of integrated optics is the capability to efficiently couple wideband signals into the optical domain. Typically, EA processors require simple and relatively low-precision analog components and thus are well suited to integrated optical implementations. This paper presents the design of a first-order, integrated optical EA modulator.

In general, all the video Super-Resolution (SR) algorithms present the important drawback of a very high computational load, mainly due to the huge amount of operations executed by the Motion Estimation (ME) stage. Commonly, there is a trade-off between the accuracy of the estimated motion, given as a Motion Vector (MV), and the computational cost associated. In this sense, the ME algorithms that explore more exhaustively the search area among images use to deliver better MVs, at the cost of a higher computational load and resources use. Due to this reason, the proper choice of a ME algorithm is a key factor not only to reach real-time applications, but also to obtain high quality video sequences independently of their characteristics. Under the hardware point of view, the preferred ME algorithms are based on matching fixed-size blocks in different frames. In this paper, a comparison of nine of the most representative Fast Block Matching Algorithms (FBMAs) is made in order to select the one which presents the best tradeoff between video quality and computational cost, thus allowing reliable real-time hardware implementations of video super-resolution systems 1 .

Many scalable video coding systems use frame downsampling in order to reduce complexity and to enable enhancement layers. Super-resolution (SR) can be used to help the up-sampling and recovering processes of those frames. We are interested in reversed-complexity (distributed) coding methods, wherein few key frames are encoded at normal resolution, while the rest are downsampled and encoded at reduced resolution along with the enhancement layers. We are only interested on the decoder side, wherein we carry motion estimation of the downsampled frames using the key frames as references. When a match is made, the high-frequency components of the key frames (KF) are used to super-resolve the non-key frames (NKF). The motion estimation process is performed using blocks of band-pass versions of the frames, rather than lowpass ones. Results indicate the improved performance of the proposed super-resolution algorithm.

Multiframe super-resolution (SR) reconstruction aims to produce a high-resolution (HR) image using a set of low-resolution (LR) images. In the process of reconstruction, fuzzy registration usually plays a critical role. It mainly focuses on the correlation between pixels of the candidate and the reference images to reconstruct each pixel by averaging all its neighboring pixels. Therefore, the fuzzy-registration-based SR performs well and has been widely applied in practice. However, if some objects appear or disappear among LR images or different angle rotations exist among them, the correlation between corresponding pixels becomes weak. Thus, it will be difficult to use LR images effectively in the process of SR reconstruction. Moreover, if the LR images are noised, the reconstruction quality will be affected seriously. To address or at least reduce these problems, this paper presents a novel SR method based on the Zernike moment, to make the most of possible details in each LR ima...