Reference Image Generation for Non-Linear Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video encoding techniques that generate additional images by calculating average or median pixel values struggle with maintaining image quality and code efficiency when encoding videos with non-linear or irregular movements, such as those seen through fluctuating water surfaces, leading to increased code amounts and deteriorated image quality.
Innovation Solution
An image generation apparatus that suppresses changes based on the property of a first subject, generating a reference image by setting specific areas within frames and using provisional reference images for motion-compensated prediction, with weighting of pixel values to maintain image stability and reduce deformation, effectively creating an additional image that is not displayed but aids in encoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional encoding generates additional images using average or median pixel values from multiple encoding target images, then coding efficiency is improved for videos with static backgrounds, but prediction accuracy decreases for videos with non-linear or irregular movements
Solution Approach 1:
The patent changes the parameters used for generating additional images from simple average or median pixel values to motion-compensated pixel values that account for actual object movement. This involves calculating motion vectors between frames and using these to select or generate pixel values that accurately represent the background, thereby improving prediction accuracy for videos with non-linear or irregular movements while maintaining coding efficiency.
2Quantity of substance
If conventional encoding uses additional images for background compensation, then code amount is reduced for static scenes, but image quality deteriorates for scenes with irregular movements
Solution Approach 1:
The patent improves image quality by changing how additional images are generated - instead of using simple statistical methods (average/median), it uses motion compensation techniques that consider actual pixel displacement. This involves calculating motion vectors, performing motion-compensated prediction, and selecting pixel values that accurately represent the background after motion compensation, thereby maintaining high image quality even for scenes with irregular movements while keeping code amount low.
3Productivity
If motion-compensated prediction is used with conventional additional images, then encoding efficiency is maintained for regular movements, but prediction accuracy decreases for non-linear and irregular movements
Solution Approach 1:
The patent enhances motion-compensated prediction by changing the parameters used to generate additional images. Instead of relying on simple temporal averaging, the system calculates motion vectors between frames and uses these to select pixel values that accurately represent the background. This involves using motion compensation to predict background pixel values at current frame positions, thereby improving prediction accuracy for non-linear and irregular movements while maintaining encoding efficiency through the use of these enhanced additional images.
Data Source
AI summary
An image generation apparatus for generating a reference image that is referenced when encoding a time-series frame group that is a set of frames in which a second subject is captured so as to be temporally apparently deformed due to a property of a first subject comprises a reference image generation unit that generates the reference image in which a change in an image based on the property of the first subject is suppressed and apparent deformation of the second subject is suppressed.