Panoramic Image Alignment via Vector-Based Moving Object Exclusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional panoramic image capturing techniques often result in unnatural impressions due to the inclusion of moving objects, as they fail to accurately align and combine images captured while the camera is moved, leading to partial object cutting or unnatural object movement in the final panoramic image.
Innovation Solution
An image processing apparatus and method that consecutively captures images, divides them into areas, detects characteristic points, calculates vectors between corresponding points in adjacent images, selects areas based on vector similarity, and adjusts images to avoid moving objects, thereby combining images naturally without cutting or incorrectly depicting moving subjects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If conventional panoramic image capturing techniques are used to capture wide-angle images, then the image capturing range is expanded beyond hardware specifications, but moving objects are incorrectly depicted or cut in the final panoramic image
Solution Approach 1:
The image is divided into multiple areas, and characteristic points are detected separately in each area. Vectors are calculated for each area independently, allowing selective combination of areas with reliable vector information while excluding areas with moving objects that would cause alignment errors.
Solution Approach 2:
Different areas of the image are treated differently based on their vector characteristics. Areas with consistent vectors (indicating no moving objects) are selected for combination, while areas with inconsistent vectors (indicating moving objects) are excluded or handled separately, ensuring local quality control in the panoramic image.
2Productivity
If images are combined using conventional techniques without area selection, then the panoramic image is generated quickly, but unnatural impressions occur due to moving object distortion
Solution Approach 1:
The image is pre-divided into areas and characteristic points are detected in advance. Vectors are calculated for each area before the final combination step, allowing the system to identify and exclude areas with moving objects. This preliminary processing ensures that only reliable areas are combined, maintaining both speed and accuracy.
3Area of stationary object
If all areas are used for image combination, then complete coverage is achieved, but moving objects cause cutting and unnatural depiction
Solution Approach 1:
Areas containing moving objects are identified through vector calculation and extracted from the combination process. By taking out these problematic areas and excluding them from the final panoramic image combination, the harmful effects of moving object distortion are eliminated while preserving the quality of static scene areas.
Data Source
AI summary
An image processing apparatus 1 includes: an adjustment unit 55 that adjusts adjacent captured images in panoramic image generation processing; and a panoramic image generation unit 56 that generates data of a panoramic image by combining the adjacent captured images based on a result of this adjustment. The adjustment unit 55 includes: a characteristic point trace unit 552 that calculates vectors of a plurality of corresponding characteristic points in the adjacent captured images; and a position adjustment unit 554 that adjusts the adjacent captured images based on the vectors of the corresponding characteristic points thus calculated, by avoiding a moving object whose image is currently captured.