Panoramic Image Stitching Ghosting Elimination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing panoramic image stitching methods and apparatuses suffer from the ghosting phenomenon due to differing sight lines of adjacent lenses in multi-lens panoramic cameras, which affects the quality of synthesized 360-degree panoramic images.
Innovation Solution
A method and apparatus that perform equirectangular expansion, rotation transformation, and pixel light stream-based stitching position adjustment on partial panoramic images to align and stitch adjacent images effectively, eliminating the ghosting issue.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If multiple lenses are set close together (25-55 mm) to ensure image quality, then the device size is reduced and image coverage is improved, but ghosting phenomenon occurs during stitching due to different sight lines of adjacent lenses
Solution Approach 1:
The patent applies preliminary action by performing equirectangular expansion and rotation transformation on each partial panoramic image before stitching. This preprocessing aligns the sight lines of adjacent lenses in advance, so that when the images are stitched together, the ghosting phenomenon is eliminated because the overlapping regions already match in orientation and projection format
Solution Approach 2:
The patent changes the projection parameters by transforming from the original lens projection format to equirectangular projection format. This parameter change in the projection method allows for consistent alignment of adjacent images while maintaining full 360-degree coverage, resolving the ghosting issue without sacrificing image coverage
2Productivity
If traditional stitching methods are used without position adjustment, then the stitching process is simple and fast, but ghosting occurs at the edges of panoramic images due to misalignment
Solution Approach 1:
The patent performs rotation transformation and stitching position adjustment as preliminary actions before the final stitching operation. By pre-aligning the images through rotation and position adjustment based on pixel light streams, the actual stitching process becomes a simple merging operation that is both fast and precise, eliminating ghosting without sacrificing productivity
Solution Approach 2:
The patent uses pixel light streams as feedback to determine the optimal stitching position. By analyzing the light stream information from adjacent images, the system automatically adjusts the stitching position to achieve perfect alignment, ensuring high precision while maintaining efficient processing through automated feedback-based adjustment
Data Source
AI summary
The present disclosure provides a panoramic image stitching method, including the steps of acquiring a partial panoramic image taken by each lens; acquiring a first equirectangular rectangular projection graph of each partial panoramic image; performing rotation transformation on the first equirectangular rectangular projection graph of each partial panoramic image to obtain a second equirectangular rectangular projection graph; and adjusting stitching positions of adjacent second equirectangular rectangular projection graphs, and performing stitching operation on the second equirectangular rectangular projection graphs according to the adjusted stitching positions.