Region Annotation Encapsulation in Media Tracks Without Data Duplication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ISOBMFF-based file formats lack efficient and interoperable means to associate annotations with individual regions within images or image sequences, leading to duplication of information and high description costs.
Innovation Solution
The method involves generating a region track with region description data structures that include persistent region identifiers across multiple samples, allowing efficient association and parsing of annotations with media data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If annotations are associated with each sample individually using existing ISOBMFF-based file formats, then annotations can be stored, but information duplication occurs and description costs increase
Solution Approach 1:
The patent merges the association of annotations with regions across multiple samples by introducing a region track that maintains persistent region identifiers. Instead of duplicating annotation information in each sample, the region track consolidates region definitions and their associated annotations, allowing multiple samples to reference the same region identifiers without repeating the full annotation data.
Solution Approach 2:
The region track serves as a universal structure that can be associated with multiple media tracks and samples. The persistent region identifiers created in the region track can be referenced by multiple samples across different time points, making the annotation structure multi-functional and reusable throughout the media presentation without requiring separate annotation definitions for each sample.
2Reliability
If region annotations are encapsulated in each sample, then annotations are preserved, but the file size and processing overhead increase
Solution Approach 1:
The patent extracts the region definition and annotation data from individual samples and places it in a separate region track. This extraction removes the redundant annotation information from each sample while preserving the essential region identifiers that samples need to reference their annotated regions, thereby reducing overall data volume while maintaining annotation integrity.
Solution Approach 2:
Instead of storing full annotation copies in each sample, the patent creates lightweight region identifier references that copy only the essential identification information. Samples reference regions by identifier rather than containing complete annotation data, significantly reducing data volume while maintaining the ability to retrieve and preserve full annotation information when needed.
3Adaptability or versatility
If generic ISOBMFF readers are used to process annotated media, then interoperability is achieved, but efficient region-annotation association is not possible
Solution Approach 1:
The patent segments the media file structure into distinct functional components: a region track that handles region definitions and annotations, and media tracks that contain the actual media samples. This segmentation allows generic ISOBMFF readers to process the file using standard box parsing while enabling specialized efficient processing of region-annotation associations through the dedicated region track structure.
Solution Approach 2:
The region track acts as an intermediary layer between the media samples and the annotation data. It provides a standardized interface that generic ISOBMFF readers can access through normal box parsing, while simultaneously enabling efficient region-annotation association through persistent identifiers. This intermediary structure maintains interoperability while improving processing efficiency.
Data Source
AI summary
The present invention concerns a method of generating a media file, the method comprising: generating a first track of media data samples; generating a region track associated with the first track, each sample of the region track comprising a region description data structure comprising geometry information of one or more regions in at least one corresponding sample of the first track; wherein the region description data structure comprises a region identifier associated with each region; and the method further comprise: generating a region annotation in a container; generating association metadata associating a region identifier with the region annotation; generating a media file including the first track, the region track, the region annotation, and the association metadata.