Graph and subspace clustering methods have become the mainstream of multi-view due to their promising performance. However, (1) since graph learn graphs directly from raw data, when data is distorted by noise outliers, performance may seriously decrease; (2) use a "two-step" strategy representation affinity matrix independently, thus fail explore high correlation. To address these issues, we propose novel method via learning <underline xmlns:mml="http://www.w3.org/1998/Math/MathML"...

Low-rank matrix approximation (LRMA)-based methods have made a great success for grayscale image processing. When handling color images, LRMA either restores each channel independently using the monochromatic model or processes concatenation of three channels model. However, these two schemes may not make full use high correlation among RGB channels. To address this issue, we propose novel low-rank quaternion (LRQA) It contains major components: first, instead modeling pixel as scalar in...

Multi-view clustering refers to the task of partitioning numerous unlabeled multimedia data into several distinct clusters using multiple features. In this paper, we propose a novel nonlinear method called joint learning multi-view (JLMVC) jointly learn kernel representation tensor and affinity matrix. The proposed JLMVC has three advantages: (1) unlike existing low-rank representation-based methods that matrix in two separate steps, learns them both; (2) "kernel trick," can handle...

Benefited from quaternion representation that is able to encode the cross-channel correlation of color images, principle component analysis (QPCA) was proposed extract features images while reducing feature dimension. A covariance matrix (QCM) input samples constructed, and its eigenvectors were derived find solution QPCA. However, eigen-decomposition leads fixed for same input. This susceptible outliers cannot be further optimized. To solve this problem, paper proposes a novel ridge...

Superpixel segmentation targets at grouping pixels in an image into atomic regions whose boundaries align well with the natural object boundaries. This paper first proposes a new feature representation for superpixel that holistically embraces color, contour, texture, and spatial features. Then, we introduce clustering-based discriminability measure to iteratively evaluate importance of different Integrating measure, propose novel content-adaptive (CAS) algorithm. CAS is able automatically...

Multiview clustering as an important unsupervised method has been gathering a great deal of attention. However, most multiview methods exploit the self-representation property to capture relationship among data, resulting in high computation cost calculating coefficients. In addition, they usually employ different regularizers learn representation tensor or matrix from which transition probability is constructed separate step, such one proposed by Wu et al.. Thus, optimal cannot be...

To learn the self-representation matrices/tensor that encodes intrinsic structure of data, existing multiview models consider only features and, thus, impose equal membership preference across samples. However, this is inappropriate in real scenarios since prior knowledge, e.g., explicit labels, semantic similarities, and weak-domain cues, can provide useful insights into underlying relationship Based on observation, article proposes a knowledge regularized (P-MVSR) model, which features,...

This paper addresses the multi-view subspace clustering problem and proposes self-paced enhanced low-rank tensor kernelized (SETKMC) method, which is based on two motivations: (1) singular values of representations multiple instances should be treated differently. The reasons are that larger usually quantify major information less penalized; samples with different degrees noise may have various reliability for clustering. (2) many existing methods cause degraded performance when features...

In multi-view subspace clustering, the low-rankness of stacked self-representation tensor is widely accepted to capture high-order cross-view correlation. However, using nuclear norm as a convex surrogate rank function, exhibits strong connectivity with dense coefficients. When noise exists in data, generated affinity matrix may be unreliable for clustering it retains connections across inter-cluster samples due lack sparsity. Since both and sparsity coefficients are curial we propose...

Estimating the time of arrival is a crucial task in intelligent transportation systems. Although considerable efforts have been made to solve this problem, most them decompose trajectory into several segments and then compute travel by integrating attributes from all segments. The segment view, though being able depict local traffic conditions straightforwardly, insufficient embody intrinsic structure trajectories on road network. To overcome limitation, study proposes multi-view...

Recent studies have shown the effectiveness of using depth information in salient object detection. However, most commonly seen images so far are still RGB that do not contain data. Meanwhile, human brain can extract geometric model a scene from an RGB-only image and hence provides 3D perception scene. Inspired by this observation, we propose new concept named RGB-'D' saliency detection, which derives pseudo then performs The be utilized as features, prior knowledge, additional channel, or...

Abstract Hydrogel dressings have significant advantages such as absorption of tissue exudate, maintenance proper moist environment, and promotion cell proliferation. However, facile preparation method high‐efficient antibacterial hydrogel are still a great challenge. In this study, approach to prepare nanocomposite dressing accelerate healing was explored. The hydrogels consisted quaternized chitosan chemically cross‐linked polyacrylamide, well silver nanoparticles (AgNPs) stabilized by...

Purpose Infiltration of activated dendritic cells and inflammatory in cornea represents an important marker for defining corneal inflammation. Deep transfer learning has presented a promising potential is gaining more importance computer assisted diagnosis. This study aimed to develop deep models automatic detection using vivo confocal microscopy images. Methods A total 3453 images was used train the models. External validation performed on independent test set 558 ground-truth label...

As the cornerstone for joint dimension reduction and feature extraction, extensive linear projection algorithms were proposed to fit various requirements. When being applied image data, however, existing methods suffer from representation deficiency since multi-way structure of data is (partially) neglected. To solve this problem, we propose a novel Low-Rank Preserving t-Linear Projection (LRP-tP) model that preserves intrinsic using t-product-based operations. The advances in four aspects:...

Linear discriminant analysis has been incorporated with various representations and measurements for dimension reduction feature extraction. In this paper, we propose two-dimensional quaternion sparse (2D-QSDA) that meets the requirements of representing RGB RGB-D images. 2D-QSDA advances in three aspects: 1) including regularization, relies only on important variables, thus shows good generalization ability to out-of-sample data which are unseen during training phase; 2) benefited from...

Low-rank tensor representation-based multi-view clustering has become an efficient method for data due to the robustness noise and preservation of high order correlation. However, existing algorithms may suffer from two common problems: (1) local view-specific geometrical structures various importance features in different views are neglected; (2) low-rank representation affinity matrix learned separately. To address these issues, we propose a novel framework learn Graph regularized Tensor...

Travel time estimation is a crucial task in practical transportation applications, while providing the reliability of important many working scenarios. Most existing studies do not consider dynamics traffic status for different road segments real time, thus yielding unsatisfactory results. To address problem, we propose to formulate network as temporal attributed graph and perform node representation learning on it. The learned capable jointly exploiting dynamic conditions topology network,...

In real scenarios, graph-based multiview clustering has clearly shown popularity owing to the high efficiency in fusing information from multiple views. Practically, graphs offer both consistent and inconsistent cues as they usually come heterogeneous sources. Previous methods illustrated importance of leveraging consistency inconsistency for accurate modeling. However, when graphs, parts are generally ignored hence valued view-specific attributes lost. To solve this problem, we propose an...

Superpixel segmentation targets at grouping pixels in an image into atomic regions that align well with the natural object boundaries. In this paper, we propose a novel superpixel method based on iterative and adaptive clustering algorithm embraces color, contour, texture, spatial features together. The adjusts weights of different automatically content-aware way, so as to fit requirements various instances. More specifically, each iteration, aggregation function are adjusted according...

Spatial optimization problems (SOPs) refer to a class of where the decision variables require spatial organization. Existing methods based on evolutionary algorithms (EAs) fit conventional operators by flattening representations SOPs one dimension, which hence loses crucial structures potential solutions. To address this issue, article proposes tensorial algorithm (TEA) formulating tensor space SOPs, population candidate solutions is represented as third-order tensor. Accordingly, we design...

Natural image matting has garnered increasing attention in various computer vision applications. The problem aims to find the optimal foreground/background (F/B) color pair for each unknown pixel and thus obtain an alpha matte indicating opacity of foreground object. This is typically modeled as a large-scale combinatorial optimization (PPCO) problem. Heuristic widely employed tackle PPCO owing its gradient-free property promising search ability. However, traditional heuristic methods often...

Subspace learning has been widely applied for joint feature extraction and dimensionality reduction, demonstrating significant efficacy. Numerous subspace methods with diverse assumptions regarding the criteria target subspaces have developed to obtain compact interpretable data representations. However, when image data, existing fail fully exploit inherent correlations within set. This paper proposes a Robust Discriminative t-Linear Learning model (RDtSL) tackle this issue using t-product....