Multi-Resolution Image Stitching for Medical Radiography
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image stitching methods are inefficient and inaccurate in combining multiple sequences of images to create a single image of interest, particularly in medical radiography applications where large objects or high resolution requirements necessitate multiple image acquisition.
Innovation Solution
A method involving the steps of determining Regions of Interest, extracting and matching features at multiple resolutions, and blending estimates to stitch images accurately, using techniques such as Gaussian smoothing, Radon transforms, and template matching algorithms to create a seamless stitched image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If multiple images are acquired to capture large objects or high resolution regions, then the completeness and resolution of the captured region of interest is improved, but the complexity of combining and stitching multiple images increases
Solution Approach 1:
The patent segments the stitching process into distinct phases: feature extraction at multiple resolutions, template matching at coarse resolution, and refinement at fine resolution. This segmentation allows each phase to focus on specific tasks, improving overall stitching accuracy while making the complex process more manageable and systematic
Solution Approach 2:
The patent introduces multi-resolution processing as an additional dimension to the stitching process. By extracting features at multiple resolutions and performing template matching at different scales, the system achieves more robust alignment without being constrained by single-resolution limitations, thereby improving accuracy without proportionally increasing complexity
2Manufacturing precision
If traditional stitching algorithms are used to combine multiple images, then the process is simple, but the stitching accuracy and quality deteriorates
Solution Approach 1:
The patent performs preliminary feature extraction at multiple resolutions before the actual template matching process. By pre-processing and organizing features at different scales, the system reduces the computational burden during the matching phase, achieving high accuracy without excessive processing time
Solution Approach 2:
The patent implements a dynamic, multi-resolution approach where the processing detail adapts to the scale being examined. Coarse-resolution features guide the initial alignment, while fine-resolution features refine the positioning. This dynamic strategy optimizes processing time by avoiding exhaustive fine-detail analysis at every stage
3Manufacturing precision
If features are extracted and matched at high resolution only, then the stitching accuracy is improved, but the computational complexity and processing time increases significantly
Solution Approach 1:
The patent segments the feature matching process into multiple resolution levels. Coarse-resolution feature extraction and matching are performed first to establish initial alignment, followed by fine-resolution processing only in the aligned regions. This segmentation dramatically reduces the total computational load compared to processing all images at full resolution
Solution Approach 2:
The patent applies partial processing by performing high-resolution feature extraction and matching only in regions of interest after coarse alignment, rather than processing entire images at full resolution. This partial action maintains high stitching accuracy while significantly improving processing efficiency
Data Source
AI summary
A method to stitch images includes the steps of: determining a template window in the first digital image and a target window in the second digital image and extracting selected features from the selected windows and a template in the template window; extracting selected features from within the template at a first resolution; matching the selected features in the target window; extracting selected features from within the template and target window at a second resolution higher than the first resolution; matching the selected features in the target window; performing a first evaluation of the second estimate of the stitching location; blending the second estimate of the stitching location; performing an evaluation of the stitching location; stitching the first digital image and a second digital image using the stitching location to create a stitched image; and saving the stitched image to memory. A system to perform the method is also described.