Downscaled Residual Blocks for Video Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video encoding and decoding systems face challenges in efficiently compressing and transmitting video sequences due to large data sizes, with existing technologies often resulting in inefficient residual block sizes and increased bitrates.
Innovation Solution
The proposed solution involves downscaling residual blocks to reduce their size, using techniques such as bicubic, bilinear, or average downscaling, and signaling the downscaling indication in the bitstream, allowing for more efficient encoding and decoding by transforming and quantizing the downscaled residual blocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If residual blocks are encoded at full size, then encoding accuracy is maintained, but bitrate increases and compression efficiency decreases
Solution Approach 1:
The patent segments the encoding process into multiple stages: full-resolution encoding for reference frame generation, followed by downscaling residual blocks for subsequent prediction. This segmentation allows the system to capture high-frequency details when needed (in reference frames) while using compressed residuals for efficiency (in subsequent frames), resolving the contradiction between accuracy and bitrate.
Solution Approach 2:
The patent dynamically changes the resolution parameter of residual blocks based on temporal context. Reference frames use full resolution parameters, while subsequent frames use downscaled residual parameters. This parameter adaptation allows the system to maintain encoding accuracy where critical (reference frames) while reducing bitrate where possible (subsequent frames with similar content).
2Quantity of substance
If residual block size is reduced through downscaling, then bitrate decreases and compression efficiency improves, but encoding precision is lost
Solution Approach 1:
The patent performs preliminary encoding at full resolution to generate accurate reference frames before downscaling residuals. By establishing high-precision reference data in advance, the system can afford to use lower-precision downscaled residuals for subsequent frames, as the reference frames provide the foundation for accurate reconstruction. This preliminary high-precision action compensates for later precision reduction.
Solution Approach 2:
The patent creates full-resolution copies of frames as reference data, then uses downscaled residuals to predict subsequent frames against these references. The copy mechanism allows the system to preserve high-precision reference data while using efficient downscaled residuals for prediction, maintaining overall precision while reducing bitrate for the residual data itself.
3Device complexity
If downsampling is applied to residual blocks, then computational complexity is reduced, but processing time for reconstruction increases
Solution Approach 1:
The patent implements periodic full-resolution encoding for reference frames, interspersed with faster downscaled residual encoding for subsequent frames. This periodic pattern creates a rhythm of high-complexity operations (reference frame generation) followed by low-complexity operations (residual prediction), allowing the system to manage computational load while maintaining overall reconstruction accuracy.
Solution Approach 2:
The patent dynamically adjusts the encoding complexity based on frame type and temporal context. Reference frames use full computational complexity for maximum accuracy, while subsequent frames use reduced complexity with downscaled residuals. This dynamic adaptation allows the system to optimize processing time by reducing complexity where full precision is not critical, while maintaining accuracy where needed.
Data Source
AI summary
A decoder may receive, in a bit stream for a current block of samples, a first residual block of samples, an indication of a displacement vector, and an indication of a downscaling. The decoder may, based on the indication of the downscaling, upscale the first residual block of samples to a second residual block of samples. The decoder may generate a decoded block of samples based on: the displacement vector; and the second residual block of samples.