Source localization in graph information propagation is essential for mitigating network disruptions, including misinformation spread, cyber threats, and infrastructure failures. Existing deep generative approaches face significant challenges in real-world applications due to limited propagation data availability. We present SIDSL (Structure-prior Informed Diffusion model for Source Localization), a generative diffusion framework that leverages topology-aware priors to enable robust source localization with limited data. SIDSL addresses three key challenges: unknown propagation patterns through structure-based source estimations via graph label propagation, complex topology-propagation relationships via a propagation-enhanced conditional denoiser with GNN-parameterized label propagation module, and class imbalance through structure-prior biased diffusion initialization. By learning pattern-invariant features from synthetic data generated by established propagation models, SIDSL enables effective knowledge transfer to real-world scenarios. Experimental evaluation on four real-world datasets demonstrates superior performance with 7.5-13.3% F1 score improvements over baselines, including over 19% improvement in few-shot and 40% in zero-shot settings, validating the framework's effectiveness for practical source localization. Our code can be found here (https://github.com/tsinghua-fib-lab/SIDSL).

Feature representation plays a key role in matching keypoints, especially for the multispectral images of large spectral difference. On such image pairs, existing methods typically use the two images only, but it is challenging to directly learn spectrum-invariant feature representation due to the complex nonlinear distortion between them. To address this issue, this letter proposes using intermediate-band images to facilitate learning spectrum-invariant feature representation. For this purpose, this work designs a spectrum adaptive gate network (SAG-Net) that consists of a SPectral gate (SPeG) module and a deep feature extractor. The SPeG module selectively activates the spectrum-invariant features according to input image content on-the-fly. It hence allows for training on the images of over two bands simultaneously with a single network without the need of an individual branch per band. To investigate the SPeG module, we also constructed a Landsat 9 Multi-Spectral Images (L9-MSI) dataset including 3167 scenes of aligned images across five spectral bands (visible, B5, B6, B7, and B10) from the Landsat 9 imagery. The experimental results demonstrate the SPeG module can learn common feature representation for varying-band images, and the intermediate B5, B6, and B7 images are useful for the SAG-Net to learn the common feature between visible and B10. On the L9-MSI dataset, the SAG-Net significantly improved the number of correct matches and the matching score (MS). Our dataset will be released athttps://github.com/bohanlee/L9MSI-Dataset.git.

Abstract The region proposal network is indispensable to two‐stage object detection methods. It generates a fixed number of proposals that are to be classified and regressed by detection heads to produce detection boxes. However, the fixed number of proposals may be too large when an image contains only a few objects but too small when it contains much more objects. Considering this, the authors explored determining a proper number of proposals according to the number of objects in an image to reduce the computational cost while improving the detection accuracy. Since the number of ground truth objects is unknown at the inference stage, the authors designed a simple but effective module to predict the number of foreground regions, which will be substituted for the number of objects for determining the proposal number. Experimental results of various two‐stage detection methods on different datasets, including MS‐COCO, PASCAL VOC, and CrowdHuman showed that equipping the designed module increased the detection accuracy while decreasing the FLOPs of the detection head. For example, experimental results on the PASCAL VOC dataset showed that applying the designed module to Libra R‐CNN and Grid R‐CNN increased over 1.5 AP 50 while decreasing the FLOPs of detection heads from 28.6 G to nearly 9.0 G.

Generating action proposals in untrimmed videos is a challenging task, since video sequences usually contain lots of irrelevant contents and the duration of an action instance is arbitrary. The quality of action proposals is key to action detection performance. The previous methods mainly rely on sliding windows or anchor boxes to cover all ground-truth actions, but this is infeasible and computationally inefficient. To this end, this article proposes a RecapNet-a novel framework for generating action proposal, by mimicking the human cognitive process of understanding video content. Specifically, this RecapNet includes a residual causal convolution module to build a short memory of the past events, based on which the joint probability actionness density ranking mechanism is designed to retrieve the action proposals. The RecapNet can handle videos with arbitrary length and more important, a video sequence will need to be processed only in one single pass in order to generate all action proposals. The experiments show that the proposed RecapNet outperforms the state of the art under all metrics on the benchmark THUMOS14 and ActivityNet-1.3 datasets. The code is available publicly at https://github.com/tianwangbuaa/RecapNet.

Accurate and robust feature matching is a critical issue in the preprocessing of multispectral image data sets. Higher order features, such as lines can provide useful matching information but are heavily affected by the unreliable detection of lines. Existing methods typically make the unrealistic assumption that end points of lines can be accurately detected across the reference and test images. To address the unreliable detection of line end points, this paper proposes mapping line intersections and then employing tentatively mapped intersections as “anchor” points to compute line descriptors. The computed line descriptors are utilized to determine whether the two lines forming an intersection are matched with the two lines forming its mapped intersection. This eliminates the reliance on the accurate detection of line end points and results in improved matching accuracy. The proposed method is tested on a large number of multispectral images containing various scenes. Experimental results show that it can effectively deal with the detection inaccuracy of end points for line matching.

Low-rank tensor completion methods have been advanced recently for modeling sparsely observed data with a multimode structure. However, low-rank priors may fail to interpret the model factors of general tensor objects. The most common method to address this drawback is to use regularizations together with the low-rank priors. However, due to the complex nature and diverse characteristics of real-world multiway data, the use of a single or a few regularizations remains far from efficient, and there are limited systematic experimental reports on the advantages of these regularizations for tensor completion. To fill these gaps, we propose a modified CP tensor factorization framework that fuses the l2norm constraint, sparseness (l1norm), manifold, and smooth information simultaneously. The factorization problem is addressed through a combination of Nesterov's optimal gradient descent method and block coordinate descent. Here, we construct a smooth approximation to the l1norm and TV norm regularizations, and then, the tensor factor is updated using the projected gradient method, where the step size is determined by the Lipschitz constant. Extensive experiments on simulation data, visual data completion, intelligent transportation systems, and GPS data of user involvement are conducted, and the efficiency of our method is confirmed by the results. Moreover, the obtained results reveal the characteristics of these commonly used regularizations for tensor completion in a certain sense and give experimental guidance concerning how to use them.

Millimeter wave (mmWave) communications are an important candidate technique in 5G networks for features, supporting ultra-dense small cells and mobile data offloading. However, ultra-dense nodes and increasing data traffic bring in vast interference. This paper investigates low complexity non-iterative interference alignment (IA) schemes for multiple-input multiple-output (MIMO) interference broadcast channels in mmWave communications. The authors focus on the three-cell mobile network model in which each base station supports no more than two users within its cell. There is already a closed-form IA solution for the case that one cell has two users, while the other two have one user in each, which can be denoted as {2,1,1}. This paper considers different settings and proposes corresponding IA schemes. First, two IA schemes based on multi-step for the asymmetric setting {2,2,1} are presented, five degree of freedom (DoF) could be achieved. Then, for the symmetric setting {2,2,2}, a novel IA solution with lower complexity and a joint method combining IA with non-iterative multi-user MIMO technique are proposed, and they can achieve six DoF. The simulation results indicate that our non-iterative schemes have similar sum-rate capacity performances with the iterative ones in existing work, and the complexity is effectively reduced.

This paper proposes a line segment-based image registration method. Edges are detected from images by a modified Canny operator, and line segments are then extracted from these edges. At registration, triplets (quaternions) of line segment correspondences are tentatively formed by applying the distance and orientation constraints, which determine an intermediate transformation. Those triplets (quaternions) of lines resulting in higher similarity metrics are preserved, and their intersections are refined by an iterative process or random sample consensus. The proposed method is tested on indoor and outdoor EO/IR image pairs, and the average registration error is calculated to be compared with existing algorithms. Experimental results show that the proposed registration method can robustly align EO/IR images containing line segments, providing more reliable and accurate registration results on multimodal images.

This paper proposes an algorithm of building keypoint matches on multimodal images by combining a bootstrap process and global information. The correct ratio of keypoint matches built with descriptors is typically very low on multimodal images of large spectral difference. To identify correct matches, global information is utilized for evaluating keypoint matches and a bootstrap technique is employed to reduce the computational cost. A keypoint match determines a transformation T and a similarity metric between the reference and the transformed test image by T. The similarity metric encodes global information over entire images, and hence, a higher similarity indicates the match can bring more image content into alignment, implying it tends to be correct. Unfortunately, exhausting triplets/quadruples of matches for affine/projective transformation is computationally intractable, when the number of keypoints is large. To reduce the computational cost, a bootstrap technique is employed that starts from single matches for a translation and rotation model, and goes increasingly to quadruples of four matches for a projective model. The global information screens for "good" matches at each stage and the bootstrap strategy makes the screening process computationally feasible. Experimental results show that the proposed method can establish reliable keypoint matches on challenging multimodal images of strong multimodality.

This letter proposes a multiresolution technique to address the high computational cost in remote sensing image registration. The scale-invariant feature transform is applied to detect keypoints and descriptors, and then, global information combined with descriptors is utilized to establish keypoint mappings. Keypoints are first classified according to their octaves. Then, in the lowest resolution, the keypoints of the largest octave are mapped with descriptors and the global information, giving an initial affine transformation$T_0$. In the next octave, the keypoints of the second largest octave are mapped by employing$T_0$to narrow the space of matching keypoints. By this means, the process of establishing keypoint correspondences is conducted from one resolution (octave) to the next as the obtained transformation gets finer until we get to the highest resolution. Due to the high computational expense of computing global information, the proposed technique is important for aligning large-size remote sensing imagery. Experimental results show that the proposed method can achieve a comparable registration accuracy but with a less computational cost than directly building keypoint mappings on images of large size.

This paper presents a similarity metric on multimodal images utilizing curves as comparing primitives. Curves are detected from images, and then junctions are detected along curves and used to partition curves into subcurves. A modified Hausdorff distance is applied to determine whether a test subcurve is matched to a reference curve. The similarity metric is defined to be the number of matched curves. The number of overlapped edge pixels between two images is also defined on the basis of matched curves, which does not require accurately localizing edge pixels. The partitioning scheme avoids addresing curve partial matching and allows for test subcurves being matched to a reference curve if they correspond to each other. Experimental results show that the presented similarity metric gives more robust and reliable results, especially under noise.

Principal component analysis (PCA) is a technique that is widely used for applications such as dimensionality reduction, image compression, feature extraction and data visualization. One of the key issues in the use of PCA for modelling is that it is very sensitive to outliers since its formulation is based on Gaussian density model. Lately, more heavy-tailed distribution (i.e., Student's t-distribution) is introduced to increase the robustness of traditional PCA. But the robust version of PCA is expressed as the maximum likelihood solution of a probabilistic latent variable model. This reformulation raises the question of how to determine the optimal number of principal components to be retained. In this paper, we develop a Bayesian model selection approach to estimate the true dimensionality of the data. The proposed algorithm is based on a new Bayesian treatment of robust Student's t-distribution PCA. A simple Expectation-Maximization (EM) solver is introduced to find approximate solutions for the model. Experiments show that the proposed model achieves simultaneous optimal dimensionality selection an.d accurate principal components recovery.

Principal component analysis (PCA) is a technique that is widely used for applications such as dimensionality reduction, image compression, feature extraction and data visualization. One of the key issues in the use of PCA for modelling is that it is very sensitive to outliers since its formulation is based on Gaussian density model. Lately, more heavy-tailed distribution (i.e., Student’s t-distribution) is introduced to increase the robustness of traditional PCA. But the robust version of PCA is expressed as the maximum likelihood solution of a probabilistic latent variable model. This reformulation raises the question of how to determine the optimal number of principal components to be retained. In this paper, we develop a Bayesian model selection approach to estimate the true dimensionality of the data. The proposed algorithm is based on a new Bayesian treatment of robust Student’s t-distribution PCA. A simple Expectation-Maximization (EM) solver is introduced to find approximate solutions for the model. Experiments show that the proposed model achieves simultaneous optimal dimensionality selection and accurate principal components recovery.

Augmenting electro-optical (EO) based target tracking systems with infrared (IR) modality has been shown to be effective in increasing the accuracy rate of the tracking system. A key issue in designing such a multimodal tracking system is how to combine information observed from different sensor types in a systematic way to obtain desirable performance. In this paper, we present an investigation into integrating EO and IR sensors within hidden Markov model (HMM) based frameworks. We propose to use a coupled hidden Markov model (CHMM) to improve upon the existing fusion schemes. Another contribution is that we propose to use a robustt-distribution based subspace representation in the CHMM to model appearance changes of the target. Numerical experiments demonstrate that the proposed CHMM tracking system has improved performance over other integration schemes for situations where the target object is corrupted by noise or occlusion.

This paper proposes an image registration method. Edges are detected from images and partitioned into segments as matching primitives. Then, corners on the edges are detected to guide registration. A similarity metric is proposed based on the number of pairs of matching segments. Corner mappings are sequentially tried along a segment, from which a transformation is obtained. The corner mappings are evaluated by the similarity metric under their resulting transformation. By this means, corner mappings are established by utilizing whole images. Since the sensitivity of transformation parameters to the accuracy of corner mappings, as many corner mappings as possible are used. Experimental results show that the proposed method is robust, especially when there is no integral corresponding edges between two images.

This paper proposes an edge-based multimodal image registration approach. It aims to address image registration as a whole rather than tackle each of its elements independently. One-pixel-wide curves are firstly extracted from images, and junctions along the curves are detected. Then, each curve is divided into subsegments as matching primitives. A similarity metric based on the number of matched pairs of subsegments is proposed and experimental results show that the presented approach is a robust and effective tool for multimodal image registration.