Multi-Focus Image Fusion for Autofluorescence Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques for z-stack imaging of biological tissues struggle to generate focused images across the entire sample and effectively remove autofluorescence, leading to low contrast images.
Innovation Solution
A method of multi-focus image fusion that involves determining a focus map based on a z-stack of images, applying this map to generate fused images, and subtracting background images to produce high-contrast images, using techniques like SIFT and RANSAC for registration and Tenengrad and Vollath's F4 for focus metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a single most in-focus plane is selected from a z-stack, then the selection process is simple, but the image is not focused across the entire sample
Solution Approach 1:
The image is divided into multiple patches, and a focus map is created that identifies the in-focus plane for each patch. This segmentation approach allows different regions of the sample to have their own optimal focus plane selected, achieving full-sample focus uniformity while maintaining computational efficiency through localized processing.
Solution Approach 2:
The solution transitions from selecting a single 2D plane to creating a 3D focus map that specifies the optimal z-plane for each (x, y) position in the image. This dimensional expansion from 2D to 3D space enables precise control over focus across the entire sample volume.
2Measurement precision
If stained images are captured to improve contrast, then gene activity detection is enhanced, but autofluorescence reduces image contrast
Solution Approach 1:
The method separates the stained sample image into two components: the autofluorescence background and the specific stain signal. By extracting and removing the autofluorescence component through subtraction, the specific gene activity signals are enhanced while eliminating the harmful background interference.
Solution Approach 2:
The autofluorescence, which normally degrades image quality, is converted into a useful component by capturing it in an unstained reference image. This reference is then subtracted from stained images to remove the harmful background, effectively transforming the harmful autofluorescence into a benefit for background normalization and contrast enhancement.
3Manufacturing precision
If complex image processing techniques are used to remove autofluorescence and achieve focus, then image quality is improved, but computational energy consumption increases
Solution Approach 1:
The unstained background image is captured and processed in advance to create a reference autofluorescence map. This preliminary action allows the background to be removed from multiple stained images without repeating the complex processing, significantly reducing computational energy consumption while maintaining high image quality.
Solution Approach 2:
The autofluorescence background is copied from the unstained reference image and applied to subtract from stained images. This copying approach avoids re-processing the background for each stained image, reducing computational complexity and energy usage while achieving consistent background removal across all images.
Data Source
AI summary
Provided herein are methods, systems, and computer program products for image fusion. A first z-stack of images of a biological sample may be received. The first z-stack of images may correspond to a first field of view, which comprises a plurality of patches. A second z-stack of images of the biological sample may be received. The second z-stack may correspond to the first field of view. A focus map may be determined based on the second z-stack, the focus map indicating, for each of a plurality of patches of the first field of view, one of the images of the second z-stack bringing into focus that patch of the first field of view. The focus map may be applied to the first and second z-stacks to generate respective first and second fused images. The first fused image may be subtracted from the second fused image to produce a subtracted image.