Package sc.fiji.snt.util

Class PointDeduplicator

java.lang.Object

sc.fiji.snt.util.PointDeduplicator

public class PointDeduplicator
extends Object

Greedy non-maximum suppression (NMS) for scored 3D points. Given a set of points with associated scores and a merging distance, returns the subset of points that survive suppression: starting from the highest-scoring point, each point is accepted only if no already-accepted point lies within the merging distance.

This is useful for deduplicating detections that appear in overlapping neighborhoods (e.g., the same varicosity detected in adjacent cross-sections).

Author:

Tiago Ferreira

Constructor Summary

Constructors

Constructor	Description
PointDeduplicator(List<? extends SNTPoint> points, double[] scores, double mergingDistance)	Creates a new deduplicator.

Method Summary

Modifier and Type	Method	Description
List<SNTPoint>	run()	Runs the greedy NMS.
List<Integer>	runIndices()	Runs the greedy NMS and returns the original indices of surviving points.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

PointDeduplicator

public PointDeduplicator(List<? extends SNTPoint> points,
 double[] scores,
 double mergingDistance)

Creates a new deduplicator.

Parameters:

points - the points to deduplicate

scores - score for each point (same length as points); higher scores are preferred during suppression

mergingDistance - points closer than this (in the same units as the point coordinates) are considered duplicates

Throws:

IllegalArgumentException - if array lengths differ or distance is negative

Method Details

run

public List<SNTPoint> run()

Runs the greedy NMS.

Returns:

the surviving points, ordered by descending score

runIndices

public List<Integer> runIndices()

Runs the greedy NMS and returns the original indices of surviving points.

Returns:

indices into the original input list, ordered by descending score