Multi-view Image Encoding Using Reference Depth Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for generating a view-synthesized image in multi-view image encoding and decoding require significant processing and memory resources, especially when generating the image for the entire frame, which can lead to increased calculation and memory usage.
Innovation Solution
An image decoding and encoding apparatus that predicts images between different views using a reference image and a reference depth map, allowing for the generation of a view-synthesized image by setting a reference depth region and using depth information to generate an inter-view predicted image, thereby reducing the need for generating a depth map for the processing target frame.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a depth map is generated for the processing target frame to create a view-synthesized image, then image quality is improved, but processing time and memory usage increase significantly
Solution Approach 1:
The patent applies preliminary action by generating a depth map for a reference frame (other than the processing target frame) in advance and storing it in the reference depth map storage unit. This pre-generated depth map is then reused for creating view-synthesized images of multiple different views, eliminating the need to regenerate the depth map for each view and significantly reducing processing time while maintaining image quality.
2Measurement precision
If a depth map is generated for the processing target frame, then view-synthesized image quality is improved, but memory usage increases
Solution Approach 1:
The patent applies universality by making the reference depth map serve multiple functions. The single reference depth map stored in the reference depth map storage unit is reused to generate view-synthesized images for multiple different views (first view, second view, third view, etc.). This multi-functional usage eliminates the need to store separate depth maps for each view, reducing memory usage while maintaining image quality across all views.
3Productivity
If disparity-compensated prediction is performed using depth information, then encoding efficiency is improved, but calculation amount increases
Solution Approach 1:
The patent applies copying by using a pre-generated reference depth map (which is a copy of depth information from another frame) instead of generating a new depth map for the processing target frame. The reference depth map is copied into the depth map storage unit and then used for disparity-compensated prediction across multiple views. This copying approach maintains encoding efficiency while significantly reducing the calculation amount required to generate depth maps for each view.
Data Source
AI summary
An image encoding/decoding apparatus for performing encoding/decoding while predicting an image between different views using a reference image for a view different from a processing target image and a reference depth map which is a depth map for an object of the reference image when a multi-view image including images of a plurality of different views is encoded/decoded includes a reference depth region setting unit configured to set a reference depth region which is a corresponding region on the reference depth map for decoding target regions into which the processing target image is divided, and an inter-view prediction unit configured to generate an inter-view predicted image for the processing target region from the reference image using depth information in the reference depth region as depth information for the processing target region.