Package net.preibisch.mvrecon.process.fusion.blk

Class BlkThinPlateSplineFusion

java.lang.Object

net.preibisch.mvrecon.process.fusion.blk.BlkThinPlateSplineFusion

public class BlkThinPlateSplineFusion
extends Object

Field Summary

Fields

Modifier and Type	Field	Description
static final double[]	defaultDisplacementFieldSpacing
static int	defaultIntervalExpansion

Constructor Summary

Constructors

Constructor	Description
BlkThinPlateSplineFusion()

Method Summary

Modifier and Type	Method	Description
static <D extends net.imglib2.type.NativeType<D> & net.imglib2.type.numeric.RealType<D>> net.imglib2.algorithm.blocks.dfield.DisplacementFields.TransformedDisplacementField<D>	concatenateBoundingBoxOffset(net.imglib2.algorithm.blocks.dfield.DisplacementFields.TransformedDisplacementField<D> dfield, net.imglib2.Interval boundingBoxInTarget)
static net.imglib2.realtransform.AffineTransform3D	fitAffineTransform(double[][] source, double[][] target)	Fit an affine that approximates the TPS mapping from source to target landmarks.
static <T extends net.imglib2.type.numeric.RealType<T> & net.imglib2.type.NativeType<T>> net.imglib2.algorithm.blocks.BlockSupplier<T>	init(net.imglib2.converter.Converter<net.imglib2.type.numeric.real.FloatType,T> converter, mpicbg.spim.data.generic.sequence.BasicImgLoader imgLoader, Collection<? extends mpicbg.spim.data.sequence.ViewId> viewIds, Map<mpicbg.spim.data.sequence.ViewId,? extends mpicbg.spim.data.generic.sequence.BasicViewDescription<?>> viewDescriptions, Map<mpicbg.spim.data.sequence.ViewId,Landmarks> viewLandmarks, FusionGUI.FusionType fusionType, double anisotropyFactor, int interpolationMethod, Comparator<mpicbg.spim.data.sequence.ViewId> fusionOrder, Map<mpicbg.spim.data.sequence.ViewId,Coefficients> intensityAdjustmentCoefficients, net.imglib2.Interval fusionInterval, T type, int[] blockSize)	TODO javadoc
static <T extends net.imglib2.type.numeric.RealType<T> & net.imglib2.type.NativeType<T>, D extends net.imglib2.type.NativeType<D> & net.imglib2.type.numeric.RealType<D>> net.imglib2.algorithm.blocks.BlockSupplier<T>	initWithLoadedDfields(net.imglib2.converter.Converter<net.imglib2.type.numeric.real.FloatType,T> converter, mpicbg.spim.data.generic.sequence.BasicImgLoader imgLoader, Collection<? extends mpicbg.spim.data.sequence.ViewId> viewIds, Map<mpicbg.spim.data.sequence.ViewId,? extends mpicbg.spim.data.generic.sequence.BasicViewDescription<?>> viewDescriptions, Map<mpicbg.spim.data.sequence.ViewId,net.imglib2.realtransform.AffineTransform3D> approximateAffines, Map<mpicbg.spim.data.sequence.ViewId,net.imglib2.Interval> viewBounds, Map<mpicbg.spim.data.sequence.ViewId,net.imglib2.algorithm.blocks.dfield.DisplacementFields.TransformedDisplacementField<D>> rawDfields, FusionGUI.FusionType fusionType, double anisotropyFactor, int interpolationMethod, Comparator<mpicbg.spim.data.sequence.ViewId> fusionOrder, Map<mpicbg.spim.data.sequence.ViewId,Coefficients> intensityAdjustmentCoefficients, net.imglib2.Interval fusionInterval, T type, int[] blockSize)	Entry point that accepts pre-sampled (un-offset) displacement fields instead of building them from landmarks.
static net.imglib2.Interval	inverseTransformedBoundingBox(net.imglib2.realtransform.ThinplateSplineTransform transform, net.imglib2.Dimensions dimensions)	Get the bounding box in render coordinates of an image of the given dimension when back-projected through the inverse of the given transform from rendered to image coordinates.

Methods inherited from class java.lang.Object

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

Field Details

defaultIntervalExpansion

public static int defaultIntervalExpansion

defaultDisplacementFieldSpacing

public static final double[] defaultDisplacementFieldSpacing

Constructor Details

BlkThinPlateSplineFusion

public BlkThinPlateSplineFusion()

Method Details

init

public static <T extends net.imglib2.type.numeric.RealType<T> & net.imglib2.type.NativeType<T>>
net.imglib2.algorithm.blocks.BlockSupplier<T> init(net.imglib2.converter.Converter<net.imglib2.type.numeric.real.FloatType,T> converter,
 mpicbg.spim.data.generic.sequence.BasicImgLoader imgLoader,
 Collection<? extends mpicbg.spim.data.sequence.ViewId> viewIds,
 Map<mpicbg.spim.data.sequence.ViewId,? extends mpicbg.spim.data.generic.sequence.BasicViewDescription<?>> viewDescriptions,
 Map<mpicbg.spim.data.sequence.ViewId,Landmarks> viewLandmarks,
 FusionGUI.FusionType fusionType,
 double anisotropyFactor,
 int interpolationMethod,
 Comparator<mpicbg.spim.data.sequence.ViewId> fusionOrder,
 Map<mpicbg.spim.data.sequence.ViewId,Coefficients> intensityAdjustmentCoefficients,
 net.imglib2.Interval fusionInterval,
 T type,
 int[] blockSize)

TODO javadoc

Type Parameters:

T - output type

Parameters:

converter - converts from FloatType to the output type. Maybe null, in which case a default converter (clamp to output type range) is used.

imgLoader - TODO

viewIds - TODO

viewDescriptions - TODO

viewLandmarks - TODO

fusionType - how to combine pixels

anisotropyFactor -

interpolationMethod - 1==linear, 0==nearest neighbor

intensityAdjustmentCoefficients - intensity adjustments, can be null

fusionInterval - TODO

type - instance of the output type

blockSize - TODO

Returns:

initWithLoadedDfields

public static <T extends net.imglib2.type.numeric.RealType<T> & net.imglib2.type.NativeType<T>,
D extends net.imglib2.type.NativeType<D> & net.imglib2.type.numeric.RealType<D>>
net.imglib2.algorithm.blocks.BlockSupplier<T> initWithLoadedDfields(net.imglib2.converter.Converter<net.imglib2.type.numeric.real.FloatType,T> converter,
 mpicbg.spim.data.generic.sequence.BasicImgLoader imgLoader,
 Collection<? extends mpicbg.spim.data.sequence.ViewId> viewIds,
 Map<mpicbg.spim.data.sequence.ViewId,? extends mpicbg.spim.data.generic.sequence.BasicViewDescription<?>> viewDescriptions,
 Map<mpicbg.spim.data.sequence.ViewId,net.imglib2.realtransform.AffineTransform3D> approximateAffines,
 Map<mpicbg.spim.data.sequence.ViewId,net.imglib2.Interval> viewBounds,
 Map<mpicbg.spim.data.sequence.ViewId,net.imglib2.algorithm.blocks.dfield.DisplacementFields.TransformedDisplacementField<D>> rawDfields,
 FusionGUI.FusionType fusionType,
 double anisotropyFactor,
 int interpolationMethod,
 Comparator<mpicbg.spim.data.sequence.ViewId> fusionOrder,
 Map<mpicbg.spim.data.sequence.ViewId,Coefficients> intensityAdjustmentCoefficients,
 net.imglib2.Interval fusionInterval,
 T type,
 int[] blockSize)

Entry point that accepts pre-sampled (un-offset) displacement fields instead of building them from landmarks. Use this when the dfields are computed elsewhere (e.g. distributed via Spark and cached on disk) and loaded back here as full ArrayImgs. The per-block evaluation downstream is perf-tuned for ArrayImg-backed dfields, so always pass arrays (not lazy CachedCellImgs).

Parameters:

viewBounds - back-projected bounding box (render coordinates) per view, as produced by inverseTransformedBoundingBox(ThinplateSplineTransform, Dimensions).

rawDfields - un-offset displacement fields per view, as produced by DisplacementFields.sample(tps, viewBounds.get(viewId), spacing). Must contain entries at least for all views overlapping fusionInterval; extra entries (for non-overlapping views) are ignored.

concatenateBoundingBoxOffset

public static <D extends net.imglib2.type.NativeType<D> & net.imglib2.type.numeric.RealType<D>>
net.imglib2.algorithm.blocks.dfield.DisplacementFields.TransformedDisplacementField<D> concatenateBoundingBoxOffset(net.imglib2.algorithm.blocks.dfield.DisplacementFields.TransformedDisplacementField<D> dfield,
 net.imglib2.Interval boundingBoxInTarget)

fitAffineTransform

public static net.imglib2.realtransform.AffineTransform3D fitAffineTransform(double[][] source,
 double[][] target)

Fit an affine that approximates the TPS mapping from source to target landmarks. Used to derive the per-view "approximate affine" that drives blending-weight adjustment in FusionTools.adjustBlending(mpicbg.spim.data.generic.sequence.BasicViewDescription<?>, float[], float[], net.imglib2.realtransform.AffineTransform3D). The same fit is invoked by init on the driver and (when caching is wired up) the result is persisted to N5 so that distributed block-fusion tasks don't have to recompute landmarks.

inverseTransformedBoundingBox

public static net.imglib2.Interval inverseTransformedBoundingBox(net.imglib2.realtransform.ThinplateSplineTransform transform,
 net.imglib2.Dimensions dimensions)

Get the bounding box in render coordinates of an image of the given dimension when back-projected through the inverse of the given transform from rendered to image coordinates.

Parameters:

transform - forward transform (from rendered to image coordinates)

dimensions - image dimensions

Returns:

bounding box in render coordinates