com.sun.scenario.effect.impl.state.GaussianRenderState Maven / Gradle / Ivy
The newest version!
/*
* Copyright (c) 2014, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation. Oracle designates this
* particular file as subject to the "Classpath" exception as provided
* by Oracle in the LICENSE file that accompanied this code.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*/
package com.sun.scenario.effect.impl.state;
import com.sun.javafx.geom.Rectangle;
import com.sun.javafx.geom.transform.Affine2D;
import com.sun.javafx.geom.transform.BaseTransform;
import com.sun.javafx.geom.transform.NoninvertibleTransformException;
import com.sun.scenario.effect.Color4f;
import com.sun.scenario.effect.Filterable;
import com.sun.scenario.effect.ImageData;
import com.sun.scenario.effect.impl.BufferUtil;
import java.nio.FloatBuffer;
/**
*/
public class GaussianRenderState extends LinearConvolveRenderState {
public static final float MAX_RADIUS = (MAX_KERNEL_SIZE - 1) / 2;
// General variables representing the convolve operation
private boolean isShadow;
private Color4f shadowColor;
private float spread;
// Values specific to this operation, calculated from the rendering context
private EffectCoordinateSpace space;
private BaseTransform inputtx;
private BaseTransform resulttx;
private float inputRadiusX; // expected radius given inputtx
private float inputRadiusY;
private float spreadPass;
// Values specific to a given filter pass
private int validatedPass;
private PassType passType;
private float passRadius; // actual radius for src ImageData
private FloatBuffer weights;
private float samplevectors[]; // dx, dy for pixel sampling, both passes
private float weightsValidRadius;
private float weightsValidSpread;
static FloatBuffer getGaussianWeights(FloatBuffer weights,
int pad,
float radius,
float spread)
{
int r = pad;
int klen = (r * 2) + 1;
if (weights == null) {
weights = BufferUtil.newFloatBuffer(128);
}
weights.clear();
float sigma = radius / 3;
float sigma22 = 2 * sigma * sigma;
if (sigma22 < Float.MIN_VALUE) {
// Avoid divide by 0 below (it can generate NaN values).
sigma22 = Float.MIN_VALUE;
}
float total = 0.0F;
for (int row = -r; row <= r; row++) {
float kval = (float) Math.exp(-(row * row) / sigma22);
weights.put(kval);
total += kval;
}
total += (weights.get(0) - total) * spread;
for (int i = 0; i < klen; i++) {
weights.put(i, weights.get(i) / total);
}
int limit = getPeerSize(klen);
while (weights.position() < limit) {
weights.put(0.0F);
}
weights.limit(limit);
weights.rewind();
return weights;
}
/**
* Constructs a {@link RenderState} for a 2 dimensional Gaussian convolution.
*
* @param xradius the Gaussian radius along the user space X axis
* @param yradius the Gaussian radius along the user space Y axis
* @param spread the spread amount
* @param isShadow true if this is a shadow operation
* @param shadowColor the color of the shadow operation
* @param filtertx the transform applied to the filter operation
*/
public GaussianRenderState(float xradius, float yradius, float spread,
boolean isShadow, Color4f shadowColor, BaseTransform filtertx)
{
/*
* The operation starts as a description of the size of a (pair of)
* Gaussian kernels measured relative to that user space coordinate
* system and to be applied horizontally and vertically in that same
* space. The presence of a filter transform can mean that the
* direction we apply the gaussian convolutions could change as well
* as the new size of that Gaussian distribution curve relative to
* the pixels produced under that transform.
*
* We will track the direction and size of the Gaussian as we traverse
* different coordinate spaces with the intent that eventually we
* will perform the math of the convolution with weights calculated
* for one sample per pixel in the indicated direction and applied as
* closely to the intended final filter transform as we can achieve
* with the following caveats:
*
* - There is a maximum kernel size that the hardware pixel shaders
* can apply so we will try to keep the scaling of the filtered
* pixels low enough that we do not exceed that data limitation.
*
* - Software prefers to apply these weights along horizontal and
* vertical vectors, but can apply them in an arbitrary direction
* if need be.
*
* - If the Gaussian kernel is large enough, then applying a smaller
* Gaussian kernel to a downscaled input is indistinguishable to
* applying the larger kernel to a larger scaled input. Our maximum
* kernel size is large enough for this effect to be hidden if we
* max out the kernel.
*
* - We can tell the inputs what transform we want them to use, but
* they can always produce output under a different transform and
* then return a result with a "post-processing" trasnform to be
* applied (as we are doing here ourselves). Thus, we can plan
* how we want to apply the convolution weights and samples here,
* but we will have to reevaluate our actions when the actual
* input pixels are created later.
*
* - If we are blurring enough to trigger the MAX_RADIUS exceptions
* then we can blur at a nice axis-aligned orientation (which is
* preferred for the software versions of the shaders) and perform
* any rotation and skewing in the final post-processing result
* transform as that amount of blurring will quite effectively cover
* up any distortion that would occur by not rendering at the
* appropriate angles.
*
* To achieve this we start out with untransformed sample vectors
* which are unit vectors along the X and Y axes. We transform them
* into the requested filter space, adjust the kernel size and see
* if we can support that kernel size. If it is too large of a
* projected kernel, then we request the input at a smaller scale
* and perform a maximum kernel convolution on it and then indicate
* that this result will need to be scaled by the caller. When this
* method is done we will have computed what we need to do to the
* input pixels when they come in if the inputtx was honored, otherwise
* we may have to adjust the values further in {@link @validateInput()}.
*/
this.isShadow = isShadow;
this.shadowColor = shadowColor;
this.spread = spread;
if (filtertx == null) filtertx = BaseTransform.IDENTITY_TRANSFORM;
double mxx = filtertx.getMxx();
double mxy = filtertx.getMxy();
double myx = filtertx.getMyx();
double myy = filtertx.getMyy();
// Transformed unit axis vectors are essentially (mxx, myx) and (mxy, myy).
double txScaleX = Math.hypot(mxx, myx);
double txScaleY = Math.hypot(mxy, myy);
boolean scaled = false;
float scaledRadiusX = (float) (xradius * txScaleX);
float scaledRadiusY = (float) (yradius * txScaleY);
if (scaledRadiusX < MIN_EFFECT_RADIUS && scaledRadiusY < MIN_EFFECT_RADIUS) {
// Entire blur is essentially a NOP in device space, we should
// set up the values to force NOP processing rather than relying
// on calculations to do it for us.
this.inputRadiusX = 0.0f;
this.inputRadiusY = 0.0f;
this.spreadPass = 0;
this.space = EffectCoordinateSpace.RenderSpace;
this.inputtx = filtertx;
this.resulttx = BaseTransform.IDENTITY_TRANSFORM;
this.samplevectors = new float[] { 1.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f };
} else {
if (scaledRadiusX > MAX_RADIUS) {
scaledRadiusX = MAX_RADIUS;
txScaleX = MAX_RADIUS / xradius;
scaled = true;
}
if (scaledRadiusY > MAX_RADIUS) {
scaledRadiusY = MAX_RADIUS;
txScaleY = MAX_RADIUS / yradius;
scaled = true;
}
this.inputRadiusX = scaledRadiusX;
this.inputRadiusY = scaledRadiusY;
// We need to apply the spread on only one pass
// Prefer pass1 if r1 is not tiny (or at least bigger than r0)
// Otherwise use pass 0 so that it doesn't disappear
this.spreadPass = (inputRadiusY > 1f || inputRadiusY >= inputRadiusX) ? 1 : 0;
if (scaled) {
this.space = EffectCoordinateSpace.CustomSpace;
this.inputtx = BaseTransform.getScaleInstance(txScaleX, txScaleY);
this.resulttx = filtertx
.copy()
.deriveWithScale(1.0 / txScaleX, 1.0 / txScaleY, 1.0);
// assert resulttx.deriveWithConcatenation(inputtx).equals(filtertx)
this.samplevectors = new float[] { 1.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f };
} else {
this.space = EffectCoordinateSpace.RenderSpace;
this.inputtx = filtertx;
this.resulttx = BaseTransform.IDENTITY_TRANSFORM;
// These values should produce 2 normalized unit vectors in the
// direction of the transformed axis vectors.
this.samplevectors = new float[] { (float) (mxx / txScaleX),
(float) (myx / txScaleX),
(float) (mxy / txScaleY),
(float) (myy / txScaleY),
0.0f, 0.0f };
}
}
// If the input honors our requested transforms then samplevectors
// will be the unit vectors in the correct direction to sample by
// pixel distances in the input texture and the inputRadii will be
// the correct Gaussian dimension to blur them.
}
/**
* Constructs a {@link RenderState} for a single dimensional, directional
* Gaussian convolution (as for a MotionBlur operation).
*
* @param radius the Gaussian radius along the indicated direction
* @param dx the delta X of the unit vector along which to apply the convolution
* @param dy the delta Y of the unit vector along which to apply the convolution
* @param filtertx the transform applied to the filter operation
*/
public GaussianRenderState(float radius, float dx, float dy, BaseTransform filtertx) {
// This is a special case of the above 2 dimensional Gaussian, most of
// the same strategies and caveats apply except as relevant to our
// directional single-axis peculiarities
this.isShadow = false;
this.spread = 0.0f;
if (filtertx == null) filtertx = BaseTransform.IDENTITY_TRANSFORM;
double mxx = filtertx.getMxx();
double mxy = filtertx.getMxy();
double myx = filtertx.getMyx();
double myy = filtertx.getMyy();
// Manually transform the unit vector and determine its added "scale"
double tdx = mxx * dx + mxy * dy;
double tdy = myx * dx + myy * dy;
double txScale = Math.hypot(tdx, tdy);
boolean scaled = false;
float scaledRadius = (float) (radius * txScale);
if (scaledRadius < MIN_EFFECT_RADIUS) {
this.inputRadiusX = 0.0f;
this.inputRadiusY = 0.0f;
this.spreadPass = 0;
this.space = EffectCoordinateSpace.RenderSpace;
this.inputtx = filtertx;
this.resulttx = BaseTransform.IDENTITY_TRANSFORM;
this.samplevectors = new float[] { 1.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f };
} else {
if (scaledRadius > MAX_RADIUS) {
scaledRadius = MAX_RADIUS;
txScale = MAX_RADIUS / radius;
scaled = true;
}
this.inputRadiusX = scaledRadius;
this.inputRadiusY = 0.0f;
this.spreadPass = 0;
if (scaled) {
// Since this is a highly directed blur and any change in
// scaling perpendicular to the blur angle could result in
// visible artifacts not absorbed by the Gaussian convolution,
// we will try to focus any changes in intermediate scaling
// on just that direction that the blur is applied along.
// We will need to calculate 2 disjoint scale factors, one
// along the blur (already calculated in txScale) and one
// perpendicular to that vector, then we will provide the
// inputs with an animorphically scaled coordinate system
// that uses a smaller scale along the direction of the blur
// and as close as possible to the original scale in the
// orthogonal direction...
// Determine the orthogonal scale factor:
double odx = mxy * dx - mxx * dy;
double ody = myy * dx - myx * dy;
double txOScale = Math.hypot(odx, ody);
this.space = EffectCoordinateSpace.CustomSpace;
Affine2D a2d = new Affine2D();
a2d.scale(txScale, txOScale);
a2d.rotate(dx, -dy);
BaseTransform a2di;
try {
a2di = a2d.createInverse();
} catch (NoninvertibleTransformException ex) {
a2di = BaseTransform.IDENTITY_TRANSFORM;
}
this.inputtx = a2d;
this.resulttx = filtertx
.copy()
.deriveWithConcatenation(a2di);
// assert resulttx.deriveWithConcatenation(inputtx).equals(filtertx)
this.samplevectors = new float[] { 1.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f };
} else {
this.space = EffectCoordinateSpace.RenderSpace;
this.inputtx = filtertx;
this.resulttx = BaseTransform.IDENTITY_TRANSFORM;
// These values should produce a normalized unit vector in the
// direction of the transformed sample vector.
this.samplevectors = new float[] { (float) (tdx / txScale),
(float) (tdy / txScale),
0.0f, 0.0f, 0.0f, 0.0f };
}
}
// If the input honors our requested transforms then samplevectors
// will be the unit vector in the correct direction to sample by
// pixel distances in the input texture and the inputRadiusX will be
// the correct Gaussian dimension to blur them.
// The second vector in samplevectors is ignored since the associated
// inputRadiusY is hard-coded to 0.
}
@Override
public boolean isShadow() {
return isShadow;
}
@Override
public Color4f getShadowColor() {
return shadowColor;
}
@Override
public float[] getPassShadowColorComponents() {
return (validatedPass == 0)
? BLACK_COMPONENTS
: shadowColor.getPremultipliedRGBComponents();
}
@Override
public EffectCoordinateSpace getEffectTransformSpace() {
return space;
}
@Override
public BaseTransform getInputTransform(BaseTransform filterTransform) {
return inputtx;
}
@Override
public BaseTransform getResultTransform(BaseTransform filterTransform) {
return resulttx;
}
@Override
public Rectangle getInputClip(int i, Rectangle filterClip) {
if (filterClip != null) {
double dx0 = samplevectors[0] * inputRadiusX;
double dy0 = samplevectors[1] * inputRadiusX;
double dx1 = samplevectors[2] * inputRadiusY;
double dy1 = samplevectors[3] * inputRadiusY;
int padx = (int) Math.ceil(dx0+dx1);
int pady = (int) Math.ceil(dy0+dy1);
if ((padx | pady) != 0) {
filterClip = new Rectangle(filterClip);
filterClip.grow(padx, pady);
}
}
return filterClip;
}
@Override
public ImageData validatePassInput(ImageData src, int pass) {
this.validatedPass = pass;
Filterable f = src.getUntransformedImage();
BaseTransform srcTx = src.getTransform();
float iRadius = (pass == 0) ? inputRadiusX : inputRadiusY;
int vecindex = pass * 2;
if (srcTx.isTranslateOrIdentity()) {
// The input effect gave us exactly what we wanted, proceed as planned
this.passRadius = iRadius;
samplevectors[4] = samplevectors[vecindex];
samplevectors[5] = samplevectors[vecindex+1];
if (validatedPass == 0) {
if ( nearOne(samplevectors[4], f.getPhysicalWidth()) &&
nearZero(samplevectors[5], f.getPhysicalWidth()))
{
passType = PassType.HORIZONTAL_CENTERED;
} else {
passType = PassType.GENERAL_VECTOR;
}
} else {
if (nearZero(samplevectors[4], f.getPhysicalHeight()) &&
nearOne(samplevectors[5], f.getPhysicalHeight()))
{
passType = PassType.VERTICAL_CENTERED;
} else {
passType = PassType.GENERAL_VECTOR;
}
}
} else {
// The input produced a texture that requires transformation,
// reevaluate our radii.
// First (inverse) transform our sample vectors from the intended
// srcTx space back into the actual pixel space of the src texture.
// Then evaluate their length and attempt to absorb as much of any
// implicit scaling that would happen into our final pixelRadii,
// but if we overflow the maximum supportable radius then we will
// just have to sample sparsely with a longer than unit vector.
// REMIND: we should also downsample the texture by powers of
// 2 if our sampling will be more sparse than 1 sample per 2
// pixels.
passType = PassType.GENERAL_VECTOR;
try {
srcTx.inverseDeltaTransform(samplevectors, vecindex, samplevectors, 4, 1);
} catch (NoninvertibleTransformException ex) {
this.passRadius = 0.0f;
samplevectors[4] = samplevectors[5] = 0.0f;
return src;
}
double srcScale = Math.hypot(samplevectors[4], samplevectors[5]);
float pRad = (float) (iRadius * srcScale);
if (pRad > MAX_RADIUS) {
pRad = MAX_RADIUS;
srcScale = MAX_RADIUS / iRadius;
}
this.passRadius = pRad;
// For a pixelRadius that was less than MAX_RADIUS, the following
// lines renormalize the un-transformed vectors back into unit
// vectors in the proper direction and we absorbed their length
// into the pixelRadius that we will apply for the Gaussian weights.
// If we clipped the pixelRadius to MAX_RADIUS, then they will not
// actually end up as unit vectors, but they will represent the
// proper sampling deltas for the indicated radius (which should
// be MAX_RADIUS in that case).
samplevectors[4] /= srcScale;
samplevectors[5] /= srcScale;
}
samplevectors[4] /= f.getPhysicalWidth();
samplevectors[5] /= f.getPhysicalHeight();
return src;
}
@Override
public Rectangle getPassResultBounds(Rectangle srcdimension, Rectangle outputClip) {
// Note that the pass vector and the pass radius may be adjusted for
// a transformed input, but our output will be in the untransformed
// "filter" coordinate space so we need to use the "input" values that
// are in that same coordinate space.
// The srcdimension is padded by the amount of extra data we produce
// for this pass.
// The outputClip is padded by the amount of extra input data we will
// need for subsequent passes to do their work.
double r = (validatedPass == 0) ? inputRadiusX : inputRadiusY;
int i = validatedPass * 2;
double dx = samplevectors[i+0] * r;
double dy = samplevectors[i+1] * r;
int padx = (int) Math.ceil(Math.abs(dx));
int pady = (int) Math.ceil(Math.abs(dy));
Rectangle ret = new Rectangle(srcdimension);
ret.grow(padx, pady);
if (outputClip != null) {
if (validatedPass == 0) {
// Pass 0 needs to retain any added area for Pass 1 to
// compute the bounds within the outputClip, so we expand
// the outputClip accordingly.
dx = samplevectors[2] * r;
dy = samplevectors[3] * r;
padx = (int) Math.ceil(Math.abs(dx));
pady = (int) Math.ceil(Math.abs(dy));
if ((padx | pady) != 0) {
outputClip = new Rectangle(outputClip);
outputClip.grow(padx, pady);
}
}
ret.intersectWith(outputClip);
}
return ret;
}
@Override
public PassType getPassType() {
return passType;
}
@Override
public float[] getPassVector() {
float xoff = samplevectors[4]; // / srcNativeBounds.width;
float yoff = samplevectors[5]; // / srcNativeBounds.height;
int ksize = getPassKernelSize();
int center = ksize / 2;
float ret[] = new float[4];
ret[0] = xoff;
ret[1] = yoff;
ret[2] = -center * xoff;
ret[3] = -center * yoff;
return ret;
}
@Override
public int getPassWeightsArrayLength() {
validateWeights();
return weights.limit() / 4;
}
@Override
public FloatBuffer getPassWeights() {
validateWeights();
weights.rewind();
return weights;
}
@Override
public int getInputKernelSize(int pass) {
return 1 + 2 * (int) Math.ceil((pass == 0) ? inputRadiusX : inputRadiusY);
}
@Override
public int getPassKernelSize() {
return 1 + 2 * (int) Math.ceil(passRadius);
}
@Override
public boolean isNop() {
if (isShadow) return false;
return inputRadiusX < MIN_EFFECT_RADIUS
&& inputRadiusY < MIN_EFFECT_RADIUS;
}
@Override
public boolean isPassNop() {
if (isShadow && validatedPass == 1) return false;
return (passRadius) < MIN_EFFECT_RADIUS;
}
private void validateWeights() {
float r = passRadius;
float s = (validatedPass == spreadPass) ? spread : 0f;
if (weights == null ||
weightsValidRadius != r ||
weightsValidSpread != s)
{
weights = getGaussianWeights(weights, (int) Math.ceil(r), r, s);
weightsValidRadius = r;
weightsValidSpread = s;
}
}
}