Examples with UnaryFunctionOp net.imagej.ops.special.function.UnaryFunctionOp used on opensource projects

Example 1 with UnaryFunctionOp

use of net.imagej.ops.special.function.UnaryFunctionOp in project imagej-ops by imagej.

the class CachedOpEnvironment method op.

@Override
public Op op(final OpRef ref) {
    final Op op = super.op(ref);
    for (final Class<?> ignored : ignoredOps) {
        for (final Type t : ref.getTypes()) {
            // FIXME: Use generic assignability test, once it exists.
            final Class<?> raw = GenericUtils.getClass(t);
            if (ignored.isAssignableFrom(raw)) {
                return op;
            }
        }
    }
    final Op cachedOp;
    if (op instanceof UnaryHybridCF) {
        cachedOp = wrapUnaryHybrid((UnaryHybridCF<?, ?>) op);
    } else if (op instanceof UnaryFunctionOp) {
        cachedOp = wrapUnaryFunction((UnaryFunctionOp<?, ?>) op);
    } else
        return op;
    getContext().inject(cachedOp);
    return cachedOp;
}

Example 2 with UnaryFunctionOp

use of net.imagej.ops.special.function.UnaryFunctionOp in project imagej-ops by imagej.

the class ConvolveTest method testConvolve.

/**
 * tests fft based convolve
 */
@Test
public void testConvolve() {
    float delta = 0.0001f;
    int[] size = new int[] { 225, 167 };
    int[] kernelSize = new int[] { 27, 39 };
    long[] borderSize = new long[] { 10, 10 };
    // create an input with a small sphere at the center
    Img<FloatType> in = new ArrayImgFactory<FloatType>().create(size, new FloatType());
    placeSphereInCenter(in);
    // create a kernel with a small sphere in the center
    Img<FloatType> kernel = new ArrayImgFactory<FloatType>().create(kernelSize, new FloatType());
    placeSphereInCenter(kernel);
    // create variables to hold the image sums
    FloatType inSum = new FloatType();
    FloatType kernelSum = new FloatType();
    FloatType outSum = new FloatType();
    FloatType outSum2 = new FloatType();
    FloatType outSum3 = new FloatType();
    // calculate sum of input and kernel
    ops.stats().sum(inSum, in);
    ops.stats().sum(kernelSum, kernel);
    // convolve and calculate the sum of output
    @SuppressWarnings("unchecked") final Img<FloatType> out = (Img<FloatType>) ops.run(ConvolveFFTF.class, in, kernel, borderSize);
    // create an output for the next test
    Img<FloatType> out2 = new ArrayImgFactory<FloatType>().create(size, new FloatType());
    // create an output for the next test
    Img<FloatType> out3 = new ArrayImgFactory<FloatType>().create(size, new FloatType());
    // Op used to pad the input
    final BinaryFunctionOp<RandomAccessibleInterval<FloatType>, Dimensions, RandomAccessibleInterval<FloatType>> padOp = (BinaryFunctionOp) Functions.binary(ops, PadInputFFTMethods.class, RandomAccessibleInterval.class, RandomAccessibleInterval.class, Dimensions.class, true);
    // Op used to pad the kernel
    final BinaryFunctionOp<RandomAccessibleInterval<FloatType>, Dimensions, RandomAccessibleInterval<FloatType>> padKernelOp = (BinaryFunctionOp) Functions.binary(ops, PadShiftKernelFFTMethods.class, RandomAccessibleInterval.class, RandomAccessibleInterval.class, Dimensions.class, true);
    // Op used to create the complex FFTs
    UnaryFunctionOp<Dimensions, RandomAccessibleInterval<ComplexFloatType>> createOp = (UnaryFunctionOp) Functions.unary(ops, CreateOutputFFTMethods.class, RandomAccessibleInterval.class, Dimensions.class, new ComplexFloatType(), true);
    final int numDimensions = in.numDimensions();
    // 1. Calculate desired extended size of the image
    final long[] paddedSize = new long[numDimensions];
    // if no getBorderSize() was passed in, then extend based on kernel size
    for (int d = 0; d < numDimensions; ++d) {
        paddedSize[d] = (int) in.dimension(d) + (int) kernel.dimension(d) - 1;
    }
    RandomAccessibleInterval<FloatType> paddedInput = padOp.calculate(in, new FinalDimensions(paddedSize));
    RandomAccessibleInterval<FloatType> paddedKernel = padKernelOp.calculate(kernel, new FinalDimensions(paddedSize));
    RandomAccessibleInterval<ComplexFloatType> fftImage = createOp.calculate(new FinalDimensions(paddedSize));
    RandomAccessibleInterval<ComplexFloatType> fftKernel = createOp.calculate(new FinalDimensions(paddedSize));
    // run convolve using the rai version with the memory created above
    ops.run(ConvolveFFTC.class, out2, paddedInput, paddedKernel, fftImage, fftKernel);
    ops.run(ConvolveFFTC.class, out3, paddedInput, paddedKernel, fftImage, fftKernel, true, false);
    ops.stats().sum(outSum, Views.iterable(out));
    ops.stats().sum(outSum2, out2);
    ops.stats().sum(outSum3, out3);
    // multiply input sum by kernelSum and assert it is the same as outSum
    inSum.mul(kernelSum);
    assertEquals(inSum.get(), outSum.get(), delta);
    assertEquals(inSum.get(), outSum2.get(), delta);
    assertEquals(inSum.get(), outSum3.get(), delta);
    assertEquals(size[0], out.dimension(0));
    assertEquals(size[0], out2.dimension(0));
}