Video Intra Prediction Using Multi-Directional Reference Samples

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video signal processing technologies face challenges in efficiently performing intra prediction during encoding and decoding, particularly when utilizing right and bottom reference samples and considering both forward and inverse directions of the intra prediction mode.

Innovation Solution

A method and apparatus for video signal encoding and decoding that derive first reference samples from the top and left of a current block, and second reference samples from the right and bottom, using temporary prediction samples generated by temporary intra prediction. The prediction sample for the current block is obtained based on a weighted sum operation of the first and second reference samples, with weights determined by the positions of the prediction sample, reference samples, and the intra prediction mode.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If only top and left reference samples are used for intra prediction, then the device complexity is low, but the prediction accuracy is insufficient

Engineering Contradiction:
Improveprediction accuracyVSAvoiddevice complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The reference samples are segmented into two groups: first reference samples from top and left boundaries, and second reference samples from right and bottom boundaries. This segmentation allows the patent to utilize multiple reference sample sources without requiring a completely new prediction structure, thereby improving prediction accuracy while controlling complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extends the traditional intra prediction by adding spatial dimensions - utilizing not only top and left reference samples (conventional approach) but also right and bottom reference samples (new dimension). This multi-directional approach improves prediction accuracy by considering gradients from all four boundaries of the current block.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Measurement precision

If temporary intra prediction is performed to generate second reference samples, then the prediction accuracy is improved, but the computational complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

Temporary intra prediction is performed in advance to generate temporary prediction samples, which are then used to derive the second reference samples. This preliminary action allows the patent to prepare high-quality reference samples before the final prediction, improving overall prediction accuracy while organizing the computational workload efficiently.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

Temporary prediction samples serve as an intermediary between the available reference samples and the final prediction result. The patent uses these intermediate values to derive the second reference samples, which then participate in the gradient calculation and final prediction, thereby improving accuracy through a multi-stage process.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Measurement precision

If both forward and inverse directions of intra prediction mode are considered, then the prediction accuracy is improved, but the processing time increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The prediction process is segmented into two directional components: forward direction (top to bottom, left to right) and inverse direction (bottom to top, right to left). By segmenting the reference sample usage into these two directions, the patent can calculate gradients from both directions and combine them, improving prediction accuracy while maintaining structured processing that controls time consumption.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentUS20250184531A1Video signal processing method and device involving modification of intra predicted sample
Publication Date: 2025.06.05 KT CORP
  • US20250184531A1 patent drawing
  • US20250184531A1 patent drawing
  • US20250184531A1 patent drawing

AI summary

Provided is an image decoding method. The method may include: deriving first reference samples located at the upper end and on the left side of a current block; deriving second reference samples located on the right side and at the lower end of the current block; and acquiring a prediction sample for the current block based on the first and second reference samples, wherein the second reference samples are derived based on the first reference samples and temporary prediction samples generated by performing temporary intra prediction on the current block based on a temporary intra prediction mode.