Inter Frame
Inter-frame processing in video analysis focuses on leveraging the relationships between consecutive frames to improve various tasks, such as video interpolation, segmentation, and object tracking. Current research emphasizes efficient model architectures, including transformers and state space models, to handle the computational demands of processing high-resolution videos and long sequences, often incorporating techniques like attention mechanisms and optical flow for improved accuracy. These advancements are significantly impacting fields like video compression, medical imaging (e.g., cardiac PET), and autonomous driving by enabling real-time processing and more robust performance in challenging scenarios. The development of more efficient and accurate inter-frame methods continues to be a key area of focus for improving the quality and speed of video-related applications.
Papers
Motion estimation and filtered prediction for dynamic point cloud attribute compression
Haoran Hong, Eduardo Pavez, Antonio Ortega, Ryosuke Watanabe, Keisuke Nonaka
Learned Video Compression for YUV 4:2:0 Content Using Flow-based Conditional Inter-frame Coding
Yung-Han Ho, Chih-Hsuan Lin, Peng-Yu Chen, Mu-Jung Chen, Chih-Peng Chang, Wen-Hsiao Peng, Hsueh-Ming Hang