Panoramic Image Stitching Using Depth Mapping to Eliminate Ghosting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing panoramic image stitching methods suffer from image ghosting at the junction of fields of view and either require large safety distances for fast stitching or fail to account for inherent parallax, leading to poor stitching effects, especially in virtual reality scenarios.
Innovation Solution
Estimate parallax information between panoramic images, generate depth information, and map images into the same device space using the depth information to eliminate ghosting and ensure accurate display of close-range objects without cropping.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional panoramic image stitching methods are used to quickly stitch images, then stitching speed is improved, but image ghosting occurs at the junction of fields of view and stitching precision deteriorates
Solution Approach 1:
The patent introduces depth information as an intermediary element to mediate between multiple panoramic images. By estimating depth values for pixels in different panoramic images and using these depth values as weighting factors during fusion, the system achieves both fast stitching and high precision without requiring complex manual alignment or cropping procedures.
Solution Approach 2:
The patent changes the parameter space by introducing depth information estimation and using depth-based weighting instead of traditional geometric alignment parameters. This allows the system to resolve parallax effects and achieve accurate stitching of close-range objects while maintaining fast processing speeds through automated depth-guided fusion.
2Productivity
If large safety distances are used for fast stitching, then stitching speed is improved, but the ability to display close-range objects deteriorates
Solution Approach 1:
Depth information serves as an intermediary that enables the system to handle close-range objects effectively. By estimating depth values and using them for weighted fusion, the system can accurately stitch images of close-range objects without requiring large safety distances, thus maintaining both fast stitching speed and accurate close-range object display.
3Device complexity
If traditional stitching methods are used without considering parallax, then device complexity is reduced, but stitching quality deteriorates due to image ghosting
Solution Approach 1:
The patent introduces depth information as an intermediary that automatically handles parallax effects. This approach maintains relatively simple device architecture while significantly improving stitching quality, as the depth-based weighting mechanism naturally resolves parallax-induced ghosting without requiring complex manual intervention or additional hardware.
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
Generates panoramic images without ghosting and maintains image details, providing an immersive experience by accurately displaying close-range objects.
Implementation Method 1
determining parallax information between the first panoramic image and the second panoramic image
Data Source
AI summary
Provided in the present disclosure are a panoramic image processing method and apparatus, and an electronic device and a storage medium. The method comprises: acquiring a first image captured by a first image acquisition apparatus and a second image captured by a second image acquisition apparatus, and respectively converting the first image and the second image into a first panoramic image and a second panoramic image; determining parallax information between the first panoramic image and the second panoramic image, and generating depth information corresponding to the parallax information; and on the basis of the depth information, mapping the first image and the second image into the same device space, and according to a mapping result in the same device space, generating a stitched panoramic image of the first image and the second image.