Head avatar generation is facilitated to construct high-fidelity 3D virtual personas from a single portrait, but it also raises the risk of unauthorized personal avatars generation. Recent 2D portrait protection methods actively prevent malicious image generation by perturbing the identity features. However, there are two key limitations when directly applied to prevent 3D head avatar generation: 1) These methods neglect the inherent 3D geometric structure of portrait, thus failing to disrupt the modeling of 3D shapes or poses. 2) They focus only on identity offset and are unable to interfere with the overall appearance, resulting in excessive preservation of facial characteristics. To overcomes these limitations, we propose a 3D defense framework termed Anti-Avatar, tailored to protect against unauthorized 3D head avatar generation from a single portrait. Specifically, Anti-Avatar consists of two key designs: Geometric Disruption and Perceptual Confusion. The former disrupts the precise reconstruction of 3D structure by interfering with the estimation of geometric parameters, thus affecting the structural accuracy of the 3D avatar. Collaboratively, the latter confuses image features by dispersing attention distribution, thereby hindering the effective perception of portrait appearance. Benefiting from the above dual-space divergence in geometry and perception, the avatars generated by our protected portraits exhibit substantial discrepancies from the originals. Extensive experiments show that our Anti-Avatar outperforms 2D methods in protection performance and effectively resists reconstruction and manipulation by state-of-the-art 3D head avatar generation methods.

Neural Radiance Fields (NeRF) has achieved remarkable success in synthesizing impressive novel views. However, existing methods usually fail to handle scenes with adverse lighting conditions caused by external time variations and different camera settings, leading to poor visual quality. To address this challenge, we propose a physical-aware NeRF for efficient exposure correction, named PHY-NeRF. Specifically, we design Adaptive Lighting Particles inspired by the theory of light scattering and absorption, which can adjust the illumination intensity during volume rendering. Subsequently, we can handle scenes with different lighting conditions by jointly optimizing camera parameters and these lighting particles. Moreover, to promote natural brightness transitions, we devise a global illumination consistency module to control the lighting intensity across views at the feature level while completing more details. Benefiting from the above designs, our PHY-NeRF can tackle arbitrary low-light or overexposed scenes in an unsupervised manner. Extensive experiments show that our PHY-NeRF achieves state-of-the-art results in addressing adverse lighting problems while ensuring high rendering efficiency.

3D Gaussian Splatting (3DGS) has recently achieved remarkable progress in novel view synthesis. However, existing methods rely heavily on high-quality data for rendering and struggle to handle degraded scenes with multi-view inconsistency, leading to inferior rendering quality. To address this challenge, we propose a novel Depth-aware Gaussian Splatting for efficient 3D scene Restoration, called RestorGS, which flexibly restores multiple degraded scenes using a unified framework. Specifically, RestorGS consists of two core designs: Appearance Decoupling and Depth-Guided Modeling. The former exploits appearance learning over spherical harmonics to decouple clear and degraded Gaussian, thus separating the clear views from the degraded ones. Collaboratively, the latter leverages the depth information to guide the degradation modeling, thereby facilitating the decoupling process. Benefiting from the above optimization strategy, our method achieves high-quality restoration while enabling real-time rendering speed. Extensive experiments show that our RestorGS outperforms existing methods significantly in underwater, nighttime, and hazy scenes.

3D editing with Gaussian splatting is exciting in creating realistic content, but it also poses abuse risks for generating malicious 3D content. Existing 2D defense approaches mainly focus on adding perturbations to single image to resist malicious image editing. However, there remain two limitations when applied directly to 3D scenes: (1) These methods fail to reflect 3D spatial correlations, thus protecting ineffectively under multiple viewpoints. (2) Such pixel-level perturbation is easily eliminated during the iterations of 3D editing, leading to failure of protection. To address the above issues, we propose a novel Defense framework against malicious 3D Editing for Gaussian splatting (DEGauss) for robustly disrupting the trajectory of 3D editing in multi-views. Specifically, to enable the effectiveness of perturbation across various views, we devise a view-focal gradient fusion mechanism that dynamically emphasizes the contributions of the most challenging views to adaptively optimize 3D perturbations. Furthermore, we design a dual discrepancy optimization strategy that both maximize the semantic deviation and the edit direction deviation of the guidance conditions to stably disrupt the editing trajectory. Benefiting from the collaborative designs, our method achieves effective resistance to 3D editing from various views while preserving photorealistic rendering quality. Extensive experiments demonstrate that our DEGauss not only performs excellent defense in different scenes, but also exhibits strong generalization across various state-of-the-art 3D editing pipelines.

Since the training data of the target model is not available in the black-box substitute attack, most recent schemes utilize generative adversarial networks (GANs) to generate data for training the substitute model. However, these GANs-based schemes suffer from low training efficiency as the generator needs to be retrained for each target model during the substitute training process, as well as low generation quality. To overcome these limitations, we consider utilizing the diffusion model (DM) to generate data and propose a novel data-free substitute attack scheme based on stable diffusion (SD) to improve the efficiency and accuracy of substitute training. Despite the data generated by the SD exhibited high quality, it presented a different distribution of domains and a large variation of positive and negative samples for the target model. For this problem, we propose latent code augmentation (LCA) to facilitate SD in generating data that aligns with the data distribution of the target model. Specifically, we augment the latent codes of the inferred member data with LCA and use them as guidance for SD. With the guidance of LCA, the data generated by the SD not only meets the discriminative criteria of the target model but also exhibits high diversity. By utilizing this data, it is possible to train the substitute model that closely resembles the target model more efficiently. Extensive experiments demonstrate that our LCA achieves higher attack success rates (ASRs) and requires fewer query budgets compared to GANs-based schemes for different target models. Our codes are available at https://github.com/LzhMeng/LCA.

3D Gaussian splatting has recently achieved remarkable progress in dynamic scene reconstruction. However, there remain two practical challenges: (1) Existing methods typically employ a strict point-wise deformation structure to model dynamic attributes, while neglecting the uncertain motion correlation in local space, leading to inferior adaptability to complex scenes. (2) The inherent low-frequency bias properties of Gaussians often lead to blurring artifacts due to the insufficient high-frequency learning of variable motions. To address these challenges, we propose a novel Frequency-aware Uncertainty Gaussian Splatting, termed FUGS, for adaptively reconstructing dynamic scenes in the Fourier space. Specifically, we design an Uncertainty-aware Deformation Model (UDM) that explicitly models motion attributes using learnable uncertainty relations with neighboring Gaussian points. Such a paradigm is capable of facilitating temporal and spatial motion correlation learning, thereby enabling flexible Gaussian deformations. Subsequently, a Dynamic Spectrum Regularization (DSR) is developed to perform coarse-to-fine Gaussian densification through low-to-high frequency filtering. By weighting the gradient with frequency distance, the Gaussian attribute is adaptively adjusted according to the scene complexity. Benefiting from the flexible optimization, our method achieves high-fidelity reconstruction of complex scenes while enjoying real-time rendering. Extensive experiments on synthetic and real-world datasets show that our FUGS exhibits significant superiority over state-of-the-art methods. The code will be available at https://github.com/KevinJoee/GS.

Deep learning-based image restoration methods trained on synthetic datasets have witnessed notable progress, but suffer from significant performance drops on real-world images due to huge domain shifts. To alleviate this issue, some recent methods strive to improve the generalization ability of models with unpaired training. However, these solutions typically handle each problem individually and ignore the shared physical properties of different harsh scenarios, i.e., heavy rain, hazy and low-light images degrade more densely with increasing scene depth. Such limitations make them generalize poorly to real-world images. In this paper, we propose a novel Physically Oriented Generative Adversarial Network (POGAN) for unpaired image restoration with depth-density priors. Specifically, our POGAN consists of two core designs: Physical Restoration Network (PRNet) and Degradation Rendering Network (DRNet). The former focuses on estimating the physical components related to the depth and density distribution for restoration, while the latter re-renders degradation effects guided by the estimated depth information. To further facilitate learning the above physical prior, we design a Spatial-Frequency Interaction Residual block (SFIR), which efficiently learns global frequency information and local spatial features in an interactive manner. Extensive experiments on synthetic and real-world datasets demonstrate the superiority of our method in heavy rain, haze, and low-light scenarios.

Hyperspectral image (HSI) denoising is an essential preprocessing step for downstream applications. Although vision transformer (ViT)-based approaches show impressive denoising performance through self-similarity modeling, these methods still fail to exploit spatial and spectral correlations while ensuring flexibility and efficacy. To address this issue, we propose a hyperspectral denoising transformer using low-rank prompt (HyLoRa), simultaneously taking the spatial self-similarity and spectral low-rank property into account for HSI denoising. Specifically, to fully utilize intrinsic similarity in spatial domain, we perform cross-shaped window-based spatial self-attention for effectively modeling local and global similarity. Moreover, to exploit low-rank inductive bias, we integrate a low-rank prompt module into attention calculation for counting corrected low-dimensional vectors from a large collection of HSIs. This helps to better refine underlying noise-free structure representations. Compared to existing works, powerful capabilities for modeling spatial and spectral correlations can be built to correct low-rank representation in the feature space. Extensive experiments on both simulated and real remote sensing noise demonstrate that our HyLoRa consistently surpasses the state-of-the-art methods.

Convolutional neural networks (CNNs) are known to be vulnerable to adversarial examples even in black-box scenarios, posing a significant threat to their reliability and security. Most existing black-box attack methods primarily focus on data-free scenarios, which often require a large number of queries and yield low attack success rates. But in practical applications, it is feasible to collect a small amount of data associated with the target network. In light of this, in this article, we propose an advancing few-shot black-box attack with alternating training scheme using few data and alternating training to improve the efficiency and attack success rate. Specifically, we propose an alternating training approach consisting of two parts, both aimed at optimizing the substitute network, which alternate and reinforce each other, leading to a significant reduction in the query budget required for a successful attack. In addition, we propose an image degradation (ID) module that expands the data volume diversity through ID techniques to mitigate the problem of generator overfitting. Furthermore, we design a model specific adapter to enable the substitute networks to dynamically adjust the parameters for different target networks. Extensive experiments demonstrate the efficacy of our approach in significantly reducing the query budget while achieving higher attack success rates compared to state-of-the-art competitors.

Although existing super-resolution (SR) techniques have made great progress, they are often tailored for either paired or unpaired scenery, thus may result in poor migration ability. In this work, we propose a generalized Frequency Domain Separation Network (FDSNet) for both paired and unpaired SR settings. Firstly, through statistical analysis, we found that real-world low-resolution (LR) images and high-resolution (HR) images differ greatly in high frequencies but less in low frequencies. Inspired by this, we perform high and low-frequency separation of LR images and guide our model to reconstruct the HR contents in the different frequency domains. Then, according to the varying attention on frequencies of traditional CNN and Transformer models, we design a parallel pipeline: LFNet based on Transformer for low-frequency feature extraction, and HFNet based on CNN for high frequencies. In LFNet, to further alleviate the high complexity and data dependency of Transformer, Simplified Multi-head Self Attention (SMSA) is proposed at a low computational cost. And original MLP is replaced by our Spatial Enhancement MLP (SEMLP) to take full advantage of local spatial contexts. Finally, to further facilitate frequency separation and learning, a Frequency attention block is designed to impose guidance on high frequencies. Experiments indicate that our FDSNet achieves promising performance in terms of quantitative and qualitative evaluations while enjoying a faster speed and much fewer parameters.