Intra-Prediction Residual Approximation for Encoding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Intra-prediction in MPEG-4 AVC/H.264 is limited by restricted prediction directions, leading to insufficient prediction accuracy and increased encoding bit amounts due to transmission of prediction direction information, which hinders encoding efficiency improvements.
Innovation Solution
An image encoding apparatus and method that generate an intra-prediction signal, calculate residual signals, and perform variable-length encoding, incorporating a residual approximation signal creation unit to linearly approximate the DC component of the residual signal, reducing the information amount of the residual signal and improving encoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If arbitrary prediction direction is used to improve prediction accuracy, then prediction efficiency is improved, but encoding bit amount increases due to transmission of prediction direction information
Solution Approach 1:
The patent extracts only the necessary prediction direction information by limiting prediction directions to only horizontal and vertical directions, eliminating the need to transmit arbitrary angle information. This extraction approach reduces the encoding bit amount while maintaining sufficient prediction accuracy for most cases.
Solution Approach 2:
The patent changes the parameter of prediction direction from continuous arbitrary angles to discrete horizontal and vertical directions only. This parameter discretization reduces the information content that needs to be encoded and transmitted, thereby reducing the encoding bit amount while still providing effective prediction capability.
2Device complexity
If same prediction value is used in single direction, then encoding complexity is reduced, but prediction accuracy is insufficient as change of pixel value along prediction direction cannot be attained
Solution Approach 1:
The patent segments the prediction process into multiple stages: first generating prediction values in horizontal and vertical directions, then creating a residual approximation signal from the residual. This segmentation allows each stage to focus on specific aspects, improving overall prediction accuracy without significantly increasing complexity.
Solution Approach 2:
The patent adds another dimension to the prediction by introducing the residual approximation signal creation step that operates on the residual values. This additional dimensional processing captures the change of pixel values along the prediction direction, thereby improving prediction accuracy while keeping the base prediction mechanism simple.
Data Source
AI summary
An intra-prediction unit generates an intra-prediction signal of an image. A subtraction unit calculates a first residual signal from a difference between an original signal of the image and the intra-prediction signal. A residual approximation signal creation unit creates a residual approximation signal by linearly approximating the first residual signal based on a DC component of the first residual signal. A subtraction unit calculates a second residual signal from a difference between the first residual signal and the residual approximation signal. An orthogonal transformation coefficient integration unit generates a third residual signal by synthesizing the DC component of the first residual signal and the second residual signal. A variable-length encoding unit performs variable-length encoding of the third residual signal.