Single image super-resolution method via refined local learning

Abstract

Recent years have witnessed the prosperity of referencebased image super-resolution (Ref-SR). By importing the high-resolution (HR) reference images into the single image super-resolution (SISR) approach, the ill-posed nature of this long-standing field has been alleviated with the assistance of texture transferred from reference images. Although the significant improvement in quantitative and qualitative results has verified the superiority of Ref-SR methods, the presence of misalignment before texture transfer indicates room for further performance improvement. Existing methods tend to neglect the significance of details in the context of comparison, therefore not fully leveraging the information contained within low-resolution (LR) images. In this paper, we propose a Detail-Enhancing Framework (DEF) for reference-based super-resolution, which introduces the diffusion model to generate and enhance the underlying detail in LR images. If corresponding parts are present in the reference image, our method can facilitate rigorous alignment. In cases where the reference image lacks corresponding parts, it ensures a fundamental improvement while avoiding the influence of the reference image. Extensive experiments demonstrate that our proposed method achieves superior visual results while maintaining comparable numerical outcomes.

Research on Novel Single Image Super-resolution Algorithm through Regularization Approach and Joint Learning Theory: Theoretical Analysis and Applications

“…Through the analysis on traditional regularization method has many shortcomings, in order to improve the quality of reconstruction image, this paper puts forward the need to rebuild image division, through the local variance to measure the spatial structure of regional information, by setting threshold, the image is divided into the smooth area, edge/texture area and the smooth area [12][13][14]. Locally linear embedding in the manifold learning ideas, put forward the high and low resolution of block manifolds with similar local structure assumption, each test low resolution block using a linear combination of the sample collection of a number of nearest neighbor, through calculate the reconstruction weights and pass a weight to a linear combination of the high resolution sample piece to generate for reconstruction of high resolution image block.…”

Generative Adversarial Network for Image Super-Resolution Combining Texture Loss

“…Yang et al [8] have applied sparse representation theory to ISRR. Tang et al [9] have proposed a refined local learning scheme to reduce the image artifacts and further improve the image visual quality. Similar algorithms for reconstructing images by learning mapping relationships include Bayesian process estimation [10], statistical learning [11] and linear regression representation algorithm [12].…”

Effect of Fly Ash Content on Mechanical Properties of Cement-Fly Ash Stabilized Crushed Stones

“…Cement-fly ash stabilized crushed stone is a new type road base course material in China, which has the advantages of cement stabilized material and lime-fly ash stabilized material [1]. The mechanical prosperities of cement-fly ash stabilized crushed stones, especially the effect of fly ash content, have been a major research topic for these years.…”