Parallel Image Coding Global Motion Vector Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In image processing, parallel coding methods often result in reduced global motion vector (GMV) precision due to the need to derive GMVs using motion information from older pictures, leading to decreased image quality.
Innovation Solution
An image processing device that divides images into multiple arrangements, allowing each coding unit to perform inter-prediction using a first global motion vector derived from finished motion information of preceding pictures, and combines streams from different coding units to enhance GMV precision and image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If parallel coding is performed using multiple encoders, then processing speed is improved, but global motion vector precision deteriorates
Solution Approach 1:
The patent applies preliminary action by deriving the global motion vector from motion information of a picture that has already been processed (finished), before the current picture coding begins. This ensures that the GMV is available and precise when needed for inter-prediction in parallel coding, without waiting for the preceding picture to be fully processed by another encoder.
2Productivity
If GMV is derived from older pictures due to parallel processing, then processing parallelism is maintained, but image quality deteriorates
Solution Approach 1:
The patent uses an intermediary approach by selecting and using motion information from a specific picture (the nearest picture whose processing is finished) as a mediator to derive the global motion vector. This intermediary picture serves as the optimal bridge between maintaining parallel processing and ensuring image quality, providing the most recent reliable motion data available.
Data Source
AI summary
An image processing device includes: a division unit that divides each picture of image data into multiple arrangements; multiple coding units, each of which codes the pictures in the mutually-different arrangements that result from the division by the division unit; and a composition unit that composites streams in the arrangements, which are obtained by each of the multiple coding units coding each picture, in which when coding a current picture, the coding unit performs inter-prediction using a first global motion vector (GMV) that is derived from motion information on a portion whose processing is finished, of a picture that precedes the current picture in the image data that is present before being divided by the division unit and whose processing is in progress in a different coding unit.