Source Images Research Articles

ASIFusion: An Adaptive Saliency Injection-Based Infrared and Visible Image Fusion Network

The purpose of infrared and visible image fusion (IVIF) is to acquire a more informative fused image by leveraging complementary information, facilitating human perception and machine vision. Among the existing fusion methods, the saliency-based methods conform to human perception characteristics and achieve relatively advantageous fusion performance. However, such methods fail to adaptively maintain the edge and intensity of salient objects, resulting in fixed fusion performance. To address these issue, we present ASIFusion, an adaptive saliency injection-based IVIF network. First, source images are inputted to the feature extraction encoder for fully extracting features. Meanwhile, the proposed adaptive saliency injection module detects salient objects in the infrared image and then learns the fusion weights of each channel, which serve as supplementary information for further fusion. These learned weights are utilized to merge the source images’ extracted features. Finally, the feature reconstruction decoder produces a fused image with injected saliency. The fused image maintains the intensity and edge of the salient objects and fully preserves the complementary information. Extensive experiments demonstrate that our proposed network outperforms state-of-the-art (SOTA) approaches with regard to fusion performance and computational efficiency.

Improving Compton camera imaging of multi-energy radioactive sources by using machine learning algorithms for event selection

Event selection and background reduction for Compton camera imaging of multi-energy radioactive sources has been performed by employing neural networks. A Compton camera prototype with detectors made of LaBr3 crystals coupled to silicon photomultiplier arrays was used to acquire experimental data from a circular array of 22Na sources. The prototype and two arrays of 22Na sources were simulated with GATE v8.2 Monte Carlo code, to obtain data for neural network training. Neural network models were trained on simulated data for event classification. The optimum models were found by using Weights & Biases platform tools. The trained models were used to classify simulated and real data for selecting signal events and rejecting background prior to image reconstruction. The models performed well on simulated data. The image obtained with experimental data showed an improvement with respect to event selection with energy cuts. The method is promising for Compton camera imaging of multi-energy radioactive sources.

Electromagnetic Source Imaging in Presurgical Evaluation of Children with Drug-Resistant Epilepsy.

For children with drug-resistant epilepsy (DRE), seizure freedom relies on the delineation and resection (or ablation/disconnection) of the epileptogenic zone (EZ) while preserving the eloquent brain areas. The development of a reliable and noninvasive localization method that provides clinically useful information for the localization of the EZ is, therefore, crucial to achieving successful surgical outcomes. Electric and magnetic source imaging (ESI and MSI) have been increasingly utilized in the presurgical evaluation of these patients showing promising findings in the delineation of epileptogenic as well as eloquent brain areas. Moreover, the combination of ESI and MSI into a single solution, namely electromagnetic source imaging (EMSI), performed on simultaneous high-density electroencephalography (HD-EEG) and magnetoencephalography (MEG) recordings has shown higher source localization accuracy than either modality alone. Despite these encouraging findings, such techniques are performed in only a few tertiary epilepsy centers, are rarely recorded simultaneously, and are underutilized in pediatric cohorts. This study illustrates the experimental setup for recording simultaneous MEG and HD-EEG data as well as the methodological framework for analyzing these data aiming to localize the irritative zone, the seizure onset zone, and eloquent brain areas in children with DRE. More specifically, the experimental setups are presented for (i) recording and localizing interictal and ictal epileptiform activity during sleep and (ii) recording visual-, motor-, auditory-, and somatosensory-evoked responses and mapping relevant eloquent brain areas (i.e., visual, motor, auditory, and somatosensory) during visuomotor task, as well as auditory and somatosensory stimulations. Detailed steps of the data analysis pipeline are further presented for performing EMSI as well as individual ESI and MSI using equivalent current dipole (ECD) and dynamic statistical parametric mapping (dSPM).

Improved multi-focus image fusion using online convolutional sparse coding based on sample-dependent dictionary

Multi-focus image fusion merges multiple images captured from different focused regions of a scene to create a fully-focused image. Convolutional sparse coding (CSC) methods are commonly employed for accurate extraction of focused regions, but they often disregard computational costs. To overcome this, an online convolutional sparse coding (OCSC) technique was introduced, but its performance is still limited by the number of filters used, affecting overall performance negatively. To address these limitations, a novel approach called Sample-Dependent Dictionary-based Online Convolutional Sparse Coding (SCSC) was proposed. SCSC enables the utilization of additional filters while maintaining low time and space complexity for processing high-dimensional or large data. Leveraging the computational efficiency and effective global feature extraction of SCSC, we propose a novel method for multi-focus image fusion. Our method involves a two-layer decomposition of each source image, yielding a base layer capturing the predominant features and a detail layer containing finer details. The amalgamation of the fused base and detail layers culminates in the reconstruction of the final image. The proposed method significantly mitigates artifacts, preserves fine details at the focus boundary, and demonstrates notable enhancements in both visual quality and objective evaluation of multi-focus image fusion.

Coherent fading suppression of distributed acoustic sensing based on time–frequency domain linkage algorithm

Abstract Due to the inherent characteristic of coherent fading phenomenon, the performance of distributed acoustic sensing system will be severely degraded, producing anomalous information and positioning errors. In order to suppress the influence of this phenomenon, the data array of demodulated results is obtained after processing the collected distributed Rayleigh scattering signals along the sensing fiber, which can be utilized to construct a two-dimension (2D) image with the time–distance domain information. Thus, this allows us to remove the abnormal information that are introduced by fading noise. Based on this principle, a morphology compositional analysis (MCA) scheme, an effective time–frequency domain linkage algorithm, is applied to the constructed 2D images to enhance the sensing accuracy. Compared with traditional methods including the 2D mean value filter (MVF) and 2D wavelet transform (WT), the proposed method makes full use of the independence of image source property and abnormity feature so that the abnormal information caused by the coherent fading can be eliminated without deteriorating the source information along the sensing fiber. Experimental results show that the anomaly rate of the time–distance domain information along the sensing fiber for the proposed method, the MVF method and the WT method can be reduced by 92.43%, 65.05% and 70.18% respectively. In addition, the isolation degree of the MCA, MVF and WT methods are 41, 16.02 and 23.35 times higher than that treated by traditional method, respectively. Also, the stability of the time-domain vibration can be improved and the spatial resolution would not be influenced by the proposed method compared with the traditional method. This scheme will promote the development of distributed acoustic sensor system with high-accuracy and high-quality without any hardware modification.

Evaluation of urban drainage capacity using different satellite image sources and SWMM

ABSTRACT This research aims to cost-effectively identify inundation due to combined rainfall with tidal surges by evaluating the existing urban drainage system. The Storm Water Management Model (SWMM) applied two scenarios based on different Digital Elevation Model (DEM) sources. These are scenario A, i.e. Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), and scenario B, i.e. Shuttle Radar Topography Mission (SRTM). Then, catchment delineation and SWMM parameterization in the low-lying urban region of Chattogram City were executed. In contrast to scenario A, i.e. ASTER DEM, identified inundated nodes in the northwest part of the city, scenario B, i.e. SRTM DEM, showed more inundated nodes. In both scenarios, drainage capacity was observed in the field at two outlet points of Ispahani Khal (R2 = 0.91 and 0.85) and Chaktai Khal (R2 = 0.90 and 0.92), which reasonably mimics the model outcome. With an intensive field dataset, this model could provide relevant information for betterment.

MeST-Former: Motion-enhanced Spatiotemporal Transformer for generalizable Deepfake detection

The rise of Deepfake technology has sparked significant concerns due to its potential for misuse and malicious manipulation of multimedia content. Various detection approaches aimed at detecting Deepfake videos have been proposed, mostly relying on the identification of spatial and temporal artifacts. However, due to the different contexts of source images and the variety of generation techniques, current Deepfake detection methods usually perform well on training datasets, and yet generalize poorly to those unseen identities in new datasets. This issue is widely known as the generalization challenge of Deepfake detection. To address this challenge, this paper proposes an advanced spatiotemporal Deepfake video detector, named Motion-enhanced Spatiotemporal Transformer (MeST-Former). MeST-Former is based on the spatiotemporal modeling capacity of the video Swin Transformer. The spatial and temporal features are obtained from the RGB and motion images, respectively. To enhance the generalization ability of MeST-Former to unseen identities in unseen datasets, the ID-related components in the spatial and temporal features are detached. Specifically, MeST-Former adopts the newly proposed Identity-Decoupling Attention (IDC-Att) module to disentangle the ID-related and ID-unrelated components. Only the ID-unrelated components are used to construct more generalizable spatiotemporal representations. This process makes the constructed spatiotemporal features identity-agnostic and more generalizable to unseen identities. We conducted extensive experiments to evaluate the performance of the MeST-Former. Our results indicate that MeST-Former achieves accurate and generalizable Deepfake detection performance. Notably, MeST-Former also demonstrates high efficacy in detecting AI-animated talking-head videos.

Reference-then-supervision framework for infrared and visible image fusion

Infrared and visible image fusion has drawn increasing attention of researchers in recent years, wherein the complementary information between two source images is extracted to synthesize a new fusion image with richer information. Deep neural network is the latest technique to tackle various image fusion problems. However, the natural absence of ground truth makes it very challenging to optimize deep learning models. While the current methods mainly consider to uses the two source images themselves or their visual features, to provide supervision for learning, which is easy to result in the imbalance between detail preservation and brightness distribution. To boost the performance of the fusion models, it is meaningful and necessary to establish a flexible framework that can combine the advantages of existing models to produce reliable supervision for model training. In this work, we propose a novel reference-then-supervision framework, which aims to fully leverage and exploit the available favorable reference information based on the performance of existing methods and then construct high-quality reliable supervision to assist in model building. For this purpose, we design an automatic filter to produce favorable reference and devise an adaptive enhancement method to construct reliable supervision, which helps to aggregate the advantages of various existing fusion methods for yielding visually pleasing results adapting to different complex scenarios. Extensive experiments on two commonly used datasets and our built challenging test set demonstrate that our framework can greatly improve the performance of existing fusion methods. Ablation study and empirical analysis also present the efficacy of our framework design. Furthermore, the applications on downstream pedestrian detection and object tracking tasks indicate the great potential of our framework. Our code and data are publicly available at https://github.com/zhenglab/ReferenceSupervisionIVIF.

MSI-DTrans: A multi-focus image fusion using multilayer semantic interaction and dynamic transformer

Multi-focus image fusion (MFIF) aims to utilize multiple images with different focal lengths to fuse into a single full-focus image. This process enhances the realism and clarity of the resulting image. In this paper, a MFIF method called MSI-DTrans was proposed. On the one hand, in order to fully utilize all the effective information that the source image carries, the proposed method adopts a multilayer semantic interaction strategy to enhance the interaction of high-frequency and low-frequency information. This approach gradually mines more abstract semantic information, guiding the generation of feature maps from coarse to fine. On the other hand, a parallel multi-scale joint self-attention computation model is designed. The model adopts dynamic sense field and dynamic token embedding to overcome the performance degradation problem when dealing with multi-scale objects. This enables self-attention to integrate long-range dependencies between objects of different scales and reduces computational overhead. Numerous experimental results show that the proposed method effectively avoids image distortion, achieves better visualization results, and demonstrates good competitiveness with many state-of-the-art methods in terms of qualitative and quantitative analysis, as well as efficiency comparison. The source code is available at https://github.com/ouyangbaicai/MSI-DTrans.

Source imaging method based on diagonal covariance bases and its applications to OPM-MEG

Magnetoencephalography (MEG) is a noninvasive imaging technique used in neuroscience and clinical research. The source estimation of MEG involves solving a highly underdetermined inverse problem, which requires additional constraints to restrict the solution space. Traditional methods tend to obscure the extent of the sources. However, an accurate estimation of the source extent is important for studying brain activity or preoperatively estimating pathogenic regions. To improve the estimation accuracy of the extended source extent, the spatial constraint of sources is employed in the Bayesian framework. For example, the source is decomposed into a linear combination of validated spatial basis functions, which is proved to improve the source imaging accuracy. In this work, we further construct the spatial properties of the source using the diagonal covariance bases (DCB), which we summarize as the source imaging method SI-DCB. In this approach, specifically, the covariance matrix of the spatial coefficients is modeled as a weighted combination of diagonal covariance basis functions. The convex analysis is used to estimate noise and model parameters under the Bayesian framework. Extensive numerical simulations showed that SI-DCB outperformed five benchmark methods in accurately estimating the location and extent of patch sources. The effectiveness of SI-DCB was verified through somatosensory stimulation experiments performed on a 31-channel OPM-MEG system. The SI-DCB correctly identified the source area where each brain response occurred. The superior performance of SI-DCB suggests that it can provide a template approach for improving the accuracy of source extent estimations under a sparse Bayesian framework.

DCFusion: Difference correlation-driven fusion mechanism of infrared and visible images

In end-to-end image fusion models, the loss function significantly impacts performance. However, most loss functions treat salient and background regions in source images equally, failing to distinguish complementary areas in multimodal images. This limits the model’s ability to effectively integrate information from these regions. Therefore, we propose difference correlation-driven fusion mechanism of infrared and visible images, which called DCFusion. Specifically, the model utilizes a dual-branch interactive network that dynamically fuses cross-modal multi-scale complementary information through element-wise multiplication, effectively integrating region-specific information. We introduce a two-stage method for generating salient target masks that adaptively focus on high-contrast regions in infrared images by analyzing pixel contrasts in local areas. Furthermore, we utilize the salient target masks to create heterogeneous images and design the LSCD loss function to minimize the information gap between the heterogeneous images and the fused image, thereby enhancing the model’s interpretability. Experiments on the RoadScene and TNO datasets show that DCFusion surpasses with existing representativity fusion approaches, achieving state-of-the-art performance in both subjective visual and objective evaluations. Our code will be publicly available at https://github.com/MinLila/DCFusion.

HBANet: A hybrid boundary-aware attention network for infrared and visible image fusion

Infrared and visible image fusion is an extensively investigated problem in infrared image processing, aiming to extract useful information from source images. However, the automatic fusion of these images presents a significant challenge due to the large domain difference and ambiguous boundaries. In this article, we propose a novel image fusion approach based on hybrid boundary-aware attention, termed HBANet, which models global dependencies across the image and leverages boundary-wise prior knowledge to supplement local details. Specifically, we design a novel mixed boundary-aware attention module that is capable of leveraging spatial information to the fullest extent and integrating long dependencies across different domains. To preserve the integrity of texture and structural information, we introduced a sophisticated loss function that comprises structure, intensity, and variation losses. Our method has been demonstrated to outperform state-of-the-art methods in terms of both visual and quantitative metrics, in our experiments on public datasets. Furthermore, our approach also exhibits great generalization capability, achieving satisfactory results in CT and MRI image fusion tasks.

RT-SNDETR: real-time supernova detection via end-to-end image transformers

ABSTRACT In large-scale astronomical surveys, traditional supernova detection pipelines rely on complex and relatively inefficient image differencing techniques. This paper proposes an end-to-end deep-learning supernova detection network, the Real-Time SuperNova DEtection TRansformer (RT-SNDETR). This network partially replaces traditional pipelines by integrating image differencing, source detection, and Real-bogus classification, achieving a speed 51.49 times that of the fastest image differencing method, SFFT. Additionally, it remains competitive with methods like YOLO v8, offering a well-balanced trade-off between speed and accuracy. Experimental results highlight RT-SNDETR’s superior performance, with an average precision(AP) of 96.30 per cent on synthetic samples and 76.60 per cent on real supernova samples. It significantly outperforms other detection networks, including RT-DETR (+5.6 per cent AP on synthetic/+5.1 per cent AP on real samples) and Cascade R-CNN (+8.9 per cent AP on synthetic/ +28.6 per cent AP on real samples). The incorporation of CycleGAN-based data generation methods plays a significant role in enhancing RT-SNDETR’s performance. These methods simulate realistic PSF variations, enabling the object detection network to learn more robust features and improving its generalization to real supernovae data. Additionally, by integrating unsupervised domain adaptation techniques, RT-SNDETR achieves an AP of 81.70 per cent on real SDSS supernova survey samples. This study demonstrates RT-SNDETR’s potential to significantly enhance both the speed and accuracy of supernova detection, making it a highly effective solution for large-scale astronomical surveys.

Advancing Nighttime Object Detection through Image Enhancement and Domain Adaptation

Due to the lack of annotations for nighttime low-light images, object detection in low-light images has always been a challenging problem. Achieving high-precision results at night is also an issue. Additionally, we aim to use a single nighttime dataset to complete the knowledge distillation task while improving the detection accuracy of object detection models under nighttime low-light conditions and reducing the computational cost of the model, especially for small targets and objects contaminated by special nighttime lighting. This paper proposes a Nighttime Unsupervised Domain Adaptation Network (NUDN) based on knowledge distillation to address these issues. To improve the detection accuracy of nighttime images, high-confidence bounding box predictions from the teacher and region proposals from the student are first fused, allowing the teacher to perform better in subsequent training, thus generating a combination of high-confidence and low-confidence pseudo-labels. This combination of feature information is used to guide model training, enabling the model to extract feature information similar to that of source images in nighttime low-light images. Nighttime images and pseudo-labels undergo random size transformations before being used as input for the student, enhancing the model’s generalization across different scales. To address the scarcity of nighttime datasets, we propose a nighttime-specific augmentation pipeline called LightImg. This pipeline enhances nighttime features, transforming them into daytime features and reducing issues such as backlighting, uneven illumination, and dim nighttime light, enabling cross-domain research using existing nighttime datasets. Our experimental results show that NUDN can significantly improve nighttime low-light object detection accuracy on the SHIFT and ExDark datasets. We conduct extensive experiments and ablation studies to demonstrate the effectiveness and efficiency of our work.

Ultrashort-Echo-Time MRI of the Disco-Vertebral Junction: Modulation of Image Contrast via Echo Subtraction and Echo Times.

The disco-vertebral junction (DVJ) of the lumbar spine contains thin structures with short T2 values, including the cartilaginous endplate (CEP) sandwiched between the bony vertebral endplate (VEP) and the nucleus pulposus (NP). We previously demonstrated that ultrashort-echo-time (UTE) MRI, compared to conventional MRI, is able to depict the tissues at the DVJ with improved contrast. In this study, we sought to further optimize UTE MRI by characterizing the contrast-to-noise ratio (CNR) of these tissues when either single echo or echo subtraction images are used and with varying echo times (TEs). In four cadaveric lumbar spines, we acquired 3D Cones (a UTE sequence) images at varying TEs from 0.032 ms to 16 ms. Additionally, spin echo T1- and T2-weighted images were acquired. The CNRs of CEP-NP and CEP-VEP were measured in all source images and 3D Cones echo subtraction images. In the spin echo images, it was challenging to distinguish the CEP from the VEP, as both had low signal intensity. However, the 3D Cones source images at the shortest TE of 0.032 ms provided an excellent contrast between the CEP and the VEP. As the TE increased, the contrast decreased in the source images. In contrast, the 3D Cones echo subtraction images showed increasing CNR values as the second TE increased, reaching statistical significance when the second TE was above 10 ms (p < 0.05). Our study highlights the feasibility of incorporating UTE MRI for the evaluation of the DVJ and its advantages over conventional spin echo sequences for improving the contrast between the CEP and adjacent tissues. Additionally, modulation of the contrast for the target tissues can be achieved using either source images or subtraction images, as well as by varying the echo times.

MEEAFusion: Multi-Scale Edge Enhancement and Joint Attention Mechanism Based Infrared and Visible Image Fusion.

Infrared and visible image fusion can integrate rich edge details and salient infrared targets, resulting in high-quality images suitable for advanced tasks. However, most available algorithms struggle to fully extract detailed features and overlook the interaction of complementary features across different modal images during the feature fusion process. To address this gap, this study presents a novel fusion method based on multi-scale edge enhancement and a joint attention mechanism (MEEAFusion). Initially, convolution kernels of varying scales were utilized to obtain shallow features with multiple receptive fields unique to the source image. Subsequently, a multi-scale gradient residual block (MGRB) was developed to capture the high-level semantic information and low-level edge texture information of the image, enhancing the representation of fine-grained features. Then, the complementary feature between infrared and visible images was defined, and a cross-transfer attention fusion block (CAFB) was devised with joint spatial attention and channel attention to refine the critical supplemental information. This allowed the network to obtain fused features that were rich in both common and complementary information, thus realizing feature interaction and pre-fusion. Lastly, the features were reconstructed to obtain the fused image. Extensive experiments on three benchmark datasets demonstrated that the MEEAFusion proposed in this research has considerable strengths in terms of rich texture details, significant infrared targets, and distinct edge contours, and it achieves superior fusion performance.

Improved Dipole Source Localization from Simultaneous MEG-EEG Data by Combining a Global Optimization Algorithm with a Local Parameter Search: A Brain Phantom Study.

Dipole localization, a fundamental challenge in electromagnetic source imaging, inherently constitutes an optimization problem aimed at solving the inverse problem of electric current source estimation within the human brain. The accuracy of dipole localization algorithms is contingent upon the complexity of the forward model, often referred to as the head model, and the signal-to-noise ratio (SNR) of measurements. In scenarios characterized by low SNR, often corresponding to deep-seated sources, existing optimization techniques struggle to converge to global minima, thereby leading to the localization of dipoles at erroneous positions, far from their true locations. This study presents a novel hybrid algorithm that combines simulated annealing with the traditional quasi-Newton optimization method, tailored to address the inherent limitations of dipole localization under low-SNR conditions. Using a realistic head model for both electroencephalography (EEG) and magnetoencephalography (MEG), it is demonstrated that this novel hybrid algorithm enables significant improvements of up to 45% in dipole localization accuracy compared to the often-used dipole scanning and gradient descent techniques. Localization improvements are not only found for single dipoles but also in two-dipole-source scenarios, where sources are proximal to each other. The novel methodology presented in this work could be useful in various applications of clinical neuroimaging, particularly in cases where recordings are noisy or sources are located deep within the brain.

Multi-Source Image Fusion Based on BEMD and Region Sharpness Guidance Region Overlapping Algorithm

Multi-focal image and multi-modal image fusion technology can fully take advantage of different sensors or different times, retaining the image feature information and improving the image quality. A multi-source image fusion algorithm based on bidimensional empirical mode decomposition (BEMD) and a region sharpness-guided region overlapping algorithm are studied in this article. Firstly, source images are decomposed into multi-layer bidimensional intrinsic mode functions (BIMFs) and residuals from high-frequency layer to low-frequency layer by BEMD. Gaussian bidimensional intrinsic mode functions (GBIMFs) are obtained by applying Gaussian filtering operated on BIMF and calculating the sharpness value of segmented regions using an improved weighted operator based on the Tenengrad function, which is the key to comparison selection and fusion. Then, the GBIMFs and residuals selected by sharpness comparison strategy are fused by the region overlapping method, and the stacked layers are weighted to construct the final fusion image. Finally, based on qualitative evaluation and quantitative evaluation indicators, the proposed algorithm is compared with six typical image fusion algorithms. The comparison results show that the proposed algorithm can effectively capture the feature information of images in different states and reduce the redundant information.

CT and MRI Image Fusion via Coupled Feature-Learning GAN

The fusion of multimodal medical images, particularly CT and MRI, is driven by the need to enhance the diagnostic process by providing clinicians with a single, comprehensive image that encapsulates all necessary details. Existing fusion methods often exhibit a bias towards features from one of the source images, making it challenging to simultaneously preserve both structural information and textural details. Designing an effective fusion method that can preserve more discriminative information is therefore crucial. In this work, we propose a Coupled Feature-Learning GAN (CFGAN) to fuse the multimodal medical images into a single informative image. The proposed method establishes an adversarial game between the discriminators and a couple of generators. First, the coupled generators are trained to generate two real-like fused images, which are then used to deceive the two coupled discriminators. Subsequently, the two discriminators are devised to minimize the structural distance to ensure the abundant information in the original source images is well-maintained in the fused image. We further empower the generators to be robust under various scales by constructing a discriminative feature extraction (DFE) block with different dilation rates. Moreover, we introduce a cross-dimension interaction attention (CIA) block to refine the feature representations. The qualitative and quantitative experiments on common benchmarks demonstrate the competitive performance of the CFGAN compared to other state-of-the-art methods.

GAN-HA: A generative adversarial network with a novel heterogeneous dual-discriminator network and a new attention-based fusion strategy for infrared and visible image fusion

Infrared and visible image fusion (IVIF) aims to preserve thermal radiation information from infrared images while integrating texture details from visible images. Thermal radiation information is mainly expressed through image intensities, while texture details are typically expressed through image gradients. However, existing dual-discriminator generative adversarial networks (GANs) often rely on two structurally identical discriminators for learning, which do not fully account for the distinct learning needs of infrared and visible image information. To this end, this paper proposes a novel GAN with a heterogeneous dual-discriminator network and an attention-based fusion strategy (GAN-HA). Specifically, recognizing the intrinsic differences between infrared and visible images, we propose, for the first time, a novel heterogeneous dual-discriminator network to simultaneously capture thermal radiation information and texture details. The two discriminators in this network are structurally different, including a salient discriminator for infrared images and a detailed discriminator for visible images. They are able to learn rich image intensity information and image gradient information, respectively. In addition, a new attention-based fusion strategy is designed in the generator to appropriately emphasize the learned information from different source images, thereby improving the information representation ability of the fusion result. In this way, the fused images generated by GAN-HA can more effectively maintain both the salience of thermal targets and the sharpness of textures. Extensive experiments on various public datasets demonstrate the superiority of GAN-HA over other state-of-the-art (SOTA) algorithms while showcasing its higher potential for practical applications.