Adaptive Color Quantization Video Bitstream
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High color depth in digital camera images leads to high bitrate and processing burdens during video encoding and storage, exceeding human visual perception capabilities, necessitating efficient quantization methods to balance resource usage and image quality.
Innovation Solution
Adaptive non-uniform quantization of color levels based on the capabilities of the decoder-side screen, using a look-up table to adjust quantization levels and distribute bits more efficiently, allowing for greater detail preservation in relevant components while reducing detail in less relevant ones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high color depth (10-12 bits per channel) is used to capture and encode video, then image quality and color accuracy are improved, but bitrate and processing burden increase significantly
Solution Approach 1:
The patent applies different quantization levels to different color components (Y, U, V) based on human visual sensitivity. The luminance component Y uses finer quantization steps while chrominance components U and V use coarser steps, optimizing the balance between quality and bitrate by matching encoding precision to perceptual importance
Solution Approach 2:
The patent dynamically adjusts quantization parameters based on the capabilities of the decoder-side screen. By detecting screen color depth capabilities and adapting the quantization table accordingly, the system optimizes bitrate usage while maintaining acceptable quality for the specific display device
2Measurement precision
If high color depth (10-12 bits per channel) is used for video encoding, then color accuracy is improved, but memory resources required for storage increase
Solution Approach 1:
The patent reduces memory requirements by applying non-uniform quantization where chrominance components (U, V) are quantized with coarser steps than luminance (Y). This selective reduction preserves perceptually important information while discarding less significant data, thereby reducing storage requirements without noticeably impacting perceived quality
Solution Approach 2:
The patent adapts quantization parameters based on screen capabilities to optimize the balance between quality and memory usage. By matching the encoded bit depth to the actual display capabilities, the system avoids storing excessive precision that would not be visible on the target device
3Ease of operation
If uniform quantization is applied to all color components, then encoding simplicity is maintained, but perceptible distortion increases in visually sensitive regions
Solution Approach 1:
The patent uses different quantization step sizes for different color components based on human visual sensitivity. The luminance component Y uses finer quantization steps (e.g., step size 1) while chrominance components U and V use coarser steps (e.g., step size 2), reducing perceptible distortion in visually sensitive luminance regions while maintaining encoding efficiency
Solution Approach 2:
The patent segments the color space into different components (Y, U, V) and applies different quantization strategies to each segment. This segmentation allows tailored quantization approaches for each component based on its perceptual importance, optimizing the balance between simplicity and quality
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A video bitstream is received from an encoder, comprising encoded image portions each having a common form representing components of a channel in a color space. Each of a plurality of the encoded image portions comprises a different set of quantized values of the components, including at least a value of a first of the components being quantized from amongst a first scheme of quantized levels. Further, the received bitstream comprises a look-up table mapping the quantized levels of the first scheme to at least partially de-quantized respective levels. A de-quantizer uses the look-up table received in the bitstream to at least partially de-quantize the different quantized values of the first component in a plurality of the image portions, by converting the quantized values of the first component to at least partially de-quantized values corresponding to ones of the at least partially de-quantized levels of the first scheme.