Motion Vector Detector Using Hierarchical Search and Gradient Correction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing motion vector detection methods, such as block matching and gradient techniques, face limitations in accurately detecting large motions on a pixel-by-pixel basis and correcting errors in hierarchical searches, leading to incomplete or inaccurate motion vector detection.

Innovation Solution

A motion vector detector comprising a searcher, resolution converter, synthesizer, and corrector that searches for high correlation areas in blocks, adjusts resolution, optimizes gradient-technique objective functions, and corrects motion vectors across different resolutions to achieve accurate pixel-by-pixel detection and error correction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If block matching method is used to detect motion vectors, then large motions can be detected in a stable manner, but motion vectors cannot be detected on a pixel-by-pixel basis

Engineering Contradiction:

Improvemotion vector detection precisionVSAvoiddetection resolution

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent segments the motion detection process into two distinct stages: block matching for coarse motion estimation and gradient technique for fine pixel-level refinement. This segmentation allows each method to operate in its optimal range, with block matching handling large motions and gradient technique providing pixel-level precision, thereby resolving the contradiction between detection precision and resolution.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces a hierarchical dimension to the motion detection process by combining results from two different detection methods. The block matching results serve as initial values for the gradient technique, creating a multi-dimensional approach where coarse and fine detections are integrated, enabling both large motion stability and pixel-level precision simultaneously.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Productivity

If gradient technique is used to detect motion vectors on a pixel-by-pixel basis, then pixel-level motion vectors can be detected, but large motions cannot be detected

Engineering Contradiction:

Improvedetection resolutionVSAvoidlarge motion detection accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by using block matching to obtain initial motion vector estimates before applying the gradient technique. These preliminary results serve as starting points that guide the subsequent pixel-level refinement process, ensuring that large motions are captured first and then refined to pixel-level precision, thus resolving the limitation of gradient technique alone.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent changes the initial parameter values for the gradient technique by setting them to the block matching results. This parameter change allows the gradient technique to start from a reasonable estimate rather than zero, enabling it to converge to accurate pixel-level motion vectors even for large motions, thereby overcoming the gradient technique's inherent limitation.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If hierarchical search with block matching is used from low to high resolution, then processing can be efficient, but false detection errors cannot be corrected in subsequent levels

Engineering Contradiction:

Improveprocessing efficiencyVSAvoiddetection accuracy

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent implements feedback by using the gradient technique to refine and correct block matching results at each hierarchical level. The pixel-level motion vectors obtained through gradient optimization serve as feedback to correct false detections from block matching, ensuring that errors do not propagate to higher resolution levels, thus maintaining both efficiency and reliability.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The gradient technique acts as an intermediary between block matching results and final motion vectors. It mediates the transition by refining the coarse block-level estimates into accurate pixel-level vectors, correcting false detections in the process, and enabling the hierarchical search to maintain both processing efficiency and detection reliability across all resolution levels.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS9582893B2Motion vector detector, distance detector, motion vector detection method and computer program product

Publication Date: 2017.02.28 KK TOSHIBA

AI summary

According to an embodiment, a motion vector detector includes a searcher, a converter, a synthesizer, and a corrector. The searcher is configured to search for, in each block in a first image having a resolution of interest, an area having a high degree of correlation with a second image having the resolution of interest, and generate a first flow representing a set of motion vectors having the resolution of interest. The converter is configured to generate a second flow representing a set of motion vectors having the resolution of interest from a set of motion vectors having a higher resolution. The synthesizer is configured to generate a third flow representing a set of motion vectors having the resolution of interest. The corrector is configured to generate a set of motion vectors having the resolution of interest by optimizing gradient objective function having initial value set to the third flow.