Person re-identification (ReID) is a challenging task that aims to identify individuals across multiple non-overlapping camera views. To enhance the performance and robustness of ReID models, it is crucial to train them over multiple data sources. However, the traditional centralized approach poses a significant challenge to privacy as it requires collecting data from distributed data owners. To overcome this challenge, we employ the federated learning approach, which enables distributed model training without compromising data privacy. In this paper, we propose a novel feature-aware local proximity and global aggregation method for federated ReID to extract robust feature representations. Specifically, we introduce a proximal term and a feature regularization term for local model training to improve local training accuracy while ensuring global aggregation convergence. Furthermore, we use the cosine distance of backbone features to determine the global aggregation weight of each local model. Our proposed method significantly improves the performance and generalization of the global model. Extensive experiments demonstrate the effectiveness of our proposal. Specifically, our method achieves an additional 27.3% Rank-1 average accuracy in federated full supervision and an extra 20.3% mean Average Precision (mAP) on DukeMTMC in federated domain generalization.

Real-world image super-resolution (Real-SR) is a challenging task due to the unknown complex image degradation. Recent research on Real-SR has achieved remarkable progress by degradation process modeling; however, there are still undesired structural distortions and over-smoothed textures in the recovered images. In this letter, we propose a structure and texture preserving network, towards reducing the structural distortions while refining the perceptual-pleasant textures. Specifically, we propose a structure tensor (ST) branch to guide the restoration of high-resolution images by extracting channel-aggregated structural information. To further adaptively optimize different local texture, we replace the global discriminator with a global-local discriminator. By “local,” we mean that the discriminator loss, imposed on local areas randomly selected from the generated SR image, is minimized to generate textures with great visual perception in the selected local areas. Experimental results on five real-world datasets demonstrate the superiority of our methods in restoring structures, generating visually realistic SR images, as well as handling images of different degradation levels.

Among the key sub-topics of image fusion, infrared-visible image fusion technology is widely used in modern military and civilian domain. Driven by the data-free and end-to-end advancedness, growing research efforts have been devoted to convert the image fusion task into a norm minimization problem. To simultaneously keep the highlighted thermal information in the infrared image and the clear appearance information in the visible image, we propose a low-complexity fusion algorithm via guided transformation minimization, namely L2GTM. In particular, we formulate the fusion task as a solely `2norm minimization problem, which enjoys the uniqueness of the optimal solution. To adaptively integrate information from both infrared and visible images, we introduce guided weights to determine the degree of information retention. Experiments on public datasets validate the competitiveness of L2GTM from both the visual quality and objective evaluation perspective.

With the fast development of multi-sensor technologies, image fusion has played an essential role in modern military and civilian. To better integrate thermal radiation information in infrared images and detailed appearance information in visible images, we investigate a novel norm formulation via joint contrast and gradient preservation, for infrared-visible image fusion. Specifically, we employ a structure tensor measurement to characterize the similarity between the fused image and the infrared image in terms of thermal radiation information, to better integrate visible appearance details. Since natural image gradients follow the hyper-Laplacian distribution, we employ$ \ell_{ p\in\left[0.5,0.8\right]} $norm instead of$ \ell_0 $or$ \ell_1 $norm to measure the gradient term to further extract texture information from visible images. To cope with the computational problem introduced by the coupled structure tensor and non-convex$ \ell_{p\in\left[0.5,0.8\right]}$norm, we propose a computational efficient solver based on half-quadratic splitting scheme. Experiments on public datasets validate the competitiveness of FCGP from both the subjective and objective evaluation perspective.