Heavy-Tailed Self-Similarity Modeling for Single Image Super Resolution

An alternating multiple residual Wasserstein regularization model for Gaussian image denoising

Residual histograms can provide meaningful statistical information in low-level visual research. However, the existing image denoising methods do not deeply explore the potential of alternate multiple residual histograms for overall optimization constraints. Considering this deficiency, this paper presents a novel unified framework of the alternating multiple residual Wasserstein regularization model (AMRW), which can tactfully embrace multiple residual Wasserstein constraints and different image prior information for image denoising. Specifically, AMRW focuses on solving the practical and meaningful problem of restoring a clean image from multiple frame degraded images. Utilizing the Wasserstein distance in the optimal transport theory, the residual histograms of the multiple degraded images are as close as possible to the referenced Gaussian noise histogram to enhance the noise estimation accuracy. Further, the proposed concrete AMRW combines the triple residual Wasserstein distance with the image total variation prior information for Gaussian image denoising. More importantly, through the alternating implementation of residual Wasserstein regularization from different image frames, the beneficial information of the image is essentially transmitted in each cycle, continuously improving the quality of the output image. Synchronously, the alternate iterative algorithm of histogram matching and Chambolle dual projection has high implementation efficiency. AMRW provides a new research idea for other visual processing tasks such as image inpainting and image deblurring. Finally, extensive numerical experiments substantiate that our AMRW can greatly boost the subjective and objective performance of the restored images compared with some popular image denoising algorithms in recent years.

RefQSR: Reference-Based Quantization for Image Super-Resolution Networks

Single image super-resolution (SISR) aims to reconstruct a high-resolution image from its low-resolution observation. Recent deep learning-based SISR models show high performance at the expense of increased computational costs, limiting their use in resource-constrained environments. As a promising solution for computationally efficient network design, network quantization has been extensively studied. However, existing quantization methods developed for SISR have yet to effectively exploit image self-similarity, which is a new direction for exploration in this study. We introduce a novel method called reference-based quantization for image super-resolution (RefQSR) that applies high-bit quantization to several representative patches and uses them as references for low-bit quantization of the rest of the patches in an image. To this end, we design dedicated patch clustering and reference-based quantization modules and integrate them into existing SISR network quantization methods. The experimental results demonstrate the effectiveness of RefQSR on various SISR networks and quantization methods.

Tchebichef Transform Domain-Based Deep Learning Architecture for Image Super-Resolution

Recent advances in the area of artificial intelligence and deep learning have motivated researchers to apply this knowledge to solve multipurpose applications in the area of computer vision and image processing. Super-resolution (SR), in the past few years, has produced remarkable results using deep learning methods. The ability of deep learning methods to learn the nonlinear mapping from low-resolution (LR) images to their corresponding high-resolution (HR) images leads to compelling results for SR in diverse areas of research. In this article, we propose a deep learning-based image SR architecture in the Tchebichef transform domain. This is achieved by integrating a transform layer into the proposed architecture through a customized Tchebichef convolutional layer (TCL). The role of TCL is to convert the LR image from the spatial domain to the orthogonal transform domain using Tchebichef basis functions. The inversion of the transform mentioned earlier is achieved using another layer known as the inverse TCL (ITCL), which converts back the LR images from the transform domain to the spatial domain. It has been observed that using the Tchebichef transform domain for the task of SR takes the advantage of high and low- frequency representation of images that makes the task of SR simplified. Furthermore, a transfer learning-based approach is adopted to enhance the quality of images by considering Covid19 medical images as an additional experiment. It is shown that our architecture enhances the quality of X-ray and CT images of COVID-19, providing a better image quality that may help in clinical diagnosis. Experimental results obtained using the proposed Tchebichef transform domain SR (TTDSR) architecture provides competitive results when compared with most of the deep learning methods employed using a fewer number of trainable parameters.