A Motion Refinement Network with Local Compensation for Video Frame Interpolation

Video frame interpolation (VFI) is a challenging yet promising task that involves synthesizing intermediate frames from two given frames. State-of-the-art approaches have made significant progress by directly images using forward optical flow. However, these methods often encounter issues such as occlusion and pixel blurring when handling large motion scenes. To address challenges, this paper proposes novel approach based on the concept of local compensation. By adopting approach, more refined flow estimation can be obtained, leading to higher-quality video results. Specifically, we introduce modules, namely Comprehensive Contextual Feature Extraction (CCFE) module Motion-Guided Fusion (MGFF) module, enable compensation estimation. The CCFE designed embedded in each layer image pyramid structure. It aims encourage model extract clean sufficiently rich contextual information input images. On other hand, MGFF guide multi-source features fusion features, making feature moving objects precise, thus providing for Extensive experimental results demonstrate incorporating our proposed modules into baseline network significantly enhances performance interpolation.