Quantization Rescaling for DCT Coefficient Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video and image compression standards face challenges in efficiently quantizing discrete cosine transform (DCT) coefficients without the overhead of sending time-varying quantization matrices, especially when dynamic adjustment is not possible, leading to suboptimal compression performance.
Innovation Solution
A novel rescaling method is employed in the encoding process, where thresholds are defined for quantizing DCT coefficients, and non-zero coefficients are set to desired reconstructed values using a uniform quantization step size, allowing for standard compliant bitstreams that can be decoded to achieve improved quality without an explicit quantization matrix.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If quantization matrices are used to achieve differentiated quantization for different DCT coefficients, then compression performance and visual quality are improved, but bit rate overhead increases due to transmitting time-varying quantization matrices
Solution Approach 1:
The patent extracts only the essential quantization information needed for differentiated quantization by using a single quantization parameter (QP) to control the quantization step size for all coefficients, while deriving individual quantization matrices at the decoder side through scaling operations. This eliminates the need to transmit full quantization matrices, removing the bit rate overhead while preserving compression performance.
Solution Approach 2:
Instead of transmitting actual quantization matrices, the patent creates copies or derived versions of a base quantization matrix at the decoder side by applying scaling factors based on the transmitted QP value. This allows the decoder to reconstruct the necessary quantization matrices locally without receiving them in the bitstream, thus avoiding the associated bit rate overhead.
2Loss of information
If uniform quantization is applied to all DCT coefficients, then bit rate overhead is reduced, but compression performance deteriorates due to inability to differentiate between important and less important coefficients
Solution Approach 1:
The patent applies local quality by allowing different quantization step sizes for different DCT coefficients through the use of quantization matrices that are derived at the decoder side. Although a single QP parameter is transmitted, it is used to scale the quantization steps differently for various frequency components, enabling important coefficients (e.g., low-frequency) to be quantized with finer precision while less important coefficients (high-frequency) use coarser quantization, thus improving compression performance without increasing bit rate overhead.
3Adaptability or versatility
If dynamic quantization matrices are adjusted for varying input signals, then adaptation to different content is improved, but device complexity increases when dynamic adjustment is not possible
Solution Approach 1:
The patent changes the parameter representation from transmitting full quantization matrices to transmitting a single quantization parameter (QP) that controls the scaling of quantization steps. This parameter change simplifies the encoder complexity significantly while maintaining adaptability, as the QP value can be dynamically adjusted based on input signal characteristics, and the decoder derives the appropriate quantization matrices from this single parameter through scaling operations.
Data Source
AI summary
A method, system, and program product for quantizing discrete cosine transform coefficients, e.g., for MPEG compression, with minimal bit rate overhead and without using a quantization matrix. This is done by scaling a uniform quantization parameter for the entire discrete cosine transform block, defining a variety of thresholds for the quantization of discrete cosine transform coefficients below which the corresponding coefficient will be quantized to zero, and setting different normative reconstructed values for coefficients that have not been quantized to zero as the decoder will still use the original, unmodified reconstructed values as long as the corresponding coefficient is not zero.