Mobile Panorama Stitching With ROI-Based Frame Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image stitching processes often select base frames containing low-quality representations of objects of interest, leading to reduced image fidelity and noticeable distortions, and fail to discriminate between objects of interest and background objects during blending, resulting in artifacts and distortions.
Innovation Solution
The process intelligently selects base frames by considering the quality of objects of interest and penalizes seams placed on these regions during blending, ensuring each region of interest is fully contained within a selected base frame and applying a computational bias to minimize seam placement on these areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional stitching algorithms select base frames from overlapping regions, then the stitching process can be completed, but the selected base frames may contain low-quality representations of objects of interest leading to reduced image fidelity
Solution Approach 1:
The patent applies local quality by differentiating between important and unimportant regions within image frames. It identifies regions of interest (such as objects of interest) and assigns them higher importance weights, then selects base frames based on the quality of these specific regions rather than overall frame quality. This ensures that base frames containing high-quality representations of objects of interest are preferentially selected, thereby resolving the contradiction between completing the stitching process and maintaining image fidelity of important regions.
Solution Approach 2:
The patent changes the selection parameter from overall frame quality to region-specific quality metrics. It computes quality measures for specific regions of interest within frames and uses these regional metrics (rather than global frame metrics) to guide base frame selection. This parameter change enables the system to prioritize frames with high-quality objects of interest even if other parts of the frame are of lower quality, thus improving both image fidelity and reliability of important regions.
2Productivity
If stitching algorithms blend all image frames uniformly, then the composite image can be created, but artifacts and distortions occur because the algorithm fails to discriminate between objects of interest and background objects
Solution Approach 1:
The patent applies local quality in the blending process by assigning different blending weights to different regions based on their importance. Regions of interest (objects of interest) are assigned higher weights and protected from seam placement, while background regions receive lower weights and can tolerate seam placement. This differential blending approach creates composite images more efficiently while maintaining high quality in important regions, resolving the contradiction between productivity and image quality.
Solution Approach 2:
The patent introduces an intermediary mechanism (the blending weight map and seam penalty field) that mediates between the need for efficient composite image generation and the requirement for high image quality. This intermediary layer allows the system to prioritize blending operations in regions of interest while allowing faster, less precise operations in background regions, thus achieving both productivity and manufacturing precision.
3Productivity
If seams are placed freely during the blending process, then the stitching can be completed efficiently, but noticeable artifacts and distortions appear on objects of interest
Solution Approach 1:
The patent applies preliminary anti-action by pre-computing penalty values for potential seam locations based on the importance of underlying regions. Before performing the actual blending and seam placement, the system creates a penalty map that identifies regions where seam placement would be harmful (objects of interest) versus regions where it is acceptable (background). This preliminary anti-action guides the blending process to avoid placing seams on important regions, thereby maintaining stitching efficiency while preventing artifacts and distortions on objects of interest.
Data Source
AI summary
Devices and methods for selecting and stitching image frames are provided. A method includes obtaining a plurality of image frames. The method also includes identifying one or more regions of interest within one or more image frames in the plurality of image frames. The method further includes selecting, based on a respective quality measure associated with each image frame of the plurality of image frames, a set of base frames, where each identified region of interest of the one or more identified regions of interest is fully contained within at least one base frame in the selected set of base frames. The method additionally includes stitching together the selected set of base frames to create a composite image.