Mobile Camera Panorama Stitching with Object-Aware Frame Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image stitching processes often select base frames containing low-quality representations of objects of interest, leading to reduced image fidelity and noticeable distortions, particularly when seams are placed over these objects.

Innovation Solution

The process intelligently selects base frames by considering the quality of objects of interest within candidate frames and penalizes seams placed on these regions during the blending phase, ensuring each region of interest is fully contained within a selected base frame and applying a computational bias to minimize seam placement on these areas.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional stitching algorithms select base frames from overlapping regions, then the stitching process is simple and fast, but the quality of objects of interest is degraded due to low-quality representations and seam artifacts

Engineering Contradiction:
Improveimage fidelityVSAvoidframe selection complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by performing frame quality assessment and region of interest detection before the stitching process. The system pre-identifies high-quality frames containing complete regions of interest and pre-calculates seam placement strategies to avoid these regions, thereby ensuring high image fidelity from the outset without compromising stitching simplicity

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements local quality by differentiating between different regions within frames during selection. Instead of evaluating frames uniformly, the system specifically identifies frames where regions of interest are fully contained and of high quality, giving priority to local quality metrics in critical areas while maintaining overall stitching efficiency

Inventive Principle:
Principle #3Local quality

2Reliability

If seams are placed in overlapping regions during blending, then the stitching process is efficient and fast, but objects of interest suffer from distortions and artifacts

Engineering Contradiction:
Improveobject qualityVSAvoidstitching efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent applies local quality by implementing differential seam placement strategies. The system identifies regions of interest within the panorama and specifically avoids placing seams through these regions, even though this requires more complex calculations. This ensures that objects of interest maintain high quality without severe distortions or artifacts

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent uses an intermediary approach by introducing a seam placement optimization layer between the blending process and the final composite image. This intermediary step calculates optimal seam locations that avoid regions of interest, acting as a mediator that preserves object quality while still achieving efficient blending in non-critical areas

Inventive Principle:
Principle #24Intermediary (Mediator)

3Reliability

If all captured frames are used for stitching, then the coverage area is maximized, but the quality of the composite image is reduced due to inclusion of low-quality frames

Engineering Contradiction:
Improvecomposite image qualityVSAvoidpanorama coverage
Core Design Contradiction:
ReliabilityVSArea of stationary object

Solution Approach 1:

The patent implements local quality by evaluating frames based on the quality of specific regions within them rather than overall frame quality. Frames are selected if they contain fully-contained regions of interest at high quality, even if this reduces the total number of frames used, thereby prioritizing local quality over maximum coverage

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent applies taking out by extracting and prioritizing high-quality regions of interest from the captured frames. The system identifies and extracts frames that completely contain regions of interest at high quality, separating these from lower-quality frames that would degrade the composite image, even if it means excluding some coverage areas

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4158585B1Panorama generation with mobile camera
Publication Date: 2025.10.01 GOOGLE LLC
  • EP4158585B1 patent drawingFigure 1
  • EP4158585B1 patent drawingFigure 2
  • EP4158585B1 patent drawingFigure 3

AI summary

Devices and methods for selecting and stitching image frames are provided. A method includes obtaining a plurality of image frames. The method also includes identifying one or more regions of interest within one or more image frames in the plurality of image frames. The method further includes selecting, based on a respective quality measure associated with each image frame of the plurality of image frames, a set of base frames, where each identified region of interest of the one or more identified regions of interest is fully contained within at least one base frame in the selected set of base frames. The method additionally includes stitching together the selected set of base frames to create a composite image.