Depth Lookup Table Coding for 3D Video Bitrate Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3D video coding techniques, such as 3D-HEVC, face inefficiencies in coding depth maps due to redundant depth value signaling and lack of inter-view prediction, leading to increased bitrates and reduced coding efficiency, especially with scene changes and varying depth values across views.
Innovation Solution
The proposed solution involves signaling depth values in a depth lookup table (DLT) by predicting and coding relative differences between depth values, and implementing inter-view DLT prediction to reduce redundancy and bit requirements, using techniques like differential coding and inter-view prediction to signal depth values more efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If depth values are coded directly without prediction, then coding simplicity is maintained, but bitrate increases due to redundant depth value signaling
Solution Approach 1:
The patent applies preliminary action by constructing a depth lookup table (DLT) before coding depth values. The DLT pre-computes and stores depth information for different view configurations, allowing the encoder to reference pre-prepared depth values rather than computing and transmitting full depth maps for each view. This preliminary preparation reduces the bitrate required for transmitting depth information while the structured DLT format maintains coding efficiency.
Solution Approach 2:
The patent uses copying by referencing depth values from the DLT rather than transmitting complete depth maps. Instead of sending redundant depth information for each view, the system copies or references indices into the pre-computed DLT, significantly reducing the amount of data that needs to be transmitted while preserving the essential depth information needed for 3D reconstruction.
2Productivity
If inter-view prediction is not used, then coding process remains simple, but coding efficiency decreases due to redundancy across views
Solution Approach 1:
The patent implements feedback through inter-view prediction where depth information from one view is used to predict depth in another view. The system encodes the difference or residual between the predicted depth and actual depth, creating a feedback loop where previously decoded view information improves the coding of subsequent views. This reduces redundancy and improves coding efficiency by leveraging the temporal and spatial correlations between different views.
Solution Approach 2:
The patent merges depth information across multiple views by using inter-view prediction. Instead of treating each view's depth map independently, the system combines information from multiple views to predict depth values, merging the redundant depth information into a single predictive model that improves overall coding efficiency while reducing the total bitrate required.
3Measurement precision
If DLT is used to represent depth values, then depth coding precision is improved, but bitstream size increases without compression
Solution Approach 1:
The patent applies parameter changes by transitioning from representing depth values as continuous or high-precision floating-point numbers to using integer indices into the DLT. This parameter transformation maintains depth precision because the DLT stores the actual high-precision depth values, while the transmitted bitstream only needs to carry compact integer indices. This changes the representation parameter from high-bit-depth depth values to low-bit-depth indices, reducing bitstream size while preserving precision.
Solution Approach 2:
The patent uses copying by storing the actual high-precision depth values in the DLT structure and copying only the indices into the bitstream. The full precision depth information is copied into the DLT during preprocessing, and then only compact index references are copied into the transmitted data, achieving both precision preservation and compression.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In an example, a method of coding video data includes determining a first depth value of a depth look up table (DLT), where the first depth value is associated with a first pixel of the video data. The method also includes determining a second depth value of the DLT, where the second depth value is associated with a second pixel of the video data. The method also includes coding the DLT including coding the second depth value relative to the first depth value.