Chroma Sample Array Inter-Component Prediction for Depth View Synthesis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video coding methods face limitations in compressing multiview video plus depth (MVD) due to high correlation between texture and depth components, leading to redundant depth map information and inefficient compression, particularly in view synthesis, where sufficient correlation between sample values of texture and depth pictures has been difficult to find.
Innovation Solution
A method is introduced to derive sample value prediction from a texture picture to a depth picture by decoding chroma components, obtaining a reference sample array, and forming a decoded depth view component using modified chroma sample arrays through quantization, segmentation, and filtering techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional video coding methods are used for MVD, then compression is performed separately for texture and depth components, but compression efficiency is limited due to high correlation between texture and depth components
Solution Approach 1:
The patent merges the coding of texture and depth components by using decoded chroma samples from texture pictures as reference samples for depth picture prediction. This inter-component prediction approach exploits the high correlation between texture and depth components, allowing depth maps to be coded with lower spatial resolution while maintaining compression efficiency.
Solution Approach 2:
The patent makes the chroma sample arrays serve multiple functions: they are used for color representation in texture pictures and simultaneously as reference samples for depth picture prediction. This multi-functionality eliminates redundant coding and improves compression efficiency by leveraging the correlation between components.
2Reliability
If all available depth maps are used in post processing steps, then view synthesis quality is maintained, but bitrate and complexity increase due to redundant depth map information
Solution Approach 1:
The patent extracts only the essential depth information by using lower spatial resolution depth maps for view synthesis. By exploiting the correlation between texture and depth through inter-component prediction, the system can discard redundant high-resolution depth maps while maintaining sufficient quality for synthesis operations.
Solution Approach 2:
The patent applies partial action by using reduced-resolution depth maps instead of full-resolution depth maps for view synthesis. This partial approach is sufficient for the intended application while significantly reducing the quantity of depth data that needs to be transmitted and processed.
3Loss of information
If motion vectors are predicted from texture picture to depth picture, then compression is improved, but sufficient correlation between sample values has been difficult to find
Solution Approach 1:
The patent introduces chroma sample arrays as an intermediary between texture and depth components. Instead of directly predicting depth sample values from texture sample values, the system uses chroma samples as a mediator that captures the correlation information, enabling more effective inter-component prediction.
Solution Approach 2:
The patent changes the parameter being predicted from motion vectors alone to include chroma sample values. By modifying the prediction input from purely geometric (motion vectors) to include color information (chroma samples), the system exploits additional correlation dimensions between texture and depth components.
Data Source
AI summary
A method comprising decoding one or both chroma components of a coded texture picture into one or two decoded chroma sample arrays, obtaining a reference sample array on the basis of said one or two decoded chroma sample arrays, and forming a decoded depth view component on the basis of said reference sample array; encoding one or two chroma components of a texture picture into one or two coded chroma components and reconstructing one or two reconstructed chroma sample arrays, obtaining a reference sample array on the basis of said one or two reconstructed chroma sample arrays, and using the reference sample array as a reference for prediction in encoding of a depth view component.