Advances in Video Processing and Analysis

The field of video processing and analysis is rapidly advancing, with a focus on improving the quality, authenticity, and security of video content. Recent developments have led to significant improvements in video generation, quality assessment, and forgery detection. Researchers are exploring new methods for integrating consistency information across long-short frames in video generation, and for assessing video quality in a more accurate and generalized way. Additionally, there is a growing need for robust video forensic techniques to counteract evolving forgery methods. Noteworthy papers in this area include: FreePCA, which proposes a training-free long video generation paradigm based on Principal Component Analysis, and VIDSTAMP, which introduces a watermarking framework for ownership and integrity in video diffusion models. HiLLIE is also notable for its human-in-the-loop training framework for low-light image enhancement.

Sources

FreePCA: Integrating Consistency Information across Long-short Frames in Training-free Long Video Generation via Principal Component Analysis

VIDSTAMP: A Temporally-Aware Watermark for Ownership and Integrity in Video Diffusion Models

Understanding and Exploiting Plasticity for Non-stationary Network Resource Adaptation

HiLLIE: Human-in-the-Loop Training for Low-Light Image Enhancement

A Rate-Quality Model for Learned Video Coding

NTIRE 2025 Challenge on UGC Video Enhancement: Methods and Results

DiffVQA: Video Quality Assessment Using Diffusion Feature Extractor

Breaking Annotation Barriers: Generalized Video Quality Assessment via Ranking-based Self-Supervision

Towards Smart Point-and-Shoot Photography

Video Forgery Detection for Surveillance Cameras: A Review

Securing Immersive 360 Video Streams through Attribute-Based Selective Encryption

Built with on top of