Modern methods of computational photography make it possible to bring the quality of images obtained by mobile cameras closer to the quality of professional cameras. One of the most important tasks is that of ensuring the consistency of colors from different cameras. In this paper, we propose a simple and efficient way to bring the colors of one camera to another, based on the approximation of the required transformation by a tone correction spline and a color transformation matrix. An experimental study was carried out in a rather complicated case, in which it was required to match colors of the images obtained from two fundamentally different sensors, as well as using diffractive optics. The results of the experiments showed that the proposed method allows one to obtain a higher accuracy of color matching between cameras than existing analogues.

Currently, more and more research projects are being carried out in the field of creating flat optics-based imaging systems. The use of flat optics allows one to significantly reduce the weight and simplify the design of the optical system, which is a great advantage over classic refractive analogs. The main drawback of flat optical elements is significant reduction in the quality of captured images due to strong chromatic aberration. This article shows how this degradation in quality can be dealt with using a Model-Agnostic Meta-Learning approach.

In previous works, the generative adversarial network (GAN) was successfully used to remove the image-wise color distortion of images captured by diffractive optics. It showed good image quality for a test sample. However, some reconstruction artifacts were produced by GAN for real scene images. In the paper, we study the nature of these artifacts. We show how overexposure and cross-like markers affect the occurrence of artifacts.

The pressure to reduce weight and improve image quality of the imaging devices continues to push research in the area of flat optics with computational image reconstruction. This paper presents a new end-to-end framework applying two convolutional neural networks (CNNs) to reconstruct images captured with multilevel diffractive lenses (MDLs). We show that the patch-wise chromatic blur and image-wise context-aware color highlights, the distortions inherent to MDLs, can be successfully addressed with the suggested reconstruction pipeline. The generative adversarial network (GAN) is first used to remove image-wise color distortion, while a patch-wise network is then used to apply chromatic deblur. The proposed approach produces better image quality improvement than the context-independent color correction with a deconvolution-based chromatic deblur. We also show that the proposed end-to-end reconstruction is equally applicable for single-and multi-aperture MDL-based imaging systems.

The article describes a technique that allows to form a planned projection of the aircraft using the minimum number of iterations (refinements based on aerodynamic and weight calculations results). The elements of the technique are described via example of the development (synthesis) of the layout scheme of a front-line aircraft with two engines and internal cargo compartments, performed according to the normal balancing scheme. Synthesis is carried out on the basis of certain pre-design parameters, limitations, as well as solutions set by the designer, based on the analysis of parametric ratios and statistical dependencies.

The use of multi-aperture cameras is one of the modern trends for imaging devices, both consumer-grade and professional. This paper presents the creation of multi-aperture cameras based on long-focus single diffraction lenses. These lenses are several times better than the common lenses in terms of weight and cost, but they are significantly inferior in quality of the resulting image, and therefore they require computational reconstruction stage. We introduce various schemes of multi-aperture diffraction lenses, allowing to increase both the viewing angle and the resolution of the imaging system. We propose a convolutional neural network for image reconstruction in multi-aperture diffraction optical systems.

This article is devoted to resolving problems with synthesis of FIR-filter restoring distortions such as defocusing. We propose a new parametrical class of finite impulse response filters (FIR-filters) based on a model of the one-dimensional radially-symmetric frequency response. In the proposed synthesis method, the one-dimensional frequency response was composed of square and exponential functions. The two-dimensional filter's impulse response (SE-filter) was constructed by sampling one-dimensional impulse responses for all directions. For filter parameters estimation, the iterative scheme was used. We considered two approaches to estimating the filter parameters: with training images and in the absence of training images. The important advantages of the proposed synthesis method are the possibility of high quality restoration with a minimum of prior information and the low amount of computing resources required. Given examples illustrates the possibility of the high-quality distortion correction. The proposed method provides higher quality restorations than optimal Wiener filter (from Open CV) due to a lack of distortion on the border of images.

Abstract Multi-aperture imaging systems are one of the modern trends for imaging devices. The paper presents the results of developing a multi-aperture imaging system based on single diffractive lenses and postprocessing pipeline based on deep learning. The pipeline includes two processing steps: color correction using a generative adversarial network and chromatic deblurring by a convolutional neural network.

In this paper, we describe our advances in manufacturing a 256-layer 7-mu m thick harmonic lens with 150 and 300 mm focal distances combined with color correction, deconvolution, and a feedforwarding deep learning neural network capable of producing images approaching photographic visual quality. While reconstruction of images taken with diffractive optics was presented in previous works, this paper is the first to use deep neural networks during the restoration step. The level of imaging quality we achieved with our imaging system can facilitate the emergence of ultralightweight remote sensing cameras for nano- and picosatellites, and for aerial remote sensing systems onboard small UAVs and solar-powered airplanes.

The paper addresses the problem of imaging quality enhancement for the Offner hyperspectrometer using a convolutional neural network. We use a deep convolutional neural network with residual training and PReLU activation, inspired by the super-resolution task for RGB images. In the case of hyperspectral imaging, it is often a problem to find a large enough ground truth dataset for training a neural network from scratch. Transfer learning using the network pretrained for RGB images with some pre- and postprocessing is one of the possible workarounds. In this paper, we propose to simulate the necessary ground truth data using non-imaging spectrometer. The obtained dataset with partially simulated ground truth is then used to train the convolutional neural network directly for hyperspectral image quality enhancement. The proposed training approach also allows to incorporate distortions specific for hyperspectral images into the enhancement procedure. It allows to successfully remove the striping distortions inherent to the Offner scheme of image acquisition. The experimental results of the proposed approach show a significant quality gain.

This paper considers a vegetation type recognition algorithm in which the conjugacy indicator with a subspace spanned by endmember vectors is taken as a proximity measure. We show that with proper data preprocessing, including vector components weighting and class partitioning into subclasses, the proposed method offers a higher recognition quality when compared to a support vector machine (SVM) method implemented in MatLab software. This implementation provides good results with the SVM method for a fairly difficult classification test using the Indian Pines dataset with 16 classes containing similar vegetation types. The difficulty of the test is caused by high correlation between the classes. Thus, the results show a possibility for the recognition of a large variety of vegetation types, including the narcotic plants.

Here, we briefly describe the real-time fMRI data that is provided for testing the functionality of the open-source Python/Matlab framework for neurofeedback, termed Open NeuroFeedback Training (OpenNFT, Koush et al. [1]). The data set contains real-time fMRI runs from three anonymized participants (i.e., one neurofeedback run per participant), their structural scans and pre-selected ROIs/masks/weights. The data allows for simulating the neurofeedback experiment without an MR scanner, exploring the software functionality, and measuring data processing times on the local hardware. In accordance with the descriptions in our main article, we provide data of (1) periodically displayed (intermittent) activation-based feedback; (2) intermittent effective connectivity feedback, based on dynamic causal modeling (DCM) estimations; and (3) continuous classification-based feedback based on support-vector-machine (SVM) estimations. The data is available on our public GitHub repository: https://github.com/OpenNFT/OpenNFT_Demo/releases.

Neurofeedback based on real-time functional magnetic resonance imaging (rt-fMRI) is a novel and rapidly developing research field. It allows for training of voluntary control over localized brain activity and connectivity and has demonstrated promising clinical applications. Because of the rapid technical developments of MRI techniques and the availability of high-performance computing, new methodological advances in rt-fMRI neurofeedback become possible. Here we outline the core components of a novel open-source neurofeedback framework, termed Open NeuroFeedback Training (OpenNFT), which efficiently integrates these new developments. This framework is implemented using Python and Matlab source code to allow for diverse functionality, high modularity, and rapid extendibility of the software depending on the user's needs. In addition, it provides an easy interface to the functionality of Statistical Parametric Mapping (SPM) that is also open-source and one of the most widely used fMRI data analysis software. We demonstrate the functionality of our new framework by describing case studies that include neurofeedback protocols based on brain activity levels, effective connectivity models, and pattern classification approaches. This open-source initiative provides a suitable framework to actively engage in the development of novel neurofeedback approaches, so that local methodological developments can be easily made accessible to a wider range of users.

Diffractive optical elements (DOE) have significant advantage over refractive optics in linear dimensions and weight. However, the quality of images produced by a diffractive optics-based system suffers from strong distortions. In previous papers, we introduced and enhanced post-capture computational image reconstruction for diffractive Fresnel lens. In this paper, we revise chromatic shift correction method. Energy redistribution between diffractive orders results in chromatic shift and affects image quality. Each of the RGB channels has nonlinear flare spots from two other channels. In this paper, we present multicomponent color correction using three-dimensional lookup tables that takes into account the nonlinear relationship between color channels. PSNR values are used to compare the quality of the presented multicomponent approach with a previously described single-channel method.

In recent years, several pioneering works were dedicated to imaging systems based on simple diffractive structures like Fresnel lenses or phase zone plates. Such systems are much lighter and cheaper than classical refractive optical systems. However, the quality of images obtained by diffractive optics suffers from stronger distortions of various types. In this paper, we show that a combination of the high-precision lens design with post-capture computational reconstruction allows one to attain a much higher image quality. The proposed reconstruction procedure uses a sequence of color correction, deconvolution, and a feedforward deep learning neural network. An improvement both in lens manufacturing and in image processing may contribute to the emergence of ultra-lightweight imaging systems varying from cameras for nano- and picosatellites to surveillance systems.

With suggested computational post-processing workflow for correcting optical distortions, the Fresnel lens can finally be used in lightweight and inexpensive computer vision sensors. Common methods for image enhancement do not comprehensively address the blurring artifacts caused by strong chromatic aberrations in images produced by a simple Fresnel optical system. To deliver image quality acceptable for general-purpose color imaging, we propose a computational post-capture processing to enhance the quality of images acquired with a 256-level Fresnel lens. The PSNR quality measure is then applied to estimate resulting quality for different deblurring techniques. A novel technique that removes chromatic blur without computationally expensive deconvolution can be considered a breakthrough as it finally enables in-camera embedded post-processing.

This paper presents a novel identification-based image correction method using a bi-illuminant dichromatic reflection model. Image patches with uniform properties over distorted and distortion-free images or image parts are used as a prior knowledge for identification. We identify the distortion correction function on a set of these patches, called spectrum shape elements, with the Hausdorffmetric. The main issue during prior knowledge representation is for each distorted spectrum shape element to find a corresponding distortion-free element. A necessary condition to find a matching spectrum shape element is presented and theoretically proved. Identification problem was solved using a RANSAC-based optimization with this necessary condition as an optimization constraint. The method works well both for color and hyperspectral images. The proposed image correction procedure was tested on a set of color images and AVIRIS hyperspectral remote sensing data and proved to provide the quality superior to the results obtained with Retinex correction. (C) 2016 Elsevier B.V. All rights reserved.

This paper describes a unified approach to correct optical distortions in images formed by a Fresnel lens with computational post-processing that opens up new opportunities to use Fresnel lenses in lightweight and inexpensive computer vision devices. Traditional methods of aberration correction do not address artifacts introduced by a Fresnel lens in a systematic way and thus fail to deliver image quality acceptable for general-purpose color imaging. In our approach, the image is restored using three steps: first, by deblurring the base color channel, then by sharpening other two channels, and finally by applying color correction. Deblurring and sharpening remove significant chromatic aberration and are similar to the restoration technique used for images formed by simple refraction lenses. Color correction stage removes strong color shift caused by energy redistribution between diffraction orders of Fresnel lens. This post-capture processing was tested on real images formed by a four-step approximation of the Fresnel lens manufactured in our optics laboratory.