DPCM Intra Prediction for Video Coding Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression techniques face challenges in achieving high compression ratios with minimal loss in image quality, particularly in intra-frame prediction, where accuracy decreases with larger block sizes, leading to increased bit-rates and residual data.
Innovation Solution
The implementation of differential pulse code modulation (DPCM) intra prediction schemes that predict pixels adaptively based on neighboring pixels, maintaining a constant distance of one between prediction and reference pixels, improving accuracy and coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If intra-frame prediction uses larger block sizes to reduce data quantity, then compression ratio improves, but prediction accuracy decreases leading to increased bit-rate
Solution Approach 1:
The patent segments the prediction process into multiple stages: first generating an initial prediction block from reference pixels, then creating residual blocks by subtracting the initial prediction from actual blocks, and finally refining predictions using the residual information. This segmentation allows large blocks to be processed while maintaining accuracy through iterative refinement.
Solution Approach 2:
The patent implements feedback by using reconstructed pixels from previously decoded blocks as reference pixels for subsequent prediction operations. The residual information is fed back into the prediction process to continuously improve prediction accuracy, creating a closed-loop system that adapts to local image characteristics.
2Productivity
If intra-frame prediction uses larger block sizes to decrease data quantity, then compression efficiency improves, but residual data increases leading to higher bit-rate
Solution Approach 1:
The patent performs preliminary prediction using reference pixels before processing the actual block. By generating an initial prediction block that approximates the target block in advance, the system prepares a starting point that reduces the magnitude of residual data, thereby decreasing the information that needs to be encoded and transmitted.
Solution Approach 2:
The patent discards the redundant information contained in the initial prediction block (which can be regenerated at the decoder) and only encodes the residual difference. This allows the system to discard large amounts of predictable data while recovering the essential information through the residual, significantly reducing bit-rate.
3Measurement precision
If intra prediction accuracy is improved by using closer reference pixels, then prediction precision increases, but the complexity of the prediction process increases
Solution Approach 1:
The patent employs dynamic adaptation by selecting reference pixels based on the actual image content and block characteristics. The prediction process dynamically adjusts which reference pixels to use and how to weight them, allowing the system to achieve high precision without requiring complex fixed algorithms for all scenarios.
Solution Approach 2:
The patent changes the parameters of the prediction process by using reconstructed pixels with updated characteristics as reference points. By modifying the reference pixel values based on previously decoded information, the system achieves higher prediction precision while maintaining relatively simple computational operations.
Data Source
AI summary
A video codec including a memory and a processor operably coupled to the memory. The processor is configured to compute a reconstructed pixel based on a residual pixel and a first prediction pixel and compute a second prediction pixel in a directional intra prediction mode based on the reconstructed pixel, where the first and second prediction pixels are located in a same block of a video frame. A method for intra prediction including computing a prediction pixel adaptively based on a plurality of reconstructed neighboring pixels, where a distance between the prediction pixel and each of the plurality of reconstructed neighboring pixels is one.