Image Fusion Processing Preserving Information Integrity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image fusion methods for medical imaging, such as CT and PET images, result in information degradation due to the synthesis of voxel values before rendering, leading to blurred contours and unnatural representations, especially when images have different resolutions.
Innovation Solution
An image fusion processing method that associates multiple image data sets based on positional relationships, projects virtual rays, determines synthesis ratios for optical parameters, calculates synthesized optical parameters, and accumulates partially reflected light to generate pixel values, thereby preserving the information integrity of the fusion image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If voxel values are synthesized before rendering to generate fusion image, then the fusion image can be generated efficiently, but the information content degenerates from multi-dimensional to one-dimensional causing blurred contours and unnatural representation
Solution Approach 1:
The patent segments the image fusion process into two distinct stages: (1) pre-processing stage where volume data from multiple modalities is prepared and registered, and (2) rendering stage where virtual rays are cast and optical parameters are synthesized along each ray path. This segmentation allows information preservation in the original volume data while enabling efficient processing through staged computation.
Solution Approach 2:
The patent transitions from synthesizing voxel values in 3D space to synthesizing optical parameters along 1D ray paths. By casting virtual rays through the registered volume data and computing optical parameter synthesis along each ray's path, the method preserves multi-dimensional information while generating the 2D fusion image, avoiding the information loss inherent in traditional 3D voxel synthesis approaches.
2Device complexity
If optical parameters are synthesized using a single synthesis ratio, then the processing is simple and fast, but the fusion image becomes unnatural when images have different resolutions
Solution Approach 1:
The patent introduces dynamic, position-dependent synthesis ratios that vary along the path of each virtual ray. Instead of applying a uniform synthesis ratio throughout the entire volume, the method computes local synthesis ratios at different positions along the ray path based on the relative resolutions and information content of the input images at those specific locations. This dynamic approach adapts to local variations in image quality and resolution, producing more natural fusion results.
Solution Approach 2:
The patent applies different synthesis ratios to different regions and depths within the volume data. By evaluating the local characteristics of the input images at each sampling position along the virtual ray path, the method determines appropriate synthesis ratios that optimize the contribution of each modality locally. This ensures that regions with higher resolution or more reliable information contribute more to the final fusion image, while maintaining processing efficiency through localized computations.
Data Source
AI summary
An image fusion processing method, image fusion processing program, and image fusion processing device for fusing a plurality of images of three or more dimensions obtained from a single watched object without degenerating the amount of information. A CPU calculates an optical parameter corresponding to a first voxel value and an optical parameter corresponding to a second voxel value. Then, the CPU calculates synthesized optical parameters based on a synthesis ratio obtained through a synthesis ratio determining process performed at each current position and updates the residual light and reflected light using these optical parameters. When the current position is an end point, a CPU sets the reflected light as a pixel value, and the fusion image data generating process ends for the single pixel configuring a frame.