Dependent Quantization Bit Hiding for Video Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression technologies face inefficiencies in quantization processes, particularly with conventional independent scalar quantization and HEVC's Rate-Distortion Optimized Quantization (RDOQ), leading to suboptimal coding efficiency and increased computational complexity.
Innovation Solution
Implementing Dependent Quantization (DQ) with more than 8 quantization states, using a state machine to determine quantizer selection based on previous quantization states, and implicitly encoding/decoding one bit of information to enhance coding efficiency and reduce bitrate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional independent scalar quantization is used, then the quantization process is simple, but coding efficiency is suboptimal
Solution Approach 1:
The patent implements dynamic quantization state selection using a state machine that adapts between different quantization states (0-7) based on previous quantization outcomes. This dynamic approach allows the system to optimize coding efficiency by selecting appropriate quantization states for different coefficient groups, rather than using a fixed independent scalar quantization approach.
Solution Approach 2:
The patent changes the quantization parameter by introducing multiple quantization states (0-7) with different step sizes. Each state corresponds to a different quantization precision level, allowing the system to adaptively adjust quantization parameters based on content characteristics and rate-distortion optimization, thereby improving coding efficiency.
2Productivity
If Rate-Distortion Optimized Quantization (RDOQ) is used, then coding efficiency improves, but computational complexity increases
Solution Approach 1:
The patent segments the transform coefficients into groups and processes each group independently through a state machine. This segmentation approach divides the complex RDOQ problem into smaller, more manageable units, reducing overall computational complexity while maintaining coding efficiency benefits through group-level optimization.
Solution Approach 2:
The patent applies a simplified version of RDOQ by using a state machine with predefined quantization states rather than exhaustive rate-distortion optimization. This partial action approach captures the essential benefits of adaptive quantization without the full computational burden of complete RDOQ, achieving a balance between complexity and performance.
3Productivity
If more than 8 quantization states are used, then coding efficiency enhances, but one additional bit must be encoded
Solution Approach 1:
The patent merges the quantization state information with the coefficient group data by implicitly encoding the state through the quantization process itself. The state machine's current state (0-7) determines the quantization step size, and this state information is embedded in the coefficient representation, allowing the system to convey state information without requiring separate explicit signaling for most cases.
Solution Approach 2:
The quantization state information serves a dual purpose: it controls the quantization process and simultaneously conveys information about the coefficient group characteristics. The state machine's internal state automatically provides the necessary information for both quantization and potential implicit signaling, making the system self-sufficient and avoiding additional bitrate overhead.
4Productivity
If Dependent Quantization with state machine is used, then quantization optimization improves, but device complexity increases
Solution Approach 1:
The patent implements a dynamic state machine that transitions between 8 quantization states (0-7) based on previous quantization outcomes. This dynamic mechanism allows the system to adapt quantization parameters in real-time, improving optimization performance while maintaining a manageable state space that limits computational overhead.
Solution Approach 2:
The state machine changes quantization parameters (step sizes) based on the current state, providing adaptive optimization. The parameter changes are confined to a discrete set of 8 states, which balances optimization capability with computational tractability, avoiding the complexity of continuous parameter adjustment.
Data Source
Figure 1A~1B
Figure 1C
Figure 2
AI summary
A method and an apparatus for encoding or decoding a video are provided wherein for at least one group of samples of a block of the video, values of samples of the at least one group are quantized using a plurality of quantizers assigned respectively to a plurality of quantization states of a dependent quantization, and wherein the plurality of quantization states comprises more than 8 quantization states. One bit of information is implicitly encoded based on a quantization state of a last sample of the at least one group in an encoding order.