Image Stitching Using Control Object Isolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image stitching methods struggle to accurately identify and remove overlapped areas between screenshots with non-solid-color chat backgrounds, leading to inaccuracies in the final stitched image.
Innovation Solution
The method involves identifying a common control object and its location information in both images, creating middle images with solid-color backgrounds to isolate the control object, and then stitching the images based on this information to eliminate overlapped content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image comparison method is used to identify overlapped area, then overlapped area can be accurately identified for solid-color chat background, but overlapped area cannot be found when non-solid-color chat background is used
Solution Approach 1:
The patent extracts the control object from the complex background environment by obtaining its location information and creating a middle image that isolates the control object from the original image background. This allows the control object to be identified and compared without interference from varying background patterns.
Solution Approach 2:
The patent introduces a middle image as an intermediary representation between the original images and the comparison process. The middle image contains only the control object with a standardized solid-color background, serving as a mediator that enables accurate comparison while being independent of the original diverse backgrounds.
2Loss of information
If personalized background information is included in image comparison, then complete image content is preserved, but control object identification accuracy deteriorates due to background interference
Solution Approach 1:
The patent segments the image into two distinct parts: the control object and the background. By obtaining location information of the control object and creating a middle image that contains only the control object with a standardized background, the method separates the identification task from the background information, allowing accurate control object identification while preserving background information in the original images for final stitching.
Data Source
AI summary
An image stitching method includes: obtaining a control object of a first image; obtaining a control object of a second image; identifying a same control object of the first image and the second image, and location information of the same control object, where the location information includes first location information and second location information, the first location information indicates a location of the same control object in the first image, and the second location information indicates a location of the same control object in the second image; and stitching the first image and the second image into a third image according to the location information of the same control object.