Fast Super-resolution Reconstruction Research Articles

Research on Detection and Recognition Technology of a Visible and Infrared Dim and Small Target Based on Deep Learning

With the increasing trend towards informatization and intelligence in modern warfare, high-intensity and continuous reconnaissance activities are becoming increasingly common in battlefield environments via airborne, vehicle, UAV, satellite and other platforms. Visible and infrared images are preferred due to their high resolution, strong contrast, rich texture details and color features, and strong information expression ability. However, the quality of imaging is easily affected by environmental factors, making it crucial to quickly and accurately filter useful information from massive image data. To this end, super-resolution image preprocessing can improve the detection performance of UAV, and reduce false detection and missed detection of targets. Additionally, super-resolution reconstruction results in high-quality images that can be used to expand UAV datasets and enhance the UAV characteristics, thereby enabling the enhancement of small targets. In response to the challenge of “low-slow small” UAV targets at long distances, we propose a multi-scale fusion super-resolution reconstruction (MFSRCNN) algorithm based on the fast super-resolution reconstruction (FSRCNN) algorithm and multi-scale fusion. Our experiments confirm the feasibility of the algorithm in reconstructing detailed information of the UAV target. On average, the MFSRCNN reconstruction time is 0.028 s, with the average confidence before and after reconstruction being 80.73% and 86.59%, respectively, resulting in an average increase of 6.72%.

Research on Image Super Resolution Reconstruction Model Based on Deep Learning

Artificial neural network has made the rapid development of artificial intelligence with its super-learning ability, making artificial neural network become a research hotspot again. At present, deep learning has been widely used in various fields such as computer vision, speech processing, and natural language processing, and has even played a leading role in some fields. The single image super-resolution reconstruction technique aims to reconstruct a low-resolution image through a series of algorithms to reconstruct a corresponding high-resolution image. This paper first briefly introduces the relevant theories of artificial neural networks, then studies the fast super-resolution reconstruction model, and improves the model layer and filter size to establish a new improved model.

Fast image super-resolution for a dual-resolution camera

High-spatial resolution and wide field of view (FOV) can be satisfied simultaneously with a dual-sensor camera. A special kind of dual-sensor camera named dual-resolution camera has been designed and manufactured; therefore, a high-resolution image with narrow FOV and another low-resolution image with wide FOV are captured by one shot. To generate a high-resolution image with wide FOV, a fast super-resolution reconstruction is proposed, which is composed of wavelet-based super-resolution and back projection. During wavelet-based super-solution, the high-resolution image captured is used to learn the co-occurrence prior by a linear regression function. At last, low-resolution image is reconstructed based on the learnt co-occurrence prior. Simulation and real experiments are carried out, and three other common super-resolution algorithms are compared. The experimental results show that the proposed method reduces time cost significantly, and achieves excellent performance with high PSNR and SSIM.

A fast super-resolution reconstruction from image sequence

Based on the mechanism of imagery, a novel method called the delaminating combining template method, used for the problem of super-resolution reconstruction from image sequence, is described in this paper. The combining template method contains two steps: a delaminating strategy and a combining template algorithm. The delaminating strategy divides the original problem into several sub-problems; each of them is only connected to one degrading factor. The combining template algorithm.is suggested to resolve each sub-problem. In addition, to verify the valid of the method, a new index called oriental entropy is presented. The results from the theoretical analysis and experiments illustrate that this method to be promising and efficient.

A fast super-resolution reconstruction algorithm for pure translational motion and common space-invariant blur

This paper addresses the problem of recovering a super-resolved image from a set of warped blurred and decimated versions thereof. Several algorithms have already been proposed for the solution of this general problem. In this paper, we concentrate on a special case where the warps are pure translations, the blur is space invariant and the same for all the images, and the noise is white. We exploit previous results to develop a new highly efficient super-resolution reconstruction algorithm for this case, which separates the treatment into de-blurring and measurements fusion. The fusion part is shown to be a very simple non-iterative algorithm, preserving the optimality of the entire reconstruction process, in the maximum-likelihood sense. Simulations demonstrate the capabilities of the proposed algorithm.