Abstract: Frame interpolation is a computational technique used in video processing to create additional frames between existing frames, thereby enhancing the smoothness and visual quality of motion in the video.Existing techniques for frame interpolation in videos include path-based and phase-based conventional methods, convolutional neural network (CNN)-based flow-based methods, kernel-based methods utilizing convolution operations over local patches, and recent advancements such as deformable convolution-based approaches like AdaCoF. Addressing the challenges in frame interpolation is essential for developing more efficient and versatile techniques applicable across various platforms and applications. High computational costs, particularly prevalent in methods reliant on deep neural networks (DNNs), hinder deployment on resource-constrained devices or real-time applications. Complexity arises from intricate model architectures or multi-stage processes, complicating both understanding and implementation. Additionally, limited generalization restricts the practical utility of certain techniques, as they may excel on specific datasets but struggle with diverse or unseen data. Methods relying solely on pixel-wise information or local kernels may falter in accurately interpolating frames with complex motion, occlusion, or fine details. Furthermore, the large size of state-of-the-art models poses challenges for storage, training, and deployment, especially on mobile or embedded devices. Addressing these issues is paramount for advancing frame interpolation methods towards greater efficiency, practicality, and applicability across a broad spectrum of contexts and platforms. Our Compression-Driven Framework for Video Interpolation (CDFI) addresses key challenges as follows: Reduced Computational Cost, Simplicity and Efficiency, Improved Generalization, Enhanced Motion Handling, Compact Model Size.

Keywords: Video frame interpolation, Optical flow-based, Real-time solutions, Visual quality, Real-time applications, Stakeholders, Video processing.

Cite:
Dhruva Kumar Shetty, Pramith A Naiga, Sidhvin P Shetty, Yash Karunakar Shetty, Dr. Maryjo M George, "Frame Interpolation Using FILM", IJARCCE International Journal of Advanced Research in Computer and Communication Engineering, vol. 13, no. 3, 2024, Crossref https://doi.org/10.17148/IJARCCE.2024.13334.


PDF | DOI: 10.17148/IJARCCE.2024.13334

Open chat
Chat with IJARCCE