Performance analysis of flow-based Video Interpolation Techniques

Video Frame Interpolation (VFI) is a crucial post-processing technique in computer vision, impacting applications in video recovery, vigilance, and enhancement. VFI aims to elevate a video’s frame rate by synthesizing novel frames between existing ones. Recent advancements in computer vision have yielded numerous VFI algorithms with diverse approaches for intermediate frame generation. This paper conducts a rigorous analysis of prominent VFI techniques, namely Real-Time Incremental Flow Estimation (RIFE) and eXtreme Video Frame Interpolation (XVFI), applied across various video formats encoded with distinct codecs. The evaluation demonstrates that RIFE achieves significantly faster processing times compared to XVFI. However, XVFI surpasses RIFE in generating frames of superior visual quality based on established metrics. Given the prevalence of diverse video formats employed in various applications, this work underscores the potential for further research and development in VFI algorithms, paving the way for advancements across multiple video processing domains.