图像与视频处理：2025–2026 年度综述

An annual survey of the eess.IV arXiv category, covering research published from September 2025 – August 2026. Built on a close-reading set of 132 papers and organised into 37 chapters, it maps the year by the field's research sub-directions.

Book framework (share of the year's output):

• Medical Image Reconstruction and Synthesis (14.7%) — Focuses on generative models and iterative reconstruction techniques for MRI, CT, and histopathology to enhance image quality and enable synthetic data generation.
• Medical Image Segmentation and Lesion Detection (14.6%) — Addresses robust segmentation of tumors, lesions, and anatomical structures in medical imaging using interactive, federated, and uncertainty-aware deep learning methods.
• Deep Learning-Based Classification and Detection (14.3%) — Applies CNNs, Transformers, and attention mechanisms to object detection, quality assessment, and disease classification in both medical and industrial imaging contexts.
• Video Compression and Transmission Optimization (10.2%) — Develops adaptive bitrate control, semantic transmission, and learned codecs to optimize video quality and efficiency for streaming and network-constrained environments.
• General Deep Learning Frameworks and Applications (9.4%) — Explores diverse deep learning architectures for tasks such as multi-disease classification, remote sensing estimation, and reinforcement learning in autonomous systems.
• Super-Resolution and Inverse Problems (9.1%) — Tackles high-resolution image recovery and inverse problem solving in medical and remote sensing imaging through unrolled algorithms and generative priors.
• Vision-Language and Foundation Models (8.8%) — Integrates vision-language models and foundation models for medical VQA, target recognition, and cross-modal fine-tuning to enhance generalization and interaction.
• Neural Network Architectures and Representations (5.6%) — Investigates efficient neural network designs, implicit representations, and graph-based models for interpretable and resource-constrained image analysis.
• Medical Image Restoration and Edge Deployment (5.1%) — Focuses on denoising, enhancement, and efficient deployment of medical image processing algorithms on edge devices and mobile platforms.
• Real-Time Processing and 3D Gaussian Splatting (4.7%) — Advances real-time streaming, 3D Gaussian splatting, and embedded systems for time-critical applications like autonomous navigation and license plate recognition.
• Self-Supervised and Weakly Supervised Learning (3.6%) — Leverages self-supervised and weakly supervised frameworks to reduce annotation dependency in ultrasound, sonar, and whole-slide image analysis.

Each sub-direction is covered across three chapters — how it framed its questions, the strength of its key findings, and its open debates and frontiers — so the reader gets a traceable picture of what this part of the field actually produced over the year.