NAL Unit Map Entry for Non-Rectangular Tile Sets in HEVC

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current mechanisms for accessing and processing video data in file formats, such as those based on the ISO base media file format, face inefficiencies when dealing with non-rectangular regions of tiles, particularly in HEVC and L-HEVC formats, which complicates the handling and transmission of specific tile sets.

Innovation Solution

The implementation of a method that generates and processes Network Abstraction Layer (NAL) Unit Map Entries and tile set group entries within files, using syntax elements like group identifiers, horizontal and vertical offsets, and width and height elements to efficiently manage and decode non-rectangular tile sets, allowing for improved handling of tile sets in HEVC and L-HEVC formats.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional rectangular tile grouping mechanisms are used in HEVC file formats, then processing and accessing video data follows standard procedures, but handling non-rectangular regions becomes inefficient and complex

Engineering Contradiction:
Improveefficiency of accessing and processing video dataVSAvoidcomplexity of handling non-rectangular tile sets
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the video picture into multiple tiles that can be independently grouped into tile sets. By dividing the picture into rectangular tiles first, then allowing flexible grouping of these tiles into non-rectangular tile sets, the system achieves both efficient rectangular processing and flexible non-rectangular region handling. The NAL Unit Map Entry box segments and maps NAL units to specific tile sets, enabling independent access and processing.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a new dimensional aspect by adding offset and size parameters (horizontal_offset, vertical_offset, width, height) to the tile set group entry syntax. This allows the representation of non-rectangular tile sets by defining their bounding box and individual tile offsets, transforming the problem from handling complex non-rectangular shapes directly to managing rectangular bounding regions with positional information.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Measurement precision

If non-rectangular tile sets are supported without additional syntax elements, then the file format remains simple, but precise identification and extraction of specific tile regions becomes impossible

Engineering Contradiction:
Improveprecision of identifying and extracting non-rectangular tile setsVSAvoidcomplexity of file format syntax
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary NAL Unit Map Entry box that acts as a mediator between the NAL units and the tile sets. This intermediary structure contains the necessary mapping information and references to tile set group entries, enabling precise identification of non-rectangular tile sets without requiring complex direct representation methods in the core file format syntax.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the need for complex geometric shape representation with a simpler parameter-based system. Instead of defining non-rectangular shapes through multiple coordinates and boundary definitions, the system uses offset and size parameters relative to a bounding box, substituting a complex mechanical geometric definition system with a simpler parametric system.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If all tiles in a picture are processed together as a single rectangular region, then processing is straightforward, but transmission and decoding of specific regions becomes inefficient

Engineering Contradiction:
Improveefficiency of transmitting and decoding specific video regionsVSAvoidease of handling and processing individual tile sets
Core Design Contradiction:
ProductivityVSEase of operation

Solution Approach 1:

The patent segments the complete picture into multiple independent tile sets, where each tile set can be individually processed, transmitted, and decoded. The NAL Unit Map Entry box enables selective access to specific tile sets by mapping NAL units to particular tile set groups, allowing receivers to efficiently process only the required regions without handling the entire picture.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary organization of tiles into named tile sets with defined groupings during the encoding phase. By pre-establishing these tile set groups and their mappings to NAL units, the system enables efficient random access and selective decoding at the receiving end, eliminating the need for complex runtime processing to identify and extract specific regions.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3437327B1Improvement on tile grouping in HEVC and l-HEVC file formats
Publication Date: 2019.12.25 QUALCOMM INC
  • EP3437327B1 patent drawingFigure 1
  • EP3437327B1 patent drawingFigure 2
  • EP3437327B1 patent drawingFigure 3

AI summary

A device generates a Network Abstraction Layer (NAL) Unit Map Entry box in a file. The NAL Unit Map Entry box associates NAL units with a tile set that includes two or more tiles of a plurality of tiles into which a current picture is partitioned. A region of the current picture covered by the tile set may be nonrectangular. Additionally, the device generates a tile set group entry in the file. The tile set group entry includes a horizontal offset, a vertical offset, a width, and a height. The horizontal offset specifies a horizontal offset of a top left pixel of a minimum rectangular region that covers tiles in the tile set. The vertical offset specifies a vertical offset of the top left pixel of the minimum rectangular region. The width specifies a width of the minimum rectangular region. The height specifies a height of the minimum rectangular region.