Multi-viewpoint Encoding Reference Index Allocation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for allocating reference image indexes in multi-viewpoint encoding are inefficient, leading to suboptimal encoding efficiency, especially when changing reference image indexes is necessary.

Innovation Solution

An image processing device that alternately allocates reference image indexes to referable images in both the view direction and time direction, allowing for flexible allocation patterns based on pattern identification information and prioritization, such as time direction prioritized or view direction prioritized patterns.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If reference image indexes are allocated sequentially in time direction only, then allocation simplicity is maintained, but encoding efficiency deteriorates in multi-viewpoint scenarios

Engineering Contradiction:
Improveencoding efficiencyVSAvoidallocation complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the reference image allocation process into two distinct directions: time direction allocation and view direction allocation. The index allocation unit alternates between allocating indexes in the time direction and allocating indexes in the view direction, allowing each direction to be handled independently while achieving overall optimization for multi-viewpoint encoding scenarios

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces dynamic allocation patterns that can adapt to different encoding scenarios. The system can switch between time-direction-prioritized patterns and view-direction-prioritized patterns based on the specific multi-viewpoint encoding requirements, making the allocation process flexible and efficient

Inventive Principle:
Principle #15Dynamics

2Measurement precision

If reference image indexes are frequently changed to optimize prediction, then prediction accuracy improves, but encoding overhead increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidencoding overhead
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent performs preliminary allocation of reference image indexes in both time and view directions before the actual encoding process. By pre-establishing the allocation pattern and identifying the most suitable reference images in advance, the system reduces the need for frequent changes during encoding, thereby minimizing encoding overhead while maintaining high prediction accuracy

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS10021386B2Image processing device which predicts an image by referring a reference image of an allocated index
Publication Date: 2018.07.10 SONY GROUP CORP
  • US10021386B2 patent drawing
  • US10021386B2 patent drawing
  • US10021386B2 patent drawing

AI summary

The present technique relates to an image processing device and a method that can improve encoding efficiency in multi-viewpoint encoding. A reference image index allocation unit alternately allocates a reference image index referred by a motion estimator/compensator to predict an image to be processed to a referable image in a view direction and to a referable image in a time direction. In allocation, the reference image index allocation unit refers to time information and view information of the reference image accumulated in a decoded picture buffer, information such as a sequence parameter set and a slice header from a syntax processing unit, and the like.