ROI Video Decoding with Tile Groups for Lower Mobile Power
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies, such as HEVC, do not efficiently support region of interest (ROI) decoding, leading to increased decoding complexity and power consumption on mobile devices with limited screen sizes.
Innovation Solution
Implementing ROI video coding using tiles and tile groups, where tiles are partitioned into groups with constraints on motion prediction within the group boundaries, allowing selective decoding of only the ROI and reducing decoding complexity by decoding a subset of tiles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all tiles in a picture are decoded to ensure complete video quality, then video quality is maintained, but decoding complexity and power consumption increase significantly on mobile devices
Solution Approach 1:
The picture is divided into multiple tiles, and each tile is further divided into tile groups. This segmentation allows the decoder to process only specific tile groups containing ROI data rather than decoding the entire picture, thereby reducing decoding complexity while maintaining video quality in the displayed region.
Solution Approach 2:
Different quality levels are applied to different regions of the picture. ROI regions are identified and marked with specific syntax elements, allowing the decoder to focus computational resources on decoding these important regions at high quality while using reduced complexity for non-ROI areas.
2Reliability
If all tiles in a picture are decoded to ensure complete video quality, then video quality is maintained, but power consumption increases on mobile devices with limited screen sizes
Solution Approach 1:
By segmenting the picture into tiles and tile groups, the system enables selective decoding of only those tile groups containing ROI data. This reduces the total computational workload and consequently lowers power consumption on mobile devices, while still maintaining video quality for the displayed region.
Solution Approach 2:
Instead of decoding the entire picture (excessive action), the system decodes only the necessary tile groups containing ROI data (partial action). This partial decoding approach reduces power consumption while maintaining sufficient video quality for the displayed content on mobile devices with limited screen sizes.
3Productivity
If tiles are partitioned into tile groups with motion prediction constraints, then selective decoding of ROI is enabled, but coding complexity increases due to additional syntax elements and constraints
Solution Approach 1:
Tile groups are pre-defined and marked with syntax elements indicating ROI status before decoding. The motion prediction constraints and tile group boundaries are established in advance, allowing the decoder to quickly identify and process only relevant regions without complex runtime analysis, thus improving decoding efficiency.
Solution Approach 2:
The ROI information is extracted and marked with specific syntax elements (tile_group_id, tile_group_flag, motion_vector_prediction_constraints) that are separate from the main video data. This extraction allows the decoder to efficiently identify and process only the necessary tile groups while maintaining relatively simple coding structures.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
Systems, methods, and instrumentalities are disclosed relating to region of interest (ROI) video coding using tiles and tile groups. An encoded video sequence including a plurality of tiles may be received. The plurality of tiles may be divided into one or more tile groups. Signaling indicating parameters of the one or more tile groups may be received. A tile group of the one or more tiles groups may be decoded and a picture relating to the decoded tile group may be displayed. The decoded tile group may overlap the ROI. The ROI may correspond to the displayed picture and the displayed picture may be a portion of the encoded video sequence. The tile groups that do not overlap the ROI may not be decoded.