Across-scale process similarity based interpolation for image super-resolution

Abstract

A pivotal step in image super-resolution techniques is interpolation, which aims at generating high resolution images without introducing artifacts such as blurring and ringing. In this paper, we propose a technique that performs interpolation through an infusion of high frequency signal components computed by exploiting 'process similarity'. By 'process similarity', we refer to the resemblance between a decomposition of the image at a resolution to the decomposition of the image at another resolution. In our approach, the decompositions generating image details and approximations are obtained through the discrete wavelet (DWT) and stationary wavelet (SWT) transforms. The complementary nature of DWT and SWT is leveraged to get the structural relation between the input image and its low resolution approximation. The structural relation is represented by optimal model parameters obtained through particle swarm optimization (PSO). Owing to process similarity, these parameters are used to generate the high resolution output image from the input image. The proposed approach is compared with six existing techniques qualitatively and in terms of PSNR, SSIM, and FSIM measures, along with computation time (CPU time).It is found that our approach is the fastest in terms of CPU time and produces comparable results. Recently, few end to end deep neural networks have been proposed for superresolution. SRCNN [30, 31] uses the convolutional neural network (CNN) [32] for image super-resolution. The technique is later extended to DRCN [9] by increasing the number of layers of the network and incorporating recursive learning.The techniques proposed in [8,11] use residual learning in deep and very deep If the aforesaid wavelet based analysis and synthesis is considered for interpolation, then, the input low resolution image can be decomposed into its immediate lower resolution approximation (LL) using (2) and high frequency details using (3). Now, one can assume that only the lower approximation is available for synthesizing back the input image, and not the high frequency details. In such a case, one can try to estimate the unavailable high frequency

Surveillance Image Super Resolution Using SR - Generative Adversarial Network

“…According to this basic paper [3], interpolation is a critical element in picture super-resolution techniques because it aims to generate high-resolution photos without aberrations such as ringing and blurring. As a result, a method for achieving interpolation by injecting high-frequency signal components predicted using "process similarity" has been proposed.…”

Hyperspectral Image Super Resolution Using Anomaly Weighted Gabor Based CNN

“…In the second category, there is no auxiliary image for hyperspectral super resolution. Different interpolation [9] and learning based approaches [10] belong to the second category. Recently, deep learning methods such as convolutional neural networks (CNNs) [11]- [12], due to their abilities in extraction of robust features invariant to local changes, and autoencoders [13] have shown great success for hyperspectral super resolution due to their abilities in extraction of robust features invariant to local changes.…”

Low-Light Image Enhancement with an Anti-Attention Block-Based Generative Adversarial Network

“…There are three main problems to be solved for low-light image enhancement. First, low-light image enhancement is different from super-resolution [3] or image restoration [4], which has ground truth as a reference comparison. The image should be as close to the ground truth as possible.…”