Multidimensional Media Segmentation for Parallel NBMP Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing Network-Based Media Processing (NBMP) standard does not support the segmentation of multidimensional media data, which is necessary for efficient processing of multidimensional media streams.
Innovation Solution
The method involves segmenting multidimensional media streams into segments using predefined multidimensional metadata, including starting vectors, length vectors, and startcodes, and deriving a network-based media processing workflow based on these vectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the NBMP standard is used for media processing, then independent processing of media segments is enabled, but the standard does not support segmentation of multidimensional media data
Solution Approach 1:
The patent divides multidimensional media data into multiple independent segments along different dimensions (temporal, spatial, frequency). Each segment is represented by sequence vectors containing metadata that enables independent processing while maintaining the ability to reconstruct the complete multidimensional media data through merging operations.
Solution Approach 2:
The patent extends the traditional one-dimensional temporal segmentation to multidimensional segmentation by introducing additional dimensions such as spatial dimensions and frequency dimensions. This is achieved through sequence vectors that encode position information across multiple dimensions, allowing segments to be defined and processed in a multidimensional space rather than just along the time axis.
2Productivity
If multidimensional media data is processed as a whole, then processing completeness is maintained, but processing efficiency is reduced
Solution Approach 1:
The patent segments multidimensional media data into multiple independent units that can be processed in parallel. Each segment contains metadata (sequence vectors) that identifies its position and dimensions, enabling the processing system to handle multiple segments simultaneously through parallel workflows while maintaining the ability to reconstruct the complete media data afterward.
Solution Approach 2:
The patent performs preliminary segmentation and metadata assignment before the actual media processing occurs. By pre-defining segment boundaries and assigning sequence vectors with position information in advance, the system prepares the multidimensional media data for efficient parallel processing without requiring complex real-time coordination during the processing phase.
3Ease of operation
If segments are processed independently, then processing flexibility is improved, but coordination overhead increases
Solution Approach 1:
The patent introduces sequence vectors as intermediary metadata structures that carry position information between segments and the merging process. These sequence vectors act as a coordination mechanism that enables independent segment processing while providing the necessary information for accurate reassembly, reducing the overhead of coordination by encoding all necessary position data in the segment metadata itself.
Data Source
AI summary
A method includes segmenting a multidimensional media stream into a plurality of segments of multidimensional media in a multidimensional space; representing each segment of the plurality of segments of multidimensional media using a respective sequence vector, the respective sequence vector comprising one or more predefined multidimensional metadata, wherein the predefined multidimensional metadata includes a startcode that is an identical unique code point and a sequence number, wherein each segment includes the identical unique code point; and deriving a network based media processing (NBMP) workflow based on the respective sequence vector of each segment of the plurality of segments.