net.imglib2.algorithm.metrics.imagequality

Class NRMSE

java.lang.Object

net.imglib2.algorithm.metrics.imagequality.NRMSE

public class NRMSE
extends Object

Compute the normalised root mean squared error (NRMSE) between a reference and a processed image. The metrics runs on the whole image, whether 2D or 3D. In order to get individual slice NRMSE, run the metrics on each slice independently.

Three normalization methods are available:

• euclidean: NRMSE = sqrt(MSE(reference, processed)) * sqrt(N) / || reference ||
• min-max: NRMSE = sqrt(MSE(reference, processed)) / |max(reference)-min(reference)|
• mean: NRMSE = sqrt(MSE(reference, processed)) / mean(reference)

Author:

Joran Deschamps

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	NRMSE.Normalization Normalization methods.

Constructor Summary

Constructors

Constructor and Description
NRMSE()

Method Summary

Modifier and Type	Method and Description
static <T extends RealType<T>> double	computeMetrics(RandomAccessibleInterval<T> reference, RandomAccessibleInterval<T> processed, NRMSE.Normalization norm) Compute the normalised root mean squared error (NRMSE) score between reference and processed images.
protected static <T extends RealType<T>> double	getEuclideanNorm(RandomAccessibleInterval<T> reference)
protected static <T extends RealType<T>> double	getMean(RandomAccessibleInterval<T> reference)
protected static <T extends RealType<T>> double	getMinMaxNorm(RandomAccessibleInterval<T> reference)

Methods inherited from class java.lang.Object

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

Constructor Detail

NRMSE

public NRMSE()

Method Detail

computeMetrics

public static <T extends RealType<T>> double computeMetrics(RandomAccessibleInterval<T> reference,
                                                            RandomAccessibleInterval<T> processed,
                                                            NRMSE.Normalization norm)

Compute the normalised root mean squared error (NRMSE) score between reference and processed images. The metrics run on the whole image (regardless of the dimensions).

Three normalization methods are available:

• euclidean: NRMSE = sqrt(MSE(reference, processed)) * sqrt(N) / || reference ||
• min-max: NRMSE = sqrt(MSE(reference, processed)) / |max(reference)-min(reference)|
• mean: NRMSE = sqrt(MSE(reference, processed)) / mean(reference)

Type Parameters:

T - Type of the image pixels

Parameters:

reference - Reference image

processed - Processed image

Returns:

Metrics score

getEuclideanNorm

protected static <T extends RealType<T>> double getEuclideanNorm(RandomAccessibleInterval<T> reference)

getMinMaxNorm

protected static <T extends RealType<T>> double getMinMaxNorm(RandomAccessibleInterval<T> reference)

getMean

protected static <T extends RealType<T>> double getMean(RandomAccessibleInterval<T> reference)