Prediction Decoding Method Using Sample Selection Intervals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video encoding standards, such as H.265/HEVC, face challenges in meeting the requirements of emerging video applications like ultra-high-definition video and virtual reality due to high search complexity and reduced prediction decoding performance.
Innovation Solution
A prediction decoding method that determines multiple neighbouring reference samples for a coding block, calculates positions corresponding to K reference samples, and selects second multiple neighbouring reference samples based on sampling intervals to improve prediction decoding performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If a linear model-based prediction method is used to predict chroma component from luma component, then compression performance is improved by about 50%, but search complexity increases and prediction decoding performance deteriorates
Solution Approach 1:
The patent segments the reference sample selection process into two stages: first selecting candidate reference samples from neighboring rows or columns, then selecting final reference samples based on sampling intervals. This segmentation reduces the search complexity by systematically narrowing down the candidate set while maintaining prediction accuracy.
Solution Approach 2:
The patent applies different selection criteria for different regions of the reference samples. By calculating K values based on the length of top or left sides of coding blocks and using preset proportions, the method selectively uses reference samples from specific locations (neighboring rows or columns) rather than uniformly treating all reference samples equally, thereby reducing search complexity while maintaining local prediction quality.
2Measurement precision
If all neighbouring reference samples are used for prediction, then prediction accuracy is maximized, but search complexity increases
Solution Approach 1:
The patent uses partial action by selecting only K reference samples based on sampling intervals rather than using all neighboring reference samples. The K value is calculated using preset proportions and the length of coding block sides, ensuring sufficient reference samples for accurate prediction while significantly reducing the search complexity by excluding redundant reference samples.
Data Source
AI summary
A prediction decoding method, a prediction decoding apparatus, and a non-transitory computer storage medium are provided. The method includes the following. First multiple neighbouring reference samples of a coding block are determined, where the first multiple neighbouring reference samples include reference samples in a reference row or a reference column neighbouring the coding block. Positions corresponding to K reference samples are determined from a starting position of the reference row or the reference column, where K is a positive integer. Second multiple neighbouring reference samples are determined according to the determined positions corresponding to the K reference samples, where the second multiple neighbouring reference samples include neighbouring reference samples among the first multiple neighbouring reference samples. Prediction decoding is performed on the coding block according to the second multiple neighbouring reference samples.