Rekframes

Rekframes are a modular data construct used in video processing and computer vision to represent frames as nodes in a connected graph rather than isolated images. The term, a blend of reframe and frame, is used in discussions of scalable video representations and interactive media pipelines.

A rekframe typically stores a minimal set of information sufficient to reconstruct or render the frame within a pipeline. Common fields include an identifier, a timestamp, a compact feature vector representing appearance, encoding parameters, spatial resolution, and a keyframe flag. It also maintains explicit links to related rekframes, such as predecessor and successor frames, reference frames for predictive coding, and any semantic annotations or edit histories. This structure enables efficient incremental updates and selective reconstruction of frames without decoding the entire sequence.

Applications of rekframes appear in non-linear editors and content-based search, where they enable fast queries by visual features and easy retiming or recontextualization of frames. In streaming and augmented/virtual reality, the graph connections support scene reconstruction, selective streaming of high-interest frames, and caching strategies. In video editing, rekframes support re-framing operations, such as cropping or stabilization, by allowing changes to be applied at frame level and propagated through the graph.

History and status: Rekframes emerged in technical discussions as a conceptual alternative to linear frame streams and plain keyframes. They are not a widely standardized format, and implementations vary by project and vendor. They share concepts with frame graphs, scene graphs, and content-addressable video representations. See also keyframe, frame graph, scene graph, and content-based video retrieval.