Scene Text Image Super-Resolution (STISR) methods improve recognition accuracy by refining text regions locally. However, most existing approaches are limited by short-range dependencies, hindering the modeling of long-range semantic relationships between characters, which reduces their effectiveness in complex text scenes. Moreover, current methods are predominantly evaluated on synthetic datasets, which do not adequately capture real-world challenges such as diverse text styles, complex backgrounds, and spatial distortions. To address these limitations, we propose a Dual-Guided Visual and Semantic (DGVS) framework. This innovative framework utilizes a recognizer-driven attention mechanism to decouple character sequences, effectively distinguishing text from background noise. Additionally, we incorporate a state-space model to establish global semantic reasoning links, enhancing the comprehension of contextual relationships within text images. Furthermore, we construct Real-World Text (RealWT), a novel real-world benchmark dataset that integrates diverse data sources, including online images and multi-device captures. This dataset incorporates factors like device variations, resolution differences, and degradation, offering a more realistic simulation of real-world conditions. It enables models to learn degradation patterns that closely mirror practical applications, offering a standardized benchmark for evaluating STISR performance. Extensive experiments demonstrate that our method outperforms existing approaches, as validated by evaluations on TextZoom and RealWT. Our dataset and code are available on https://github.com/yoursmith/sde-DGVS.

This article delves into the exploration of state-space methods applied to the modeling and estimation of systems with time-varying parameters. While typically existing approaches rely on the assumption that the parameters satisfy the Markov evolution and require the knowledge of the transfer matrix, this article develops an explicit autoregressive (AR) model for time-varying parameters in which the invariant matrix represents the dynamic changes in the parameters. Unlike the previous work, the state-space model is constructed by stacking the invariant matrix and time-varying parameters into the unknown state vector. Then, the joint state estimation (JSE) algorithm is deduced based on the Kalman filtering principle, aiming to reduce the dependence on the prior knowledge of the invariant matrix. Through the numerical simulation and Monte Carlo test, it is indicated that the developed algorithm maintains reliability under various random white noises. In addition, the practical estimation results with the real-time series also verify the validity.

Video has become increasingly widespread in information transmission and daily communication. However, video data often contain sensitive privacy information, which leads to privacy concerns. To address these concerns, video encryption has become one of the most effective privacy protection methods. Current efforts in video encryption focus on block-level or frame-level encryption, failing to provide fine-grained privacy protection. In response, we propose a personalized pixel-level video encryption approach for privacy protection. Our approach operates before encoding and is robust to lossy compression. It applies a quadruple encryption algorithm directly to the pixels within privacy-sensitive areas segmented by instance segmentation, achieving fine-grained, pixel-level encryption. Different encryption keys are managed using Attribute-Based Encryption (ABE) technology, which enables personalized access control. This approach assigns different permissions based on user identity, enabling multi-level access to encrypted video content.We deploy our approach across different devices and perform extensive experimental evaluations. Experimental results show that on different devices, the single-frame encryption overhead of our approach for videos of various resolutions consistently remains below 0.12 seconds, with the encrypted area achieving an average PSNR of 8.50 dB and an average SSIM of 0.079. When using a key with one bit difference from the correct key for decryption, the average number of pixels changing rate (NPCR) of the test sequence is 99.59% and the average unified average change intensity (UACI) is 32.40%. This shows that unauthorized users cannot obtain valid information when decrypting with an incorrect key.

The brain has great plasticity. Cognitive training can change the functional connectivity of brain networks and then improve trainees’ cognitive ability. Most cognitive training studies have focused on revealing the correlation between cognitive ability improvement and structural changes in functional brain networks. The attention to change in brain dynamics induced by cognitive training is relatively less. No report was found for the alteration of state distributions in phase spaces. Abacus-based mental calculation (AMC) training is one of cognitive training methods, often used to enhance the cognitive ability of subjects. We construct a dynamical model of the brain based on the Aihara chaotic neuron model and resting-state functional brain networks. The differences on phase space distributions between the AMC group and the control group are investigated. Compared to the control group, the state distributions in the phase space of brain regions associated with memory function exhibite aggregation in the AMC group, while the state trajetories of AMC group are diffused in larger area in the phase spaces with bain regions related to the somatomotor network. Our results indicate that AMC training can promote the memory function of participants and the functional segregation of community structures in the brain. Furthermore, our work suggests that greater improvements in memory function of the right cerebral than that of the left cerebral hemisphere in AMC training.

Abacus-based mental calculation (AMC) training is one of cognitive training methods, and can affect the functional segregation and integration of functional brain networks. However, there is no report about the impact of AMC training on functional brain networks from the dynamic behavior. The community synchronization of brain networks can promote information propagation and functional integration among brain regions. In this work, we focus on the effect of AMC training on the community synchronization of functional brain networks. The dynamic model based on and Hindmarsh-Rose neuron model is used to the functional brain networks of the AMC training group and the control group. The community synchronization of the functional brain networks are measured with the dynamic model. It is found that there are no significant differences between the AMC training group and the control group for synchronization at the level of the whole brain, but the synchronization of visual network associated with AMC training and the default network associated with resting-state increase significantly in the AMC training group, which indicates that information integration ability and information transmission ability between the brain regions of the two communities may be enhanced through AMC training. In addition, we notice the decrease in the synchronization of the auditory network for the AMC training group.