Layered Video Coding with High-Frequency Interlayer Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing layered video coding schemes suffer from the elimination of high-frequency components during base-layer spatial decimation, leading to insufficient predictive coding and increased signal levels in enhancement-layer frames, especially in areas with continuous high-frequency components.
Innovation Solution
A method that includes spatial interpolation and high-frequency components estimation and scaling to upscale decoded signals from lower coding layers, allowing for accurate interframe predictive coding with reduced data amounts by subtracting the upscaled signals from the input video signal to produce a predictive-error signal for encoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If base-layer spatial decimation is employed to produce base-layer frames, then spatial resolution is reduced and bandwidth is saved, but high-frequency components are eliminated leading to insufficient predictive coding
Solution Approach 1:
The patent segments the predictive signal into two distinct components: a low-frequency component obtained through base-layer spatial decimation and upscaling, and a high-frequency component obtained through high-frequency component extraction and upscaling. This segmentation allows each component to be processed appropriately, with the low-frequency component providing the base structure and the high-frequency component restoring the lost detail, thereby resolving the contradiction between bandwidth savings and predictive coding accuracy.
Solution Approach 2:
The patent applies different processing methods to different frequency components of the predictive signal. The low-frequency components are handled through standard spatial decimation and upscaling, while the high-frequency components are handled through specialized extraction and restoration processes. This local quality approach ensures that each frequency range is treated with the most appropriate method, maintaining overall predictive coding accuracy while managing bandwidth efficiently.
2Productivity
If predictive coding is performed with eliminated high-frequency components, then coding is still possible, but signal level and data amount increase especially in zones with continuous high-frequency components
Solution Approach 1:
The patent extracts the high-frequency components from the predictive signal separately from the low-frequency components. By taking out the high-frequency components through high-frequency component extraction and processing them through upscaling, the patent prevents these components from increasing the overall signal level and data amount in the enhanced-layer predictive signal, while still maintaining coding feasibility.
3Quantity of substance
If base-layer spatial decimation is used, then bandwidth is reduced, but high-frequency components are lost resulting in inaccurate enhancement-layer predictive frames
Solution Approach 1:
The patent performs preliminary action by extracting and processing the high-frequency components of the predictive signal before they are needed for enhancement-layer coding. The high-frequency component extraction and upscaling operations are performed in advance, creating a restored high-frequency component that can be combined with the upscaled base-layer signal to form an accurate predictive frame, thereby ensuring reliability before the actual coding process.
Data Source
AI summary
An input video signal is encoded at a plurality of coding layers exhibiting different spatial resolutions. Decoded is a given signal coded at a lower coding layer lower than a specific coding layer among the plurality of coding layers to generate a decoded signal of the lower coding layer. Spatial interpolation is applied to the decoded signal of the lower coding layer to generate an upscaled decoded video signal. The spatial interpolation is an upscaling procedure to upscale the decoded signal of the lower coding layer into a spatial resolution of the specific coding layer. A spatial high-frequency components estimation and scale up procedure is applied to the decoded signal of the lower coding layer to generate a high-frequency components signal. The upscaled decoded video signal and the high-frequency components signal are subtracted from the input video signal exhibiting a spatial resolution of the specific coding layer to produce a predictive-error signal. Then, the predictive-error signal is encoded at the specific coding layer to generate an output bitstream. The decoding, spatial interpolating, spatial high-frequency components estimating and upscaling, and encoding procedures are executed at, at least, the specific coding layer among the plurality of coding layers except the lowest coding layer exhibiting the lowest spatial resolution thereamong.