Panorama Image Stitching via Distance-Weighted Feature Registration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing panorama image generation methods face challenges in minimizing distortion, especially at the central overlap region, due to inadequate feature weighting and transform estimation techniques.

Innovation Solution

Assigning weights to image features based on their distance from the central portion of the image and using these weights to determine transforms between overlapping images, selecting matching features with higher weights for accurate registration and transformation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If uniform weights are assigned to all image features during panorama generation, then the processing is simple and fast, but distortion occurs at the central overlap region

Engineering Contradiction:
Improveprocessing simplicityVSAvoidimage stitching accuracy
Core Design Contradiction:
Ease of manufactureVSManufacturing precision

Solution Approach 1:

The patent applies local quality by assigning different weights to image features based on their distance from the center of the image. Features closer to the center are assigned higher weights, while features at the periphery receive lower weights. This localized weighting strategy resolves the contradiction by maintaining processing simplicity through an automated algorithm while significantly improving stitching accuracy in the central overlap region, where visual quality is most critical to human perception.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent changes the parameter of feature weights from uniform to distance-dependent values. By introducing a weighting function that varies with radial distance from the image center, the system transforms the feature selection process to prioritize central regions. This parameter change enables the system to achieve high stitching precision in critical areas without substantially increasing computational complexity, as the weighting can be computed efficiently for each feature.

Inventive Principle:
Principle #35Parameter changes

2Reliability

If all matching features are used for transform estimation, then the registration is comprehensive, but errors and noise increase in the final panorama

Engineering Contradiction:
Improveregistration completenessVSAvoidfeature matching accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by selectively emphasizing matching features based on their spatial location. Instead of treating all features equally, the system assigns higher importance to matches in the central region of the image where accuracy is most critical. This selective approach improves measurement precision by filtering out noisy peripheral matches while maintaining registration completeness through the use of sufficient central matches for robust transform estimation.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent applies partial action by using only a weighted subset of matching features for transform estimation rather than all available features. By focusing computational resources on the most reliable central features and downweighting peripheral features, the system achieves higher accuracy without the need to process every single match, thereby improving measurement precision while maintaining adequate registration through the sufficient number of high-weight matches.

Inventive Principle:
Principle #16Partial or excessive action

3Manufacturing precision

If distortion correction is applied aggressively to minimize overlap region distortion, then central image quality improves, but peripheral image regions become distorted

Engineering Contradiction:
Improvecentral region stitching accuracyVSAvoidperipheral image geometry
Core Design Contradiction:
Manufacturing precisionVSShape

Solution Approach 1:

The patent applies local quality by implementing distortion correction with spatially varying intensity. The weighting function ensures that distortion correction is applied most strongly in the central region where stitching accuracy is critical, while gradually reducing the correction intensity toward the periphery. This localized approach resolves the contradiction by prioritizing central image quality without causing excessive peripheral distortion, as the peripheral regions receive gentler transformation that preserves their geometric integrity.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP2842105B1Method, apparatus and computer program product for generating panorama images
Publication Date: 2017.10.11 NOKIA TECHNOLOGIES OY
  • EP2842105B1 patent drawingFigure 1
  • EP2842105B1 patent drawingFigure 2
  • EP2842105B1 patent drawingFigure 3

AI summary

In accordance with an example embodiment a method,apparatus and computer program product are provided. The method comprise sassigning weights to at least one first feature and at least one second feature. The at least one first feature may be associated with a first image and the at least one second feature may be associated with a second image. The weights are assigned based on a distance of the at least one first feature and the at least one second feature from a central portion of the corresponding one of the first image and the second image. The method further includes registering the first image and the second image based at least on the assigned weights to determine transforms between the first image and the second image. The first image and the second image may be based on the determined transform.