Selective Sub-Pixel ITMP Merge Candidates for Video Decoding

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing video coding systems face challenges in efficiently utilizing sub-pixel precision for intra template matching prediction (ITMP) merge candidates, leading to suboptimal compression and decoding performance.

Innovation Solution

Implementing a video coding device that determines neighboring coding units with sub-pixel precision, adds ITMP merge candidates to a list, ranks and refines them using rate-distortion optimization (RDO) cost, and selects a subset for final decoding, incorporating both integer and sub-pixel candidates for improved prediction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If sub-pixel precision ITMP merge candidates are added to improve prediction accuracy, then decoding accuracy is improved, but computational complexity increases

Engineering Contradiction:

Improvedecoding accuracyVSAvoidcomputational complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies partial action by selectively adding only certain sub-pixel precision ITMP merge candidates to the candidate list based on specific conditions (e.g., when neighboring CUs are coded with sub-pel precision). This avoids the excessive computation of evaluating all possible candidates while still capturing the beneficial sub-pixel precision cases, thus resolving the contradiction between decoding accuracy and computational complexity.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The patent implements local quality by applying sub-pixel precision ITMP merge candidate addition only in specific local contexts where neighboring CUs are coded with sub-pel precision. Rather than uniformly applying sub-pixel precision across all blocks, the method adaptively applies it only where beneficial, improving decoding accuracy in relevant regions while limiting overall computational complexity.

Inventive Principle:

Principle #3Local quality

2Productivity

If multiple ITMP merge candidates with sub-pixel precision are evaluated using RDO cost ranking, then compression efficiency is improved, but processing time increases

Engineering Contradiction:

Improvecompression efficiencyVSAvoidprocessing time

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The patent segments the candidate evaluation process by first identifying and adding sub-pixel precision ITMP merge candidates that meet specific criteria, then performing RDO cost ranking only on this refined subset. This segmentation avoids the time-consuming evaluation of all possible candidates while still achieving improved compression efficiency through targeted RDO optimization on the most promising candidates.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentEP4633140A1ITMP merge candidate with sub-pixel or frac-pixel precision

Publication Date: 2025.10.15 INTERDIGITAL CE PATENT HOLDINGS SAS

Figure 1A
Figure 1B
Figure 1C

AI summary

Features are described herein for intra template matching prediction (ITMP) merge candidates with subpixel precision. An example device may determine that a neighboring coding unit (CU) is coded with sub-pel precision. Based on the determination that the neighboring CU is coded with sub-pel precision, the device may add an ITMP merge candidate associated with the neighboring CU to an ITMP merge candidate list that includes a sub-pel candidate. The device may rank the ITMP merge candidate list to generate a ranked ITMP candidate list. The device may select a subset of the ranked ITMP candidate list that includes at least the sub-pel candidate. The device may rank the subset of the ranked ITMP candidate list to generate a final list of ITMP candidates. The device may decode a current block using one of the ITMP candidates of the final list of ITMP candidates.