Overlapping Image Stitching with Weighted Similarity Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image acquisition methods struggle to combine multiple picked-up images with high accuracy due to abnormal positional deviations caused by container edges with high contrast, leading to unreliable image composition.
Innovation Solution
An image acquisition apparatus and method that utilizes template matching to generate similarity maps, identifies specific image pairs with directivity, and adjusts weights based on the reliability of positional deviations to combine images accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all image pairs including those with container edges are used for combination, then more images are combined, but the combination accuracy deteriorates due to abnormal positional deviations
Solution Approach 1:
The patent applies local quality by differentiating between reliable and unreliable image pairs based on their local characteristics. Specifically, image pairs containing container edges are identified as having abnormal positional deviations and are excluded from the combination process. This selective approach ensures that only high-quality image pairs with normal positional relationships are used, thereby maintaining high combination accuracy without unnecessarily complicating the overall system.
Solution Approach 2:
The patent segments the set of all image pairs into two distinct groups: those containing container edges (abnormal image pairs) and those without container edges (normal image pairs). This segmentation is achieved by detecting container edges in the overlapping regions and identifying image pairs that contain these edges. By separating the image pairs in this manner, the system can selectively process only the normal image pairs for combination, avoiding the negative impact of abnormal positional deviations.
2Productivity
If template matching is performed on all overlapping regions, then positional deviation is calculated for all image pairs, but processing time increases due to unnecessary calculations on abnormal regions
Solution Approach 1:
The patent applies preliminary action by performing container edge detection in the overlapping regions before conducting template matching. This preliminary step allows the system to identify which image pairs contain container edges and should be excluded from further processing. By performing this preliminary classification, the system avoids wasting computational resources on template matching for abnormal image pairs, thereby improving processing speed while maintaining the accuracy of positional deviation calculations for the remaining normal image pairs.
Data Source
AI summary
In an image acquisition apparatus, acquired are picked-up images 71 representing divided regions, respectively, which are obtained by dividing a predetermined region on a target object, and in the picked-up images 71, each image pair 72 representing adjacent divided regions has an overlapping region 73. The amount of relative positional deviation in each image pair 72 is specified by generating a similarity map 74 indicating a distribution of the degree of similarity. Combination positions of the picked-up images 71 are determined on the basis of the amounts of positional deviation in image pairs 72 included in the picked-up images 71 while it is assumed that an image pair 72 whose similarity map 74 has a directivity is a specific image pair 72 and a weight of the specific image pair 72 is set to be lower than those of other image pairs 72.