Partial Reconstructed Pixel Buffer for Video Intra Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video decoding methods require a large picture buffer to store reconstructed pixel data for intra prediction, which is cost-inefficient and inefficient in terms of memory access latency.
Innovation Solution
A video processing apparatus and method that uses a smaller storage device, such as on-chip SRAM, to store partial reconstructed pixel data, allowing for intra prediction of subsequent blocks by generating padded pixels when necessary, thereby reducing the required buffer size and improving efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a picture buffer with capacity not smaller than reconstructed data size of one picture is used for buffering reconstructed pixels, then intra prediction of current block can be performed, but production cost increases and memory resources are wasted
Solution Approach 1:
The patent extracts only the necessary portion of reconstructed pixel data (one row or column of pixels) from the full picture buffer requirement. Instead of buffering the entire reconstructed picture, the invention takes out only the minimal data needed for intra prediction of the current block, thereby reducing memory capacity requirements and production cost while maintaining prediction capability.
Solution Approach 2:
The patent applies local quality by providing different buffering solutions for different spatial locations. The reference buffer stores only the local pixel data adjacent to the current block that is needed for prediction, rather than uniformly buffering the entire picture. This localized approach reduces overall memory requirements while ensuring prediction accuracy where needed.
2Reliability
If a large picture buffer is used to store all reconstructed pixel data, then complete reference data is available for intra prediction, but memory access latency increases due to larger buffer size
Solution Approach 1:
The patent extracts only the essential reference pixels (one row or column adjacent to the current block) from the full picture buffer. By taking out only this minimal necessary data and storing it in a small reference buffer, the invention reduces memory access latency while maintaining complete reference data for intra prediction operations.
Solution Approach 2:
The patent performs preliminary action by pre-fetching and storing the necessary reference pixel data in a small buffer before it is needed for intra prediction. The decoded pixel data is immediately stored in the reference buffer as decoding progresses, making the data readily available when needed for prediction of subsequent blocks, thereby reducing access latency.
3Ease of manufacture
If a small storage device is used to store partial reconstructed pixel data, then production cost decreases and memory access latency is reduced, but the storage capacity is insufficient to hold the entire picture
Solution Approach 1:
The patent extracts only the necessary reference pixels (one row or column) from the full picture data and stores them in a small reference buffer. This extraction approach allows the use of a small storage device with limited capacity while still providing sufficient data for intra prediction operations.
Solution Approach 2:
The patent applies dynamics by making the reference buffer content dynamically update as decoding progresses. The buffer continuously receives new decoded pixel data and provides reference data for prediction, allowing a small storage device to effectively handle the entire picture through dynamic data replacement and reuse.
4Ease of manufacture
If a small storage device is used for buffering reconstructed pixels, then production cost is reduced, but the ability to buffer all needed reference pixels is compromised
Solution Approach 1:
The patent extracts only the critical reference pixels (one row or column adjacent to current block) that are essential for intra prediction. By taking out only this minimal necessary data, the invention enables the use of a small, cost-effective storage device while maintaining complete reference pixel buffering capability for prediction operations.
Solution Approach 2:
The patent performs preliminary action by pre-storing the necessary reference pixel data in the small buffer before it is needed for prediction. As decoding progresses, the reference buffer is continuously updated with new pixel data, ensuring that all needed reference pixels are available when required for intra prediction of subsequent blocks.
Data Source
AI summary
A video processing apparatus includes a reconstruct circuit, a storage device, and an intra prediction circuit. The reconstruct circuit generates reconstructed pixels of a first block of a picture. The storage device at least stores a portion of the reconstructed pixels of the first block, wherein a capacity of the storage device is smaller than a reconstructed data size of the picture. The intra prediction circuit performs intra prediction of a second block of the picture based at least partly on pixel data obtained from the storage device.