Position Dependent Reference Sample Smoothing for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges in efficiently compressing and decompressing video sequences due to large data sizes, requiring innovative methods to reduce redundant information and improve compression efficiency.
Innovation Solution
The implementation of advanced video coding/decoding systems that utilize quadtree and multi-type tree partitioning, intra and inter prediction techniques, and adaptive interpolation filters to efficiently encode and decode video sequences, reducing bitrates while maintaining video quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple reference lines are used for video prediction, then prediction accuracy is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent divides the reference picture into multiple reference lines (first reference line, second reference line, etc.) at different distances from the current picture. Each reference line is processed independently with position-dependent smoothing applied selectively based on its distance, breaking down the complex multi-reference processing into manageable segments that can be handled efficiently
Solution Approach 2:
The patent applies position-dependent reference sample smoothing selectively: reference samples from the first reference line (closer to current picture) undergo smoothing processing, while reference samples from the second reference line (farther from current picture) may skip smoothing or apply different smoothing strength. This local differentiation optimizes prediction accuracy for each reference line based on its specific characteristics and distance from the current picture
2Measurement precision
If position-dependent smoothing is applied to reference samples, then prediction accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies smoothing processing partially rather than uniformly to all reference samples. By determining whether to apply smoothing based on the reference line's distance from the current picture and the specific prediction mode being used, the system applies smoothing only where it provides significant benefit, avoiding unnecessary processing time for cases where smoothing would provide minimal improvement
3Adaptability or versatility
If reference samples from multiple distances are used, then prediction flexibility is improved, but data redundancy increases
Solution Approach 1:
The patent changes the parameter of reference sample smoothing strength based on the distance of the reference line from the current picture. Reference lines at different distances undergo different amounts of smoothing, with closer reference lines receiving stronger smoothing and farther reference lines receiving lighter or no smoothing. This parameter adaptation reduces redundancy by tailoring the processing to the specific characteristics of each reference line
Data Source
AI summary
A method includes projecting, in an angular direction, a location of a sample to a first point on a first reference line and a second point on a second reference line. The method includes interpolating between a first and second reference sample on the first reference line to determine an interpolated value at the first point. The interpolating includes applying a first filter coefficient to the first reference sample and a second filter coefficient to the second reference sample. Both the first and the second filter coefficients are determined based on a first function that is related to a first distance from the second point to a closest integer position in a first direction along the second reference line and a second distance from the second point to a closest integer position in a second direction along the second reference line. The method includes determining a prediction of the sample based on the interpolated value.