Motion Vector Detection in Distorted Omnidirectional Images

Solution Overview

Problem

Existing image processing techniques for omnidirectional cameras fail to accurately detect motion vectors in distorted hemispherical field-of-view images, leading to inaccuracies in shake correction due to varying enlargement and movement direction across the image field.

Innovation Solution

An image processing apparatus that generates distortion-corrected images for each partial area using position-dependent conversion parameters, detects motion vectors by comparing these images, and performs inverse coordinate conversion to generate accurate motion vectors, effectively addressing the distortion and enlargement issues.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Shape

If distortion correction is performed on hemispherical field-of-view images, then the images can be converted to normal perspective images, but the enlargement amount of the subject image varies in accordance with position and the movement direction does not match the actual movement

Engineering Contradiction:

Improveimage distortionVSAvoidmotion vector accuracy

Core Design Contradiction:

ShapeVSMeasurement precision

Solution Approach 1:

The image is divided into multiple regions (central region and peripheral regions), and motion vectors are calculated separately for each region. This segmentation allows the system to handle the varying enlargement amounts and movement directions in different parts of the image independently, thereby maintaining motion detection accuracy despite distortion correction.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different processing methods are applied to different regions of the image. The central region uses one motion vector calculation approach while peripheral regions use another approach that accounts for the varying enlargement and movement direction characteristics specific to those regions.

Inventive Principle:

Principle #3Local quality

2Ease of operation

If local motion vectors are detected in distorted hemispherical field-of-view images, then block matching processing can be performed, but the accuracy of block matching deteriorates as the subject moves due to varying enlargement and movement direction

Engineering Contradiction:

Improveblock matching processingVSAvoidblock matching accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The image is divided into multiple regions (central region and peripheral regions), and motion vectors are calculated separately for each region. This segmentation allows the system to handle the varying enlargement amounts and movement directions in different parts of the image independently, thereby maintaining motion detection accuracy despite distortion correction.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different processing methods are applied to different regions of the image. The central region uses one motion vector calculation approach while peripheral regions use another approach that accounts for the varying enlargement and movement direction characteristics specific to those regions.

Inventive Principle:

Principle #3Local quality

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach enables the generation of more appropriate motion vectors, improving the accuracy of shake correction and movement detection in distorted images, thereby enhancing the overall image processing and stabilization in omnidirectional camera systems.

Implementation Method 1

a first captured image and a second captured image generated by photo-electrically converting a subject image formed by an optical system

Methodology Applied

Scientific EffectPhotoelectric conversion: Photoelectric Effect

Data Source

PatentUS10038847B2Image processing technique for generating more appropriate motion vectors for a captured image that is distorted

Publication Date: 2018.07.31 CANON KK

AI summary

A generation unit generates N distortion corrected images (N≥2) by performing coordinate conversion on each of N partial areas of a first captured image, using a conversion parameter that depends on a position of the partial area, and generates N distortion corrected images by performing coordinate conversion on each of corresponding N partial areas of a second captured image, using a conversion parameter that depends on a position of the partial area. A detection unit detects N motion vectors by comparing each of the N distortion corrected images generated from the first captured image with the corresponding distortion corrected image generated from the second captured image. A conversion unit generates N converted motion vectors by performing inverse coordinate conversion on each of the N motion vectors using a conversion parameter that depends on a position of the corresponding partial area.