The synergy between foveal and peripheral processing is fundamental to the efficiency of biological vision. While hybrid Convolutional Neural Network (CNN)-Transformer architectures aim to capture both local and global features, they often rely on static, predefined structures that struggle to dynamically align information and adaptively allocate computational resources, ultimately limiting their performance. To address this limitation, we introduce the Central-Peripheral Vision Transformer (CPVT), a novel architecture that explicitly and hierarchically mimics this biological dichotomy. CPVT employs fine-grained, convolutionally modulated attention in its shallow layers to emulate foveal vision, while seamlessly transitioning to a coarse-grained, global attention mechanism in deeper layers to emulate peripheral vision. This design is enhanced by two specialized Feed-Forward Networks that facilitate synergistic information interaction. Rigorously validated on diverse medical imaging benchmarks, CPVT achieves state-of-the-art performance, attaining classification accuracies of 87.98% on the International Skin Imaging Collaboration (ISIC) 2018 challenge dataset and 90.41% on the Kvasir dataset. These results demonstrate that an adaptive, hierarchical integration of biological vision principles can significantly enhance machine perception for medical image analysis.

As vehicular edge computing (VEC) grows increasingly demanding, distributed task offloading is brought up as a potential solution. However, the high mobility and limited resources of vehicles, coupled with uncertain service reliability, make it difficult to guarantee timely and correct execution of computation-intensive tasks. Existing task offloading models overlook critical aspects such as punctuality and correctness, limiting their ability to select reliable service vehicles in dynamic and potentially adversarial environments. To address this gap, we propose a performance-reputation integrated task offloading scheme for reliable VEC. We introduce a dual-dimensional reputation model (D2Rep) that jointly evaluates vehicles based on the degree of punctuality, result correctness probability, and historical reputation value. Building on this, we develop the performance value-maximized vehicular computation offloading (PFVMax-VCO) scheme, which combines reputation value, link reliability, and projected success rate into a unified performance value to guide multi-objective optimization of delay, energy consumption, and reliability. To support the realistic evaluation, we implement a SUMO-Veins-OMNeT++ integrated simulation framework. Experimental results demonstrate that the proposed PFVMax-VCO scheme outperforms existing solutions in terms of execution cost, result correctness probability, and task success rate, while enabling robust and dynamic reputation management for VEC systems.