Texture-Based Fusion Renderer Using Slice Sorting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fusion renderers require resampling and alignment of volumetric datasets, leading to increased processing time and memory usage, especially when volumes have different orientations and resolutions, making real-time registration and rendering inefficient.
Innovation Solution
A novel fusion renderer treats each volume as a stack of parallel 2D slices, sorts and renders them in visibility order, and applies slice-by-slice invisible space skipping to avoid resampling, allowing independent rendering parameters and orientations, and utilizes a binary space partitioning tree for efficient intersection computation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If resampling is performed to align and standardize volumetric datasets, then image quality and consistency are improved, but processing time and memory requirements increase significantly
Solution Approach 1:
The volumetric dataset is divided into multiple 2D slices that can be processed and rendered independently. Each slice is treated as a separate rendering target, allowing parallel processing and avoiding the need to resample the entire volume. This segmentation enables efficient rendering by processing only visible slices in visibility order.
Solution Approach 2:
The patent transforms the 3D volume rendering problem into a series of 2D slice rendering problems. By slicing the volume along the viewing direction and rendering each 2D slice independently, the system avoids the computational complexity of 3D resampling while maintaining rendering quality. The volume is represented as a stack of 2D slices rather than a continuous 3D grid.
2Stability of the object's composition
If all volumes are re-sampled to match the reference volume resolution, then data consistency is improved, but memory requirements increase significantly
Solution Approach 1:
Different volumes can have different resolutions and are only resampled locally where necessary for the current rendering operation. The system processes each volume at its native resolution and performs selective resampling only for visible portions, rather than forcing all volumes to a uniform high resolution. This local approach maintains data consistency where needed while conserving memory.
3Measurement precision
If registration with visual feedback is performed to refine alignment, then registration accuracy is improved, but response time decreases making real-time interaction impossible
Solution Approach 1:
The system performs preliminary alignment using transformation parameters (rotation, translation, scaling) before rendering. The volumes are pre-aligned using these parameters, and the rendering system then efficiently processes the aligned volumes without requiring iterative visual feedback during the rendering operation itself. This preliminary alignment enables fast rendering while maintaining registration accuracy.
Data Source
AI summary
An exemplary method of combining two images to form a fused image is provided. A hierarchical structure for each of a plurality of volumes is built. All blocks in each of the hierarchical structures that intersect a slicing plane are found. Each of the plurality of volumes is divided into stacks of parallel slices, and the parallel slices are sorted by visibility order. A next slice in the sorted parallel slices is chosen. The next slice belongs to a current volume. Rendering parameters are changed if the current volume is different from a previous volume associated with the chosen slice in a previous iteration of the step of choosing the slice. Based on the rendering parameters, the next slice is rendered by intersecting the slicing plane with the blocks corresponding to the current volume. The steps of choosing the slice, changing the rendering parameters, and rendering the slice are repeated until all of the sorted parallel slices are rendered.