Image super-resolution (SR) reconstruction is essentially an ill-posed problem, so it important to design effective prior. For this purpose, we propose a novel image SR method by learning both non-local and local regularization priors from given low-resolution image. The prior takes advantage of the redundancy similar patches in natural images, while assumes that target pixel can be estimated weighted average its neighbors. Based on above considerations, utilize means filter learn steering...

Until now, neighbor-embedding-based (NE) algorithms for super-resolution (SR) have carried out two independent processes to synthesize high-resolution (HR) image patches. In the first process, neighbor search is performed using Euclidean distance metric, and in second optimal weights are determined by solving a constrained least squares problem. However, separate not optimal. this paper, we propose sparse selection scheme SR reconstruction. We predetermine larger number of neighbors as...

The neighbor-embedding (NE) algorithm for single-image super-resolution (SR) reconstruction assumes that the feature spaces of low-resolution (LR) and high-resolution (HR) patches are locally isometric. However, this is not true SR because one-to-many mappings between LR HR patches. To overcome or at least to reduce problem NE-based reconstruction, we apply a joint learning technique train two projection matrices simultaneously map original onto unified subspace. Subsequently, k -nearest...

Example learning-based image super-resolution (SR) is recognized as an effective way to produce a high-resolution (HR) with the help of external training set. The effectiveness SR methods, however, depends highly upon consistency between supporting set and low-resolution (LR) images be handled. To reduce adverse effect brought by incompatible high-frequency details in set, we propose single approach learning multiscale self-similarities from LR itself. proposed based observation that small...

Reconstruction- and example-based super-resolution (SR) methods are promising for restoring a high-resolution (HR) image from low-resolution (LR) image(s). Under large magnification, reconstruction-based usually fail to hallucinate visual details while sometimes introduce unexpected details. Given generic LR image, reconstruct photo-realistic SR suppress artifacts in the reconstructed we multi-scale dictionary novel method that simultaneously integrates local non-local priors. The prior...

Defect detection holds significant importance in improving the overall quality of fabric manufacturing. To improve effectiveness and accuracy defect detection, we propose PRC-Light YOLO model for establish a system. Firstly, have improved YOLOv7 by integrating new convolution operators into Extended-Efficient Layer Aggregation Network optimized feature extraction, reducing computations while capturing spatial features effectively. Secondly, to enhance performance fusion network, use...

Neighbor embedding algorithm has been widely used in example-based super-resolution reconstruction from a single frame, which makes the assumption that neighbor patches embedded are contained manifold. However, it is not always true for complicated texture structure. In this paper, we believe textures may be multiple manifolds, corresponding to classes. Under assumption, present novel image with clustering and supervised (CSNE). First, class predictor low-resolution (LR) learnt by an...

Multiframe super-resolution (SR) reconstruction aims to produce a high-resolution (HR) image using set of low-resolution (LR) images. In the process reconstruction, fuzzy registration usually plays critical role. It mainly focuses on correlation between pixels candidate and reference images reconstruct each pixel by averaging all its neighboring pixels. Therefore, fuzzy-registration-based SR performs well has been widely applied in practice. However, if some objects appear or disappear among...

Wastewater treatment produces a large amount of sludge, where the minimizing disposed sludge is essential for environmental protection. The co-combustion with coal preferable method sewage disposal from economic and perspective. has been widely used in industry advantages processing capacity. melting characteristics ash are an important criterion selection methods furnace types. In this study, two types four different points were selected, behavior upon investigated by experimental...

In this paper, inspired by an inherent characteristic of human visual system capable recognizing salient regions from a complicated scene, we treat defective region as in fabric images. A novel defect detection method, which is based on saliency metric for color dissimilarity and positional aggregation, proposed. the RGB space given image first converted into CIE L*a*b feature representation. Then, distance between similar patches are jointly used to measure values. To improve contrast...

This paper develops a coarse-to-fine framework for single-image super-resolution (SR) reconstruction. The approach achieves high-quality SR recovery based on the complementary properties of both example learning-and reconstruction-based algorithms: learning-based approaches are useful generating plausible details from external exemplars but poor at suppressing aliasing artifacts, while methods propitious preserving sharp edges yet fail to generate fine details. In coarse stage method, we use...

Small targets exist in large numbers various fields. They are broadly used aerospace, video monitoring, and industrial detection. However, because of its tiny dimensions modest resolution, the precision small-target detection is low, erroneous rate high. Therefore, based on YOLOv5, an improved model proposed. First, order to improve number detected while enhancing performance, additional head added. Second, involution between backbone neck increase channel information feature mapping. Third,...

As well known, Gaussian process regression (GPR) has been successfully applied to example learning-based image super-resolution (SR). Despite its effectiveness, the applicability of a GPR model is limited by remarkably computational cost when large number examples are available learning task. For this purpose, we alleviate problem GPR-based SR and propose novel method, called active-sampling (AGPR). The newly proposed approach employs an active strategy heuristically select more informative...

For regression-based single-image super-resolution (SR) problem, the key is to establish a mapping relation between high-resolution (HR) and low-resolution (LR) image patches for obtaining visually pleasing quality image. Most existing approaches typically solve it by dividing model into several single-output regression problems, which obviously ignores circumstance that pixel within an HR patch affects other spatially adjacent pixels during training process, thus tends generate serious...

Considering that manual inspection of the yarn-dyed fabric can be time consuming and inefficient, we propose a defect classification method by using convolutional neural network (CNN) based on modified AlexNet. CNN shows powerful ability in performing feature extraction fusion simulating learning mechanism human brain. The local response normalization layers AlexNet are replaced batch layers, which enhance both computational efficiency accuracy. In training process network, characteristics...