Layered Lightfields for Occlusion Handling in Medical Imaging

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional lightfield rendering techniques struggle with occlusion, preventing cameras from moving inside or through objects, such as in medical imaging applications like colonoscopy, due to the convex hull of the object occluding interior views.

Innovation Solution

The method employs layered lightfields, where multiple lightfields are generated for different depths and surfaces, allowing camera positions within the convex hull by selecting appropriate lightfields based on depth information, enabling rendering from within concave regions and through objects.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If standard lightfield rendering is used, then interactivity is achieved, but camera positions are restricted to outside the convex hull of the object due to occlusion

Engineering Contradiction:
Improvecamera position freedomVSAvoidocclusion by convex hull
Core Design Contradiction:
Ease of operationVSObject-generated harmful factors

Solution Approach 1:

The lightfield data is segmented into multiple layers corresponding to different depth ranges or isosurfaces. Each layer contains lightfield information for a specific depth interval, allowing the system to selectively composite layers based on camera position and desired visibility, thereby enabling views from within the convex hull that would otherwise be occluded.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a depth layer dimension to the traditional lightfield representation. By organizing lightfield data into multiple depth layers and using per-pixel depth values, the system adds a dimensional organization that enables selective visibility control, allowing cameras to virtually move into concave regions and through objects without being constrained by the convex hull occlusion.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Manufacturing precision

If volume rendering with Monte-Carlo sampling is used for high quality, then image quality is improved, but rendering speed decreases making interactivity difficult

Engineering Contradiction:
Improveimage qualityVSAvoidrendering speed
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent performs preliminary rendering to pre-compute lightfield data for multiple camera positions and depth layers before runtime. This pre-computation stores radiance and orientation information in an organized layered structure, enabling fast interactive rendering by simply compositing pre-rendered layers rather than performing expensive Monte-Carlo sampling during interaction.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system dynamically composites static pre-rendered lightfield layers based on runtime camera position and viewing parameters. This dynamic composition of pre-computed layers provides both high image quality (inherited from careful pre-rendering) and fast interactivity (from avoiding repeated expensive sampling), resolving the contradiction between quality and speed.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3246881B1Layered lightfields for occlusion handling
Publication Date: 2022.08.03 SIEMENS HEALTHCARE GMBH
  • EP3246881B1 patent drawingFigure 1
  • EP3246881B1 patent drawingFigure 2~3
  • EP3246881B1 patent drawingFigure 4~5

AI summary

For occlusion handing in lightfield rendering, layered lightfields are created. Rather than use one lightfield for one camera position and orientation, multiple lightfields representing different depths or surfaces at different depths relative to that camera position and orientation are created. By using layered lightfields for the various camera positions and orientations, the camera may be located within the convex hull or scanned object. The depths of the layers are used to select the lightfields for a given camera position and location.