Motion Estimation for Non-Natural Video Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional motion estimation techniques designed for natural video data are inefficient when applied to non-natural video data, as they assume error decreases with the best match candidate block, which is not true for non-natural data with sharp transitions and high spatial frequencies, leading to suboptimal performance and increased computational complexity.

Innovation Solution

Implementing a multi-stage approximated error cost computation with early exit mechanisms and a modified initial search in a moving diamond pattern, which focuses on partial samples and adjusts the search area to better capture the characteristics of non-natural video data, reducing computational resources and improving efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional motion estimation techniques are used for non-natural video data, then the error calculation assumes monotonic decrease with best match candidate block, but this leads to suboptimal performance and increased computational complexity

Engineering Contradiction:
Improvemotion estimation accuracyVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The search space is segmented into multiple diamond patterns (first diamond pattern and second diamond pattern) that are horizontally and/or vertically displaced from each other. This segmentation allows the algorithm to divide the computational task into smaller, manageable sections, reducing the overall computational complexity while maintaining accuracy for non-natural video data with sharp transitions

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The algorithm performs preliminary error calculations on partial samples of candidate blocks before completing the full error calculation. By calculating error for only a subset of samples initially, the system can make early decisions about whether to continue with full calculation, thereby reducing computational complexity while preserving reliability

Inventive Principle:
Principle #10Preliminary action

2Productivity

If conventional search patterns are used, then the error calculation covers all samples, but this increases computational resources without improving efficiency for non-natural video data

Engineering Contradiction:
Improvemotion estimation efficiencyVSAvoidcomputational resources
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The algorithm performs partial error calculations on subsets of samples rather than calculating errors for all samples in the candidate block. This partial action approach reduces computational resources consumed while maintaining sufficient accuracy for non-natural video data, thereby improving productivity without excessive energy use

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The candidate block samples are segmented into different groups, and error calculations are performed on specific segments (partial samples) rather than the entire block. This segmentation strategy reduces the number of computations required, improving efficiency while lowering computational resource consumption

Inventive Principle:
Principle #1Segmentation

3Measurement precision

If full error calculation is performed on all candidate blocks, then accurate motion estimation is achieved, but computational complexity increases significantly

Engineering Contradiction:
Improvematching error accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The algorithm performs preliminary error calculations on partial samples before committing to full error calculation. This preliminary action allows the system to identify and eliminate poor candidate blocks early, maintaining measurement precision for promising candidates while reducing computational complexity overall

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

Instead of performing full error calculations on all candidate blocks, the algorithm applies partial calculations to subsets of samples. This approach maintains sufficient measurement precision for identifying the best match while significantly reducing the computational complexity associated with processing all samples of all candidates

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3120557B1Method for motion estimation of non-natural video data
Publication Date: 2020.04.15 QUALCOMM INC
  • EP3120557B1 patent drawingFigure 1A
  • EP3120557B1 patent drawingFigure 1B
  • EP3120557B1 patent drawingFigure 2

AI summary

A method for motion estimation for screen and non-natural content coding is disclosed. In one aspect, the method may include selecting a candidate block of a first frame of the video data for matching with a current block of a second frame of the video data, calculating a first partial matching cost for matching a first subset of samples of the candidate block to the current block, and determining whether the candidate block has a lowest matching cost with the current block based at least in part on the first partial matching cost.