Multi-view Coding Using Depth Map Estimate and Motion Vector Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current multi-view video coding techniques face challenges in efficiently exploiting interdependencies between views, leading to increased bit rates and reduced coding efficiency, especially when using autostereoscopic displays that require multiple views.
Innovation Solution
The proposed solution involves an apparatus and method for reconstructing and encoding multi-view signals by deriving motion vector predictor candidates using disparity vectors, and estimating depth maps to reduce inter-view redundancies and improve coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If multiple views are coded to enable autostereoscopic display, then the quality of stereo viewing impression is improved, but the bit rate increases approximately linearly with the number of coded views
Solution Approach 1:
The patent uses depth map information to generate virtual views by warping reference views, effectively copying and transforming existing view data rather than encoding all views independently. This allows multiple virtual views to be synthesized from a limited set of actual coded views, reducing the linear increase in bit rate while maintaining autostereoscopic display quality
Solution Approach 2:
The patent introduces depth map information as an additional dimension that enables view synthesis. By coding depth maps separately and using them to warp reference views, the system creates intermediate virtual views that would otherwise require separate encoding. This dimensional approach to view generation reduces the total number of views that need to be coded directly
2Quantity of substance
If depth maps are transmitted to enable view synthesis, then the number of coded views can be reduced, but additional data transmission is required
Solution Approach 1:
The patent changes the representation parameters by encoding depth information as depth maps rather than transmitting full additional views. This parameter transformation allows the system to represent multiple views using a combination of reference views and depth data, which is more efficient than transmitting all views independently
Solution Approach 2:
The patent performs depth map estimation and view synthesis operations at the decoder end using previously transmitted depth maps and reference views. This preliminary preparation of view data allows the system to generate multiple virtual views from limited transmitted information, reducing the overall data transmission requirement
3Productivity
If disparity-compensated prediction is used to exploit inter-view dependencies, then view interdependencies are exploited, but only a small subset of image samples can be predicted
Solution Approach 1:
The patent merges disparity-compensated prediction with motion-compensated prediction by combining both prediction techniques in a unified approach. This combination allows the system to exploit both spatial (inter-view) and temporal dependencies, significantly increasing the subset of image samples that can be effectively predicted while maintaining operational simplicity
Data Source
AI summary
This disclosure is directed to coding a multi-view signal, which includes processing a list of plurality of motion vector candidates associated with a coding block of a current picture in a dependent view of the multi-view signal. Such processing includes estimating a first motion vector based on a second motion vector associated with a reference block in a current picture of a reference view of the multi-view signal, the reference block corresponding to the coding block of the current picture in the dependent view. The first motion vector is added into the list, and an index is used that specifies at least one candidate from the list to be used for motion-compensated prediction. The coding block in the current picture is coded by performing the motion-compensated prediction based on the at least one candidate indicated by the index.