Add (P)SNR, RMSE, and MAE implementations

Adds various metrics for comparing reference images with (noisy) test
images.

Author: stelfrich

src/main/java/net/imagej/ops/image/ImageNamespace.java

@@ -42,6 +42,7 @@
 import net.imglib2.type.Type;
 import net.imglib2.type.numeric.RealType;
 import net.imglib2.type.numeric.integer.IntType;
+import net.imglib2.type.numeric.real.DoubleType;
 
 import org.scijava.plugin.Plugin;
 
@@ -456,6 +457,80 @@ <T extends RealType<T>> IterableInterval<T> normalize(
 		return result;
 	}
 
+	// -- quality --
+
+	@OpMethod(op = net.imagej.ops.image.quality.DefaultMAE.class)
+	public <T extends RealType<T>> DoubleType mae(final DoubleType out,
+		final IterableInterval<T> reference, final IterableInterval<T> test)
+	{
+		final DoubleType result = (DoubleType) ops().run(
+			net.imagej.ops.image.quality.DefaultMAE.class, out, reference, test);
+		return result;
+	}
+
+	@OpMethod(op = net.imagej.ops.image.quality.DefaultMAE.class)
+	public <T extends RealType<T>> DoubleType mae(
+		final IterableInterval<T> reference, final IterableInterval<T> test)
+	{
+		final DoubleType result = (DoubleType) ops().run(
+			net.imagej.ops.image.quality.DefaultMAE.class, reference, test);
+		return result;
+	}
+
+	@OpMethod(op = net.imagej.ops.image.quality.DefaultPSNR.class)
+	public <T extends RealType<T>> DoubleType psnr(final DoubleType out,
+		final IterableInterval<T> reference, final IterableInterval<T> test)
+	{
+		final DoubleType result = (DoubleType) ops().run(
+			net.imagej.ops.image.quality.DefaultPSNR.class, out, reference, test);
+		return result;
+	}
+
+	@OpMethod(op = net.imagej.ops.image.quality.DefaultPSNR.class)
+	public <T extends RealType<T>> DoubleType psnr(
+		final IterableInterval<T> reference, final IterableInterval<T> test)
+	{
+		final DoubleType result = (DoubleType) ops().run(
+			net.imagej.ops.image.quality.DefaultPSNR.class, reference, test);
+		return result;
+	}
+
+	@OpMethod(op = net.imagej.ops.image.quality.DefaultRMSE.class)
+	public <T extends RealType<T>> DoubleType rmse(final DoubleType out,
+		final IterableInterval<T> reference, final IterableInterval<T> test)
+	{
+		final DoubleType result = (DoubleType) ops().run(
+			net.imagej.ops.image.quality.DefaultRMSE.class, out, reference, test);
+		return result;
+	}
+
+	@OpMethod(op = net.imagej.ops.image.quality.DefaultRMSE.class)
+	public <T extends RealType<T>> DoubleType rmse(
+		final IterableInterval<T> reference, final IterableInterval<T> test)
+	{
+		final DoubleType result = (DoubleType) ops().run(
+			net.imagej.ops.image.quality.DefaultRMSE.class, reference, test);
+		return result;
+	}
+
+	@OpMethod(op = net.imagej.ops.image.quality.DefaultSNR.class)
+	public <T extends RealType<T>> DoubleType snr(final DoubleType out,
+		final IterableInterval<T> reference, final IterableInterval<T> test)
+	{
+		final DoubleType result = (DoubleType) ops().run(
+			net.imagej.ops.image.quality.DefaultSNR.class, out, reference, test);
+		return result;
+	}
+
+	@OpMethod(op = net.imagej.ops.image.quality.DefaultSNR.class)
+	public <T extends RealType<T>> DoubleType snr(
+		final IterableInterval<T> reference, final IterableInterval<T> test)
+	{
+		final DoubleType result = (DoubleType) ops().run(
+			net.imagej.ops.image.quality.DefaultSNR.class, reference, test);
+		return result;
+	}
+
 	// -- watershed --
 
 	/** Executes the "watershed" operation on the given arguments. */

src/main/java/net/imagej/ops/image/quality/DefaultMAE.java

@@ -0,0 +1,93 @@
+/*
+ * #%L
+ * ImageJ software for multidimensional image processing and analysis.
+ * %%
+ * Copyright (C) 2014 - 2018 ImageJ developers.
+ * %%
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * 
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ * 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ * #L%
+ */
+
+package net.imagej.ops.image.quality;
+
+import net.imagej.ops.Contingent;
+import net.imagej.ops.Ops;
+import net.imagej.ops.map.Maps;
+import net.imagej.ops.special.hybrid.AbstractBinaryHybridCF;
+import net.imglib2.Cursor;
+import net.imglib2.IterableInterval;
+import net.imglib2.type.numeric.RealType;
+import net.imglib2.type.numeric.real.DoubleType;
+import net.imglib2.util.Intervals;
+
+import org.scijava.plugin.Plugin;
+
+/**
+ * Computes the mean absolute error (MAE) between a reference image and a
+ * (noisy) test image.
+ * <p>
+ * Computations are based on the definitions of Gonzalez (R.C. Gonzalez and R.E.
+ * Woods, "Digital Image Processing," Prentice Hall 2008).
+ * </p>
+ *
+ * @author Stefan Helfrich (University of Konstanz)
+ * @param <I> type of input elements
+ */
+@Plugin(type = Ops.Image.MAE.class)
+public class DefaultMAE<I extends RealType<I>> extends
+	AbstractBinaryHybridCF<IterableInterval<I>, IterableInterval<I>, DoubleType>
+	implements Ops.Image.MAE, Contingent
+{
+
+	@Override
+	public void compute(final IterableInterval<I> input1,
+		final IterableInterval<I> input2, final DoubleType output)
+	{
+		final Cursor<I> cursor = input1.cursor();
+		final Cursor<I> cursor2 = input2.cursor();
+
+		double denominatorSum = 0d;
+		while (cursor.hasNext()) {
+			final double r = cursor.next().getRealDouble();
+			final double t = cursor2.next().getRealDouble();
+			final double abs = Math.abs(r - t);
+			denominatorSum += abs;
+		}
+
+		denominatorSum *= 1d / Intervals.numElements(input1);
+		output.setReal(denominatorSum);
+	}
+
+	@Override
+	public boolean conforms() {
+		return Intervals.equalDimensions(in1(), in2()) && //
+			Maps.compatible(in1(), in2());
+	}
+
+	@Override
+	public DoubleType createOutput(final IterableInterval<I> input1,
+		final IterableInterval<I> input2)
+	{
+		return new DoubleType();
+	}
+
+}

src/main/java/net/imagej/ops/image/quality/DefaultPSNR.java

@@ -0,0 +1,101 @@
+/*
+ * #%L
+ * ImageJ software for multidimensional image processing and analysis.
+ * %%
+ * Copyright (C) 2014 - 2018 ImageJ developers.
+ * %%
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * 
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ * 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ * #L%
+ */
+
+package net.imagej.ops.image.quality;
+
+import net.imagej.ops.Contingent;
+import net.imagej.ops.OpService;
+import net.imagej.ops.Ops;
+import net.imagej.ops.map.Maps;
+import net.imagej.ops.special.hybrid.AbstractBinaryHybridCF;
+import net.imglib2.Cursor;
+import net.imglib2.IterableInterval;
+import net.imglib2.type.numeric.RealType;
+import net.imglib2.type.numeric.real.DoubleType;
+import net.imglib2.util.Intervals;
+
+import org.scijava.plugin.Parameter;
+import org.scijava.plugin.Plugin;
+
+/**
+ * Computes peak signal-to-noise ratio (PSNR) between a reference image and a
+ * (noisy) test image. The resulting PSNR is expressed in decibel.
+ * <p>
+ * Computations are based on the definitions of Gonzalez (R.C. Gonzalez and R.E.
+ * Woods, "Digital Image Processing," Prentice Hall 2008).
+ * </p>
+ *
+ * @author Stefan Helfrich (University of Konstanz)
+ * @param <I> type of input elements
+ */
+@Plugin(type = Ops.Image.PSNR.class)
+public class DefaultPSNR<I extends RealType<I>> extends
+	AbstractBinaryHybridCF<IterableInterval<I>, IterableInterval<I>, DoubleType>
+	implements Ops.Image.PSNR, Contingent
+{
+
+	@Parameter
+	private OpService opService;
+
+	@Override
+	public void compute(final IterableInterval<I> input1,
+		final IterableInterval<I> input2, final DoubleType output)
+	{
+		final Cursor<I> cursor = input1.cursor();
+		final Cursor<I> cursor2 = input2.cursor();
+		double max = opService.stats().max(input1).getRealDouble();
+		max *= max;
+
+		double denominatorSum = 0d;
+		while (cursor.hasNext()) {
+			final double r = cursor.next().getRealDouble();
+			final double t = cursor2.next().getRealDouble();
+			denominatorSum += Math.pow(r - t, 2);
+		}
+
+		denominatorSum *= 1d / Intervals.numElements(input1);
+		final double psnr = 10 * Math.log10(max / denominatorSum);
+
+		output.setReal(psnr);
+	}
+
+	@Override
+	public boolean conforms() {
+		return Intervals.equalDimensions(in1(), in2()) && //
+			Maps.compatible(in1(), in2());
+	}
+
+	@Override
+	public DoubleType createOutput(final IterableInterval<I> input1,
+		final IterableInterval<I> input2)
+	{
+		return new DoubleType();
+	}
+
+}

src/main/java/net/imagej/ops/image/quality/DefaultRMSE.java

@@ -0,0 +1,95 @@
+/*
+ * #%L
+ * ImageJ software for multidimensional image processing and analysis.
+ * %%
+ * Copyright (C) 2014 - 2018 ImageJ developers.
+ * %%
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * 
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ * 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ * #L%
+ */
+
+package net.imagej.ops.image.quality;
+
+import net.imagej.ops.Contingent;
+import net.imagej.ops.Ops;
+import net.imagej.ops.map.Maps;
+import net.imagej.ops.special.hybrid.AbstractBinaryHybridCF;
+import net.imglib2.Cursor;
+import net.imglib2.IterableInterval;
+import net.imglib2.type.numeric.RealType;
+import net.imglib2.type.numeric.real.DoubleType;
+import net.imglib2.util.Intervals;
+
+import org.scijava.plugin.Plugin;
+
+/**
+ * Computes the root mean square error (RMSE) between a reference image and a
+ * (noisy) test image.
+ * <p>
+ * Computations are based on the definitions of Gonzalez (R.C. Gonzalez and R.E.
+ * Woods, "Digital Image Processing," Prentice Hall 2008).
+ * </p>
+ *
+ * @author Stefan Helfrich (University of Konstanz)
+ * @param <I> type of input elements
+ */
+@Plugin(type = Ops.Image.RMSE.class)
+public class DefaultRMSE<I extends RealType<I>> extends
+	AbstractBinaryHybridCF<IterableInterval<I>, IterableInterval<I>, DoubleType>
+	implements Ops.Image.RMSE, Contingent
+{
+
+	@Override
+	public void compute(final IterableInterval<I> input1,
+		final IterableInterval<I> input2, final DoubleType output)
+	{
+		final Cursor<I> cursor = input1.cursor();
+		final Cursor<I> cursor2 = input2.cursor();
+
+		double denominatorSum = 0d;
+		while (cursor.hasNext()) {
+			final double r = cursor.next().getRealDouble();
+			final double t = cursor2.next().getRealDouble();
+			denominatorSum += Math.pow(r - t, 2);
+		}
+
+		denominatorSum *= 1d / Intervals.numElements(input1);
+
+		final double rmse = Math.sqrt(denominatorSum);
+
+		output.setReal(rmse);
+	}
+
+	@Override
+	public boolean conforms() {
+		return Intervals.equalDimensions(in1(), in2()) && //
+			Maps.compatible(in1(), in2());
+	}
+
+	@Override
+	public DoubleType createOutput(final IterableInterval<I> input1,
+		final IterableInterval<I> input2)
+	{
+		return new DoubleType();
+	}
+
+}