Decoder Quantization Offset Derivation for Adaptive Block Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in optimizing quantization processes to improve image quality, particularly in deriving quantization shifting offsets that are content-dependent and effective at the decoder side.
Innovation Solution
The proposed solution involves deriving quantization shifting offsets at the decoder side by generating a new predictor based on neighboring blocks, performing inverse transforms, and adjusting dequantized blocks using these offsets to enhance image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If quantization is performed with a fixed quantization step size, then the encoding process is simple and fast, but the image quality and reconstruction accuracy deteriorate
Solution Approach 1:
The patent applies dynamics by transitioning from a fixed quantization step size to a dynamic, content-adaptive quantization process. The quantization step size is adjusted based on the local characteristics of the video block (e.g., variance, gradient, or texture complexity), allowing the system to use finer quantization for important regions and coarser quantization for less critical regions, thereby improving overall image quality while maintaining encoding efficiency.
Solution Approach 2:
The patent changes the quantization parameter (step size) based on content characteristics. By analyzing local features such as block variance, edge density, or texture complexity, the system dynamically selects appropriate quantization step sizes for different regions, optimizing the balance between compression ratio and reconstruction quality.
2Manufacturing precision
If a single predictor is used for the current block, then the prediction process is simple, but the prediction accuracy and reconstruction quality worsen
Solution Approach 1:
The patent merges multiple prediction methods by combining intra-prediction (spatial prediction from neighboring blocks within the same picture) and inter-prediction (temporal prediction from reference pictures). The system generates both predictors and selects or combines them based on rate-distortion optimization, improving prediction accuracy while maintaining manageable complexity through systematic comparison and selection.
3Manufacturing precision
If quantization shifting offset is not applied, then the decoding process is simple and fast, but the quantization error and image quality deterioration increase
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing quantization shifting offsets during the encoding phase or in advance during decoding setup. These offsets are derived from the quantized coefficients and are prepared beforehand, allowing the decoder to quickly apply them during reconstruction without performing complex real-time calculations, thus improving reconstruction accuracy while maintaining decoding speed.
Data Source
AI summary
Processing circuitry performs a dequantization on a first quantized block that is decoded from the coded information of the current block to generate a dequantized block. The processing circuitry performs an inverse transform on the dequantized block to obtain an inverse transform block, combines the first predictor of the current block with the inverse transform block of the dequantized block to generate an intermediate reconstruction of the current block, generates a second predictor of the current block based on one or more neighboring blocks of the current block, performs a transform on a difference of the intermediate reconstruction of the current block and the second predictor of the current block to generate a transform block, performs a quantization on the transform block to generate a second quantized block, and determines at least a quantization shifting offset based on the dequantized block and the second quantized block.