Tiled Region Adjacency Graphs for Large-Image Watershed Segmentation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Marker-based watershed segmentation algorithms consume excessive computational resources when applied to large images generated by scientific instruments, and existing techniques fail to generate region adjacency graphs (RAGs) in a tiled context, leading to inconsistencies and 'marker-based watershed leaks'.

Innovation Solution

The development of pixel-region adjacency graphs (PRAGs) allows for the construction of RAGs in a tiled fashion by converting image tiles into pixel adjacency graphs (PAGs), generating minimum spanning forests, flagging border and interior trees, merging border regions, and condensing interior regions into nodes, enabling marker-based watershed segmentation on large images without excessive resource consumption.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If marker-based watershed segmentation is applied to large images, then segmentation accuracy is improved, but computational resources consumed increase excessively

Engineering Contradiction:

Improvesegmentation accuracyVSAvoidcomputational resources

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent divides the large image into multiple smaller tiles, processes each tile independently to generate pixel-region adjacency graphs, and then merges the results. This segmentation approach reduces computational complexity while maintaining segmentation accuracy across the entire large image.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces region adjacency graphs (RAGs) as an intermediary data structure between the image pixels and the segmentation results. The RAG condenses pixel-level information into region-level representations, enabling efficient computational processing while preserving essential segmentation information.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Use of energy by moving object

If region adjacency graphs are constructed from image tiles, then computational resources are reduced, but inconsistencies and segmentation leaks occur

Engineering Contradiction:

Improvecomputational resourcesVSAvoidsegmentation consistency

Core Design Contradiction:

Use of energy by moving objectVSReliability

Solution Approach 1:

The patent merges the pixel-region adjacency graphs from multiple tiles into a single consolidated region adjacency graph. This merging process ensures consistency across tile boundaries by integrating information from adjacent tiles, eliminating segmentation leaks while maintaining the computational efficiency of tile-based processing.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent implements feedback mechanisms during the merging process where region adjacency information from adjacent tiles is used to refine and correct segmentation results. This feedback loop ensures consistency and reliability by adjusting segmentations at tile boundaries based on contextual information from neighboring regions.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20250292413A1Tiled region adjacency graph computation via pixel-region adjacency graphs

Publication Date: 2025.09.18 FEI CO

AI summary

Systems or techniques are provided for facilitating tiled region adjacency graph computation via pixel-region adjacency graphs. In various embodiments, a system can access an image generated by a scientific instrument. In various aspects, the system can perform marker-based watershed segmentation on a region adjacency graph of the image, wherein the region adjacency graph can be constructed from a plurality of pixel-region adjacency graphs respectively corresponding to a plurality of tiles of the image.