Ophthalmologic Image Alignment via Directional Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques face challenges in precisely aligning ophthalmologic images that have undergone position shifts in both the first and second directions, leading to distortion and reduced alignment precision, especially when the same part of a subject eye is photographed multiple times.
Innovation Solution
An ophthalmologic image processing device and medium that acquire intermediate information excluding position shift influences in one direction, allowing for precise alignment in both directions by performing first alignment in the second direction and then in the first direction between pixels at the same position, using methods like phase-only correlation and dynamic programming.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional alignment techniques are used to align ophthalmologic images with position shifts in both first and second directions, then alignment can be performed, but alignment precision deteriorates due to distortion caused by simultaneous position shifts in both directions
Solution Approach 1:
The alignment process is segmented into two independent sequential steps: first performing alignment in the second direction (B-scanning direction) to correct position shifts in that direction, then performing alignment in the first direction (A-scanning direction) to correct position shifts in the depth direction. This segmentation allows each alignment operation to focus on one direction at a time, avoiding the distortion problems that arise when attempting to correct both directions simultaneously.
Solution Approach 2:
The alignment in the second direction is performed as a preliminary action before alignment in the first direction. By first correcting the position shifts in the B-scanning direction, the images are pre-aligned in that dimension, which then enables more accurate subsequent alignment in the A-scanning direction without the compounding effects of uncorrected multi-directional shifts.
2Measurement precision
If both alignment with respect to the first direction and alignment with respect to the second direction are performed simultaneously, then comprehensive alignment is achieved, but it becomes difficult to precisely perform alignment in one direction while maintaining position shift correction in the other direction
Solution Approach 1:
The complex two-dimensional alignment problem is segmented into two simpler one-dimensional alignment tasks performed sequentially. First, alignment in the second direction is performed independently, then alignment in the first direction is performed on the already-aligned images. This segmentation reduces the mathematical complexity and computational difficulty compared to attempting simultaneous two-directional alignment.
Solution Approach 2:
Alignment in the second direction is performed as a preliminary step before alignment in the first direction. This preliminary alignment simplifies the subsequent alignment operation by reducing the degrees of freedom that need to be corrected in the second step, making the overall process more manageable and precise.
Data Source
AI summary
A processor of an ophthalmologic image processing device acquires intermediate information from which an influence of a position shift with respect to a first direction at each position in a second direction is excluded, for each of a first ophthalmologic image and a second ophthalmologic image (S4). The processor performs alignment with respect to the second direction between the first ophthalmologic image and the second ophthalmologic image, based on the intermediate information (S5, S6). The processor performs alignment with respect to the first direction between pixels at the same position with respect to the second direction in the first ophthalmologic image and the second ophthalmologic image for which the alignment with respect to the second direction has been performed (S7).