Skip to content

I
imagej-elphel
Commit 5b0abf38 authored by Andrey Filippov's avatar Andrey Filippov
Browse files
Options

Bug fixes/improvements in "Calculate grids"

parent d3b2b2c6
Show whitespace changes
Inline Side-by-side
Showing with 269 additions and 52 deletions
src/main/java/Aberration_Calibration.java
View file @ 5b0abf38
......@@ -180,7 +180,10 @@ public class Aberration_Calibration extends PlugInFrame implements ActionListene
0.4, // corrRingWidth - ring (around r=0.5 dist to opposite corr) width , center circle r=0.5*PATTERN_DETECT.corrRingWidth
5.0, // minCorrContrast - discrimination threshold between good and bad pattern correlation
0.0, // minGridPeriod
0.0 // maxGridPeriod
0.0, // maxGridPeriod
0.0, // debugX+"");
0.0, // debugY+"");
-1.0 // this.debugRadius+"");
);
......@@ -313,9 +316,14 @@ public static MatchSimulatedPattern.DistortionParameters DISTORTION =new MatchSi
1.5, // 2.5, //6.0, // 2.0, // 0.0, // correlationHighPassSigma, - pixels in frequency domain
0.6, //2.0, //0.5, //0.0, // correlationLowPassSigma, - fraction of the frequency range
0.4, // correlationRingWidth- ring (around r=0.5 dist to opposite corr) width , center circle r=0.5*PATTERN_DETECT.corrRingWidth
3.0, // correlationMaxOffset, // maximal distance between predicted and actual pattern node
2.0, // increase back to .5? was needed with fisheye. 5.0, // double correlationMinContrast, // minimal contrast for the pattern to pass
2.5, // correlationMinInitialContrast, // minimal contrast for the pattern of the center (initial point)
8.0, // 3.0, // correlationMaxOffset, // maximal distance between predicted and actual pattern node
1.0, // 2.0, // increase back to .5? was needed with fisheye. 5.0, // double correlationMinContrast, // minimal contrast for the pattern to pass
1.5, // 2.5, // correlationMinInitialContrast, // minimal contrast for the pattern of the center (initial point)
0.8, // scaleFirstPassContrast, // Decrease contrast of cells that are too close to the border to be processed in rifinement pass
0.1, // contrastSelectSigma, // Gaussian sigma to select correlation centers (fraction of UV period), 0.1
0.5, //contrastAverageSigma, // Gaussian sigma to average correlation variations (as contrast reference) 0.5
150, // minimalPatternCluster minimal pattern cluster size (0 - disable retries)
2.0, // scaleMinimalInitialContrast increase/decrease minimal contrast if initial cluster is >0 but less than minimalPatternCluster
0.5, // when searching for grid, step this amount of the FFTSize
......@@ -16162,7 +16170,9 @@ private double [][] jacobianByJacobian(double [][] jacobian, boolean [] mask) {
WVgreens=matrix2x2_mul(patternMap[nTileY][nTileX],invConvMatrix);
contrast= matchSimulatedPattern.correlationContrast (model_corr, // square pixel array
WVgreens, // wave vectors (same units as the pixels array)
patternDetectParameters.corrRingWidth, // ring (around r=0.5 dist to opposite corr) width
// patternDetectParameters.corrRingWidth, // ring (around r=0.5 dist to opposite corr) width
0.1, // contrastSelectSigma
0.5, // contrastAverageSigma
0.0, // x0, // center coordinates
0.0, //y0,
title); // title base for optional plots names
......@@ -18951,10 +18961,10 @@ use the result to create a rejectiobn mask - if the energy was high, (multiplica
gd.addNumericField("Minimal pattern correlation contrast" , patternDetectParameters.minCorrContrast, 3); //5.0; // Discrimination threshold between good and bad pattern correleation
gd.addNumericField("Minimal pattern grid period (<=0.0 - do not check)" , patternDetectParameters.minGridPeriod, 2,5,"pix");
gd.addNumericField("Maximal pattern grid period (<=0.0 - do not check)" , patternDetectParameters.maxGridPeriod, 2,5,"pix");
// 0.0, // minGridPeriod
// 0.0 // maxGridPeriod
gd.addMessage("----- debug -----");
gd.addNumericField("Debug grid near pixel X" , patternDetectParameters.debugX, 1,6,"pix");
gd.addNumericField("Debug grid near pixel X" , patternDetectParameters.debugY, 1,6,"pix");
gd.addNumericField("Debug grid nodes at this distance of (x,Y) - <0 - no debug", patternDetectParameters.debugRadius, 1,5,"pix");
gd.showDialog();
if (gd.wasCanceled()) return false;
......@@ -18971,6 +18981,9 @@ use the result to create a rejectiobn mask - if the energy was high, (multiplica
patternDetectParameters.minCorrContrast= gd.getNextNumber();
patternDetectParameters.minGridPeriod= gd.getNextNumber();
patternDetectParameters.maxGridPeriod= gd.getNextNumber();
patternDetectParameters.debugX= gd.getNextNumber();
patternDetectParameters.debugY= gd.getNextNumber();
patternDetectParameters.debugRadius= gd.getNextNumber();
return true;
}
public boolean showPatternMinMaxPeriodDialog(MatchSimulatedPattern.PatternDetectParameters patternDetectParameters) {
......@@ -19209,6 +19222,10 @@ use the result to create a rejectiobn mask - if the energy was high, (multiplica
gd.addNumericField("Correlation minimal contrast:", distortionParameters.correlationMinContrast, 3);
gd.addNumericField("Correlation minimal contrast for initial search:", distortionParameters.correlationMinInitialContrast, 3);
gd.addNumericField("Decrease contrast of cells that are too close to the border to be processed in rifinement pass", distortionParameters.scaleFirstPassContrast, 3);
gd.addNumericField("Gaussian sigma to select correlation centers (fraction of UV period), 0.1", distortionParameters.contrastSelectSigma, 3);
gd.addNumericField("Gaussian sigma to average correlation variations (as contrast reference), 0.5", distortionParameters.contrastAverageSigma, 3);
gd.addNumericField("Minimal initial pattern cluster size (0 - disable retries)", distortionParameters.minimalPatternCluster, 0);
gd.addNumericField("Scale minimal contrast if the initial cluster is nonzero but smaller", distortionParameters.scaleMinimalInitialContrast, 3);
gd.addNumericField("Overlap of FFT areas when searching for pattern", distortionParameters.searchOverlap, 3);
......@@ -19284,6 +19301,10 @@ use the result to create a rejectiobn mask - if the energy was high, (multiplica
distortionParameters.correlationMinContrast= gd.getNextNumber();
distortionParameters.correlationMinInitialContrast= gd.getNextNumber();
distortionParameters.scaleFirstPassContrast= gd.getNextNumber();
distortionParameters.contrastSelectSigma= gd.getNextNumber();
distortionParameters.contrastAverageSigma= gd.getNextNumber();
distortionParameters.minimalPatternCluster=(int) gd.getNextNumber();
distortionParameters.scaleMinimalInitialContrast=gd.getNextNumber();
distortionParameters.searchOverlap= gd.getNextNumber();
......
src/main/java/MatchSimulatedPattern.java
View file @ 5b0abf38
......@@ -962,14 +962,19 @@ public class MatchSimulatedPattern {
/* ======================================================================== */
public double correlationContrast ( double [] pixels, // square pixel array
double [][] wVectors, // wave vectors (same units as the pixels array)
double ringWidth, // ring (around r=0.5 dist to opposite corr) width
// double ringWidth, // ring (around r=0.5 dist to opposite corr) width
double contrastSelectSigma, // Gaussian sigma to select correlation centers (fraction of UV period), 0.1
double contrastAverageSigma, // Gaussian sigma to average correlation variations (as contrast reference) 0.5
double x0, // center coordinates
double y0,
String title){
return correlationContrast (
pixels, // square pixel array
wVectors, // wave vectors (same units as the pixels array)
ringWidth, // ring (around r=0.5 dist to opposite corr) width
// ringWidth, // ring (around r=0.5 dist to opposite corr) width
contrastSelectSigma, // Gaussian sigma to select correlation centers (fraction of UV period), 0.1
contrastAverageSigma, // Gaussian sigma to average correlation variations (as contrast reference) 0.5
x0, // center coordinates
y0,
title, // title base for optional plots names
......@@ -1061,16 +1066,25 @@ public class MatchSimulatedPattern {
return contrast;
}
public double correlationContrast ( double [] pixels, // square pixel array
public double correlationContrast (
double [] pixels, // square pixel array
double [][] wVectors, // wave vectors (same units as the pixels array)
double ringWidth, // ring (around r=0.5 dist to opposite corr) width
double sigma,
double sigmaNorm, // to measure variations for normalization of the contrast
double x0, // center coordinates
double y0,
String title, // title base for optional plots names
int debugLevel){
double sigma=0.1;
// TODO: make configurable parameters
// double sigma=0.1;
// double sigmaNorm=0.5; // to measure variations for normalization of the contrast
double sigma32=9*sigma*sigma;
double k=-0.5/(sigma*sigma);
double sigmaNorm32=9*sigmaNorm*sigmaNorm;
double kNorm=-0.5/(sigmaNorm*sigmaNorm);
double [][] sampleCentersXY={{0.0,0.0},{0.0,0.5},{0.5,0.0},{0.0,-0.5},{-0.5,0.0}};
int [] sampleTypes = {0,1,1,1,1};
int size=(int) Math.sqrt(pixels.length);
......@@ -1084,9 +1098,11 @@ public class MatchSimulatedPattern {
for (int n=0;n<dbgMask.length;n++) dbgMask[n]=0.0;
double [] s={0.0,0.0};
double [] w={0.0,0.0};
double S0=0.0,S1=0.0,S2=0.0;
for (i=0;i<size;i++) {
xy[1]=i-size/2-y0;
for (j=0;j<size;j++) {
int index=i*size+j;
xy[0]=j-size/2-x0;
uv=matrix2x2_mul(wVectors,xy);
for (int np=0;np<sampleCentersXY.length;np++){
......@@ -1095,10 +1111,17 @@ public class MatchSimulatedPattern {
r2=dx*dx+dy*dy;
if (r2<sigma32){
double m=Math.exp(k*r2);
dbgMask[i*size+j]+=m;
dbgMask[index]+=m;
w[sampleTypes[np]]+=m;
s[sampleTypes[np]]+=m*pixels[i*size+j];
s[sampleTypes[np]]+=m*pixels[index];
}
}
r2=uv[0]*uv[0]+uv[1]*uv[1];
if (r2<sigmaNorm32){
double m=Math.exp(kNorm*r2);
S0+=m;
S1+=m*pixels[index];
S2+=m*pixels[index]*pixels[index];
}
}
}
......@@ -1106,8 +1129,16 @@ public class MatchSimulatedPattern {
if (debugLevel>1) System.out.println("Not enough data for correlation contrast: center - w[0]="+w[0]+" opposite - w[1]="+w[1]);
return -1.0;
}
double contrast=Math.sqrt((s[0]/w[0])/(s[1]/w[1]));
double ref=Math.sqrt(S2*S0-S1*S1)/S0;
double contrast=((s[0]/w[0]) -(s[1]/w[1]))/ref;
if (debugLevel>2){
System.out.println("correlationContrast() contrast="+contrast+" w[0]="+w[0]+" w[1]="+w[1]+" s[0]="+s[0]+" s[1]="+s[1]+" S0="+S0+" S1="+S1+" S2="+S2+" ref="+ref);
}
// if (contrast>3.0){
// System.out.println("correlationContrast() contrast="+contrast+" w[0]="+w[0]+" w[1]="+w[1]+" s[0]="+s[0]+" s[1]="+s[1]+" S0="+S0+" S1="+S1+" S2="+S2+" ref="+ref);
// }
// double contrast=Math.sqrt((s[0]/w[0]) /(s[1]/w[1]));
// double contrast=((s[0]/w[0]) -(s[1]/w[1]))/(size*size);
if (debugLevel>2) {
System.out.println("Correlation contrast is "+contrast);
double [][] dbgPixels={pixels,dbgMask};
......@@ -2897,10 +2928,23 @@ public class MatchSimulatedPattern {
this.PATTERN_GRID,
getWaveList (waveFrontList,i),
extrapolationWeights, // quadrant of sample weights
true, // useContrast
!distortionParameters.useQuadratic, // use linear approximation (instead of quadratic)
1.0E-10, // thershold ratio of matrix determinant to norm for linear approximation (det too low - fail)
1.0E-20 // thershold ratio of matrix determinant to norm for quadratic approximation (det too low - fail)
);
if (wave[i]==null) { // try w/o contrast, just x,y
wave[i]=estimateCell(
this.PATTERN_GRID,
getWaveList (waveFrontList,i),
extrapolationWeights, // quadrant of sample weights
false, // do not use Contrast, keep old contrast (even if it is NaN)
!distortionParameters.useQuadratic, // use linear approximation (instead of quadratic)
1.0E-10, // thershold ratio of matrix determinant to norm for linear approximation (det too low - fail)
1.0E-20 // thershold ratio of matrix determinant to norm for quadratic approximation (det too low - fail)
);
}
}
// set new values, removed failed cells (normally should not be any)
for (int i=wave.length-1;i>=0;i--) {
......@@ -3146,10 +3190,15 @@ public class MatchSimulatedPattern {
setPatternGridCell(
this.PATTERN_GRID,
centerUV,
centerXY,
centerXY, // contrast OK?
node[1],
node[2]);
putInWaveList(waveFrontList, centerUV, 0);
// Mark initial point as debug one (seems to be largest correction later
// patternDetectParameters.debugX=centerXY[0];
// patternDetectParameters.debugY=centerXY[1];
// System.out.println(">>>>>> Set debugX="+patternDetectParameters.debugX + " debugY="+patternDetectParameters.debugY);
} else { // create initial wave from the border nodes of existent grid
// start with clearing all invalid nodes
for (iUV[1]=0;iUV[1]<this.PATTERN_GRID.length;iUV[1]++) for (iUV[0]=0;iUV[0]<this.PATTERN_GRID[0].length;iUV[0]++)
......@@ -3201,6 +3250,9 @@ public class MatchSimulatedPattern {
final AtomicBoolean cleanup=new AtomicBoolean(false); // after the wave dies, it will be restored for all cells with defined neigbors to try again. maybe - try w/o therads?
final AtomicInteger debugCellSet= new AtomicInteger(0); // cells added at cleanup stage
// special case (most common, actually) when initial wave has 1 node. Remove it after processing
ArrayList<Integer> initialWave=new ArrayList<Integer>();
for (Integer I:waveFrontList) initialWave.add(I);
while (waveFrontList.size()>0) {
// process current list, add new wave layer (moving in one of the 4 directions)
......@@ -3225,16 +3277,6 @@ public class MatchSimulatedPattern {
// TODO: find how it could get negative coordinates
if ((ix<0) || (iy<0) || (ix>=distortionParameters.gridSize) || (iy>=distortionParameters.gridSize)) hasNeededNeighbor=false; //???
else hasNeededNeighbor=focusMask[iy*getImageWidth()+ix]; //* OOB -1624 java.lang.ArrayIndexOutOfBoundsException: -1624, at MatchSimulatedPattern.distortions(MatchSimulatedPattern.java:3063), at LensAdjustment.updateFocusGrid(LensAdjustment.java:121), at Aberration_Calibration.measurePSFMetrics(Aberration_Calibration.java:5994)
/*
Image timestamp=1330935278.115151
distortions(): this.PATTERN_GRID[0][45][0][0]=1617.777509046684
distortions(): this.PATTERN_GRID[0][45][0][1]=-1.8391484099869946
distortions(): ix=1618
distortions(): iy=-2
distortions(): focusMask.length=5018112
java.lang.ArrayIndexOutOfBoundsException: -3566
*/
}
for (dir=0;dir<directionsUV.length;dir++) {
iUV[0]=uvdir[0]+directionsUV[dir][0];
......@@ -3297,6 +3339,7 @@ java.lang.ArrayIndexOutOfBoundsException: -3566
PATTERN_GRID,
iUVdir,
extrapolationWeights, // quadrant of sample weights
true, // useContrast
!distortionParameters.useQuadratic, // use linear approximation (instead of quadratic)
1.0E-10, // thershold ratio of matrix determinant to norm for linear approximation (det too low - fail)
1.0E-20 // thershold ratio of matrix determinant to norm for quadratic approximation (det too low - fail)
......@@ -3498,6 +3541,17 @@ java.lang.ArrayIndexOutOfBoundsException: -3566
}
if (debugLevel>1) System.out.println("***** Starting cleanup, wave length="+waveFrontList.size());
}
// end of layer
if (initialWave!=null){ // just after the first layer (usually one cell) - delete it and add next time - otherwise first one needs large correction
if (debugLevel>0) System.out.println("Removing "+initialWave.size()+" initial wave cells");
while (initialWave.size()>0){
uvdir= getWaveList (initialWave,0);
clearPatternGridCell(PATTERN_GRID, uvdir);
initialWave.remove(0);
}
initialWave=null;
}
}//while (waveFrontList.size()>0)
debugLevel=was_debug_level;
/*
......@@ -3674,6 +3728,16 @@ java.lang.ArrayIndexOutOfBoundsException: -3566
}
/* ================================================================*/
public void scaleContrast(double scale){
for (double [][][] patternRow:this.PATTERN_GRID){
if (patternRow!=null) for (double [][] node:patternRow){
if ((node!=null) && (node.length>0) && (node[0]!=null) && (node[0].length>2)) {
node[0][2]*=scale;
}
}
}
}
/* ================================================================*/
public double refineDistortionCorrelation (
final DistortionParameters distortionParameters, //
final MatchSimulatedPattern.PatternDetectParameters patternDetectParameters,
......@@ -3684,6 +3748,7 @@ java.lang.ArrayIndexOutOfBoundsException: -3566
final int threadsMax,
final boolean updateStatus,
final int debug_level){// debug level used inside loops
scaleContrast(distortionParameters.scaleFirstPassContrast);
final double [][][][] patternGrid=this.PATTERN_GRID;
final int debugThreshold=1;
final Rectangle selection=new Rectangle(0, 0, imp.getWidth(), imp.getHeight());
......@@ -4192,12 +4257,21 @@ java.lang.ArrayIndexOutOfBoundsException: -3566
}
gridIntensity[iUV[1]][iUV[0]]=sum/sumW;
} else {
// trying alternative
// double [][][][] patternGrid_same=patternGrid;
gridIntensity[iUV[1]][iUV[0]]=Double.NaN;
if (isCellDefined(patternGrid,iUV[0],iUV[1])) {
double [][] patternCell=patternGrid[iUV[1]][iUV[0]];
if (patternCell[0].length>2) gridIntensity[iUV[1]][iUV[0]]=patternCell[0][2];
}
/*
gridIntensity[iUV[1]][iUV[0]]=localGridContrast(
imp,
equalizeGreens,
patternGrid,
iUV[0],
iUV[1]);
*/
}
}
}
......@@ -4751,6 +4825,16 @@ java.lang.ArrayIndexOutOfBoundsException: -3566
return locsNeib;
}
/* ======================================================================== */
public void zeroNaNContrast(){
for (double [][][] row:this.PATTERN_GRID){
for (double [][] node:row){
if ((node!=null) && (node.length>0) && (node[0]!=null) && (node[0].length>2)){
if (Double.isNaN(node[0][2])) node[0][2]=0.0;
}
}
}
}
public double[][][][] recalculateWaveVectors (
// double[][][][] patternGrid,
......@@ -4927,15 +5011,15 @@ java.lang.ArrayIndexOutOfBoundsException: -3566
private void setPatternGridCell(
double [][][][] grid,
int [] uv,
double [] xy,
double [] xy, // may be a 3-element, with contrast
double [] wv1,
double [] wv2){
int i;
initPatternGridCell(grid,uv);
if (xy!=null) {
double [] grid_xy= new double[2];
for (i=0;i<2;i++) grid_xy[i]=xy[i];
grid[uv[1]][uv[0]][0]= grid_xy;
// double [] grid_xy= new double[2];
// for (i=0;i<2;i++) grid_xy[i]=xy[i];
grid[uv[1]][uv[0]][0]= xy.clone(); // grid_xy;
}
if (wv1!=null) {
double [] grid_wv1= new double[2];
......@@ -5008,7 +5092,59 @@ java.lang.ArrayIndexOutOfBoundsException: -3566
int [] uv){
return isCellDefined(this.PATTERN_GRID,uv);
}
/*
// with contrast
private double getCellContrast(double [][][][] grid,
int [] uv){
if ((uv[1]>=0) && (uv[0]>=0) && (uv[1]<grid.length) && (uv[0]<grid[uv[1]].length) &&
(grid[uv[1]][uv[0]]!=null) && (grid[uv[1]][uv[0]][0]!=null) && (grid[uv[1]][uv[0]][0].length>2)) {
return grid[uv[1]][uv[0]][0][2];
} else {
return Double.NaN;
}
}
private double getCellContrast(double [][][][] grid,
int u,
int v){
if ((v>=0) && (u>=0) && (v<grid.length) && (u<grid[v].length) &&
(grid[v][u]!=null) && (grid[v][u][0]!=null) && (grid[v][u][0].length>2)) {
return grid[v][u][0][2];
} else {
return Double.NaN;
}
}
public double getCellContrast(int [] uv){
return getCellContrast(this.PATTERN_GRID,uv);
}
public double getCellContrast(int u, int v){
return getCellContrast(this.PATTERN_GRID,u,v);
}
private boolean isCellDefinedC(
double [][][][] grid,
int [] uv){
return ((uv[1]>=0) && (uv[0]>=0) && (uv[1]<grid.length) && (uv[0]<grid[uv[1]].length) &&
(grid[uv[1]][uv[0]]!=null) && (grid[uv[1]][uv[0]][0]!=null) && (grid[uv[1]][uv[0]][0].length>2) && !Double.isNaN(grid[uv[1]][uv[0]][0][2]));
}
private boolean isCellDefinedC(
double [][][][] grid,
int u,
int v){
return ((v>=0) && (u>=0) && (v<grid.length) && (u<grid[v].length) &&
(grid[v][u]!=null) && (grid[v][u][0]!=null) && (grid[v][u][0].length>2) && !Double.isNaN(grid[v][u][0][2]));
}
public boolean isCellDefinedC(
int u,
int v){
return isCellDefinedC(this.PATTERN_GRID,u,v);
}
public boolean isCellDefinedC(
int [] uv){
return isCellDefinedC(this.PATTERN_GRID,uv);
}
/*
private double [] cellXY(int u, int v){
if (!isCellDefined(u,v)) return null;
return this.PATTERN_GRID[v][u][0];
......@@ -5307,7 +5443,7 @@ java.lang.ArrayIndexOutOfBoundsException: -3566
String msg="Single correlation FFT size used: "+(1<<maxLn2);
if (global_debug_level>0) System.out.println(msg);
}
zeroNaNContrast(); // replace grid NaN with 0
int numPointers=(laserPointer!=null)?laserPointer.laserUVMap.length:0;
double [][] pointersXY=(numPointers>0)?getPointersXY(imp, numPointers):null;
......@@ -7356,9 +7492,11 @@ y=xy0[1] + dU*deltaUV[0]*(xy1[1]-xy0[1])+dV*deltaUV[1]*(xy2[1]-xy0[1])
}
// Verify contrast (if specified) - only for the center sample (numNeib==0)
if (numNeib==0) {
contrast= correlationContrast (modelCorr,
contrast= correlationContrast(modelCorr,
WVgreens, // wave vectors (same units as the pixels array)
distortionParameters.correlationRingWidth, // ring (around r=0.5 dist to opposite corr) width
// distortionParameters.correlationRingWidth, // ring (around r=0.5 dist to opposite corr) width
distortionParameters.contrastSelectSigma, // Gaussian sigma to select correlation centers (fraction of UV period), 0.1
distortionParameters.contrastAverageSigma, // Gaussian sigma to average correlation variations (as contrast reference) 0.5
//TODO: verify that displacement is correct here (sign, direction)
centerXY[0], // x0, // center coordinates
centerXY[1], //y0,
......@@ -7426,10 +7564,10 @@ y=xy0[1] + dU*deltaUV[0]*(xy1[1]-xy0[1])+dV*deltaUV[1]*(xy2[1]-xy0[1])
double [][] locsNeib, // locations and weights of neighbors to average
int debug_level
){
double dbg_x=1005.0;
double dbg_y=1284.0;
double dbg_tolerance=5.0;
boolean dbgThis=(Math.abs(beforeXY[0]-dbg_x)<dbg_tolerance) && (Math.abs(beforeXY[1]-dbg_y)<dbg_tolerance);
boolean dbgThis=
(Math.abs(beforeXY[0]-patternDetectParameters.debugX)<patternDetectParameters.debugRadius) &&
(Math.abs(beforeXY[1]-patternDetectParameters.debugY)<patternDetectParameters.debugRadius);
if (dbgThis) {
System.out.println("correctedPatternCrossLocationAverage4(), beforeXY[0]="+beforeXY[0]+", beforeXY[1]="+beforeXY[1]);
debug_level+=3;
......@@ -7669,9 +7807,12 @@ y=xy0[1] + dU*deltaUV[0]*(xy1[1]-xy0[1])+dV*deltaUV[1]*(xy2[1]-xy0[1])
for (int i=0;i<2;i++) for (int j=0;j<2;j++) WVgreens[i][j]*=0.5;
}
contrast= correlationContrast (modelCorr,
contrast= correlationContrast(modelCorr,
WVgreens, // wave vectors (same units as the pixels array)
distortionParameters.correlationRingWidth, // ring (around r=0.5 dist to opposite corr) width
// distortionParameters.correlationRingWidth, // ring (around r=0.5 dist to opposite corr) width
distortionParameters.contrastSelectSigma, // Gaussian sigma to select correlation centers (fraction of UV period), 0.1
distortionParameters.contrastAverageSigma, // Gaussian sigma to average correlation variations (as contrast reference) 0.5
//TODO: verify that displacement is correct here (sign, direction)
centerXY[0], // x0, // center coordinates
centerXY[1], //y0,
......@@ -8408,6 +8549,7 @@ d()/dy=C*x+2*B*y+E=0
double [][][][] grid,
int [] uv0,
double [][] weights, // quadrant of sample weights
boolean useContrast, // do not use cells with undefined contrast
boolean forceLinear, // use linear approximation (instead of quadratic)
double thresholdLin, // thershold ratio of matrix determinant to norm for linear approximation (det too low - fail)
double thresholdQuad // thershold ratio of matrix determinant to norm for quadratic approximation (det too low - fail)
......@@ -8422,7 +8564,7 @@ d()/dy=C*x+2*B*y+E=0
for (int iDv=-dist;iDv<=dist;iDv++) for (int iDu=-dist;iDu<=dist;iDu++) {
uv[0]=uv0[0]+iDu;
uv[1]=uv0[1]+iDv;
if (isCellDefined(grid,uv)) {
if ((!useContrast && isCellDefined(grid,uv)) || isCellDefinedC(grid,uv)) {
w=weights[(iDv>=0)?iDv:-iDv][(iDu>=0)?iDu:-iDu];
if (w!=0.0){
if (maxU<iDu) maxU=iDu;
......@@ -8434,13 +8576,16 @@ d()/dy=C*x+2*B*y+E=0
if (maxUmV<(iDu-iDv)) maxUmV= iDu-iDv;
if (minUmV>(iDu-iDv)) minUmV= iDu-iDv;
samples0[index][0]=new double[2];
samples0[index][1]=new double[2];
samples0[index][1]=new double[useContrast?3:2];
samples0[index][2]=new double[1];
samples0[index][2][0]=w;
samples0[index][0][0]=iDu;
samples0[index][0][1]=iDv;
samples0[index][1][0]=grid[uv[1]][uv[0]][0][0];
samples0[index][1][1]=grid[uv[1]][uv[0]][0][1];
if (useContrast){
samples0[index][1][2]=grid[uv[1]][uv[0]][0][2]; // contrast
}
index++;
}
}
......@@ -8463,6 +8608,14 @@ d()/dy=C*x+2*B*y+E=0
forceLinear || (diameter<5), // use linear approximation diameter <4 should be enough, 5 - just to be safe
thresholdLin, // thershold ratio of matrix determinant to norm for linear approximation (det too low - fail)
thresholdQuad); // thershold ratio of matrix determinant to norm for quadratic approximation (det too low - fail)
if ((estimatedCell==null) || (estimatedCell[0]==null) || useContrast) return estimatedCell;
double contrast=Double.NaN;
if (isCellDefined(grid,uv0)) {
double [] xycOld=grid[uv0[1]][uv0[0]][0];
if (xycOld.length>2) contrast=xycOld[2];
}
double [] xyc={estimatedCell[0][0],estimatedCell[0][1],contrast};
estimatedCell[0]=xyc;
return estimatedCell;
}
......@@ -8535,8 +8688,6 @@ d()/dy=C*x+2*B*y+E=0
System.out.println(i+": wv="+IJ.d2s(wv[i][0],3)+":"+IJ.d2s(wv[i][1],3));
}
}
return result;
}
// calculate simulation parameters for quadratic distortion of the pattern, compatible with SimulationPattern class
......@@ -8986,6 +9137,10 @@ error=Sum(W(x,y)*(F^2 + 2*F*(A*x^2+B*y^2+C*x*y+D*x+E*y-Z(x,y)) +(...) )
public double minCorrContrast;
public double minGridPeriod;
public double maxGridPeriod;
public double debugX;
public double debugY;
public double debugRadius;
public PatternDetectParameters(
double gaussWidth,
......@@ -9000,7 +9155,11 @@ error=Sum(W(x,y)*(F^2 + 2*F*(A*x^2+B*y^2+C*x*y+D*x+E*y-Z(x,y)) +(...) )
double corrRingWidth,
double minCorrContrast,
double minGridPeriod,
double maxGridPeriod) {
double maxGridPeriod,
double debugX,
double debugY,
double debugRadius
) {
this.gaussWidth=gaussWidth;
this.corrGamma = corrGamma;
this.corrSigma = corrSigma;
......@@ -9014,7 +9173,9 @@ error=Sum(W(x,y)*(F^2 + 2*F*(A*x^2+B*y^2+C*x*y+D*x+E*y-Z(x,y)) +(...) )
this.minCorrContrast = minCorrContrast;
this.minGridPeriod=minGridPeriod;
this.maxGridPeriod=maxGridPeriod;
this.debugX=debugX;
this.debugY=debugY;
this.debugRadius=debugRadius;
}
public void setProperties(String prefix,Properties properties){
......@@ -9031,7 +9192,11 @@ error=Sum(W(x,y)*(F^2 + 2*F*(A*x^2+B*y^2+C*x*y+D*x+E*y-Z(x,y)) +(...) )
properties.setProperty(prefix+"minCorrContrast",this.minCorrContrast+"");
properties.setProperty(prefix+"minGridPeriod",this.minGridPeriod+"");
properties.setProperty(prefix+"maxGridPeriod",this.maxGridPeriod+"");
properties.setProperty(prefix+"debugX",this.debugX+"");
properties.setProperty(prefix+"debugY",this.debugY+"");
properties.setProperty(prefix+"debugRadius",this.debugRadius+"");
}
public void getProperties(String prefix,Properties properties){
this.gaussWidth=Double.parseDouble(properties.getProperty(prefix+"gaussWidth"));
this.corrGamma=Double.parseDouble(properties.getProperty(prefix+"corrGamma"));
......@@ -9050,6 +9215,12 @@ error=Sum(W(x,y)*(F^2 + 2*F*(A*x^2+B*y^2+C*x*y+D*x+E*y-Z(x,y)) +(...) )
if (properties.getProperty(prefix+"maxGridPeriod")!=null)
this.minGridPeriod=Double.parseDouble(properties.getProperty(prefix+"maxGridPeriod"));
else this.maxGridPeriod=0.0;
if (properties.getProperty(prefix+"debugX")!=null)
this.debugX=Double.parseDouble(properties.getProperty(prefix+"debugX"));
if (properties.getProperty(prefix+"debugY")!=null)
this.debugY=Double.parseDouble(properties.getProperty(prefix+"debugY"));
if (properties.getProperty(prefix+"debugRadius")!=null)
this.debugRadius=Double.parseDouble(properties.getProperty(prefix+"debugRadius"));
}
}
......@@ -9072,8 +9243,13 @@ error=Sum(W(x,y)*(F^2 + 2*F*(A*x^2+B*y^2+C*x*y+D*x+E*y-Z(x,y)) +(...) )
public double correlationMaxOffset; // maximal distance between predicted and actual pattern node
public double correlationMinContrast; // minimal contrast for the pattern to pass
public double correlationMinInitialContrast; // minimal contrast for the pattern of the center (initial point)
public double scaleFirstPassContrast; // Decrease contrast of cells that are too close to the border to be processed in rifinement pass
public double contrastSelectSigma; // Gaussian sigma to select correlation centers (fraction of UV period), 0.1
public double contrastAverageSigma; // Gaussian sigma to average correlation variations (as contrast reference) 0.5
public int minimalPatternCluster; // minimal pattern cluster size (0 - disable retries)
public double scaleMinimalInitialContrast; // increase/decrease minimal contrast if initial cluster is >0 but less than minimalPatternCluster
public double searchOverlap; // when searching for grid, step this amount of the FFTSize
public int patternSubdiv;
public double correlationDx; // not saved
......@@ -9133,6 +9309,9 @@ error=Sum(W(x,y)*(F^2 + 2*F*(A*x^2+B*y^2+C*x*y+D*x+E*y-Z(x,y)) +(...) )
double correlationMaxOffset, // maximal distance between predicted and actual pattern node
double correlationMinContrast, // minimal contrast for the pattern to pass
double correlationMinInitialContrast, // minimal contrast for the pattern of the center (initial point)
double scaleFirstPassContrast, // Decrease contrast of cells that are too close to the border to be processed in rifinement pass
double contrastSelectSigma, // Gaussian sigma to select correlation centers (fraction of UV period), 0.1
double contrastAverageSigma, // Gaussian sigma to average correlation variations (as contrast reference) 0.5
int minimalPatternCluster, // minimal pattern cluster size (0 - disable retries)
double scaleMinimalInitialContrast, // increase/decrease minimal contrast if initial cluster is >0 but less than minimalPatternCluster
double searchOverlap, // when searching for grid, step this amount of the FFTSize
......@@ -9189,6 +9368,9 @@ error=Sum(W(x,y)*(F^2 + 2*F*(A*x^2+B*y^2+C*x*y+D*x+E*y-Z(x,y)) +(...) )
this.correlationMaxOffset=correlationMaxOffset;
this.correlationMinContrast=correlationMinContrast;
this.correlationMinInitialContrast=correlationMinInitialContrast;
this.scaleFirstPassContrast=scaleFirstPassContrast; // Decrease contrast of cells that are too close to the border to be processed in rifinement pass
this.contrastSelectSigma=contrastSelectSigma; // Gaussian sigma to select correlation centers (fraction of UV period), 0.1
this.contrastAverageSigma=contrastAverageSigma; // Gaussian sigma to average correlation variations (as contrast reference) 0.5
this.minimalPatternCluster=minimalPatternCluster; // minimal pattern cluster size (0 - disable retries)
this.scaleMinimalInitialContrast=scaleMinimalInitialContrast; // increase/decrease minimal contrast if initial cluster is >0 but less than minimalPatternCluster
this.searchOverlap=searchOverlap; // when searching for grid, step this amount of the FFTSize
......@@ -9247,6 +9429,9 @@ error=Sum(W(x,y)*(F^2 + 2*F*(A*x^2+B*y^2+C*x*y+D*x+E*y-Z(x,y)) +(...) )
this.correlationMaxOffset, // maximal distance between predicted and actual pattern node
this.correlationMinContrast, // minimal contrast for the pattern to pass
this.correlationMinInitialContrast,
this.scaleFirstPassContrast, // Decrease contrast of cells that are too close to the border to be processed in rifinement pass
this.contrastSelectSigma, // Gaussian sigma to select correlation centers (fraction of UV period), 0.1
this.contrastAverageSigma, // Gaussian sigma to average correlation variations (as contrast reference) 0.5
this.minimalPatternCluster, // minimal pattern cluster size (0 - disable retries)
this.scaleMinimalInitialContrast, // increase/decrease minimal contrast if initial cluster is >0 but less than minimalPatternCluster
this.searchOverlap, // when searching for grid, step this amount of the FFTSize
......@@ -9305,6 +9490,9 @@ error=Sum(W(x,y)*(F^2 + 2*F*(A*x^2+B*y^2+C*x*y+D*x+E*y-Z(x,y)) +(...) )
properties.setProperty(prefix+"correlationMaxOffset",this.correlationMaxOffset+"");
properties.setProperty(prefix+"correlationMinContrast",this.correlationMinContrast+"");
properties.setProperty(prefix+"correlationMinInitialContrast",this.correlationMinInitialContrast+"");
properties.setProperty(prefix+"scaleFirstPassContrast",this.scaleFirstPassContrast+"");
properties.setProperty(prefix+"contrastSelectSigma",this.contrastSelectSigma+"");
properties.setProperty(prefix+"contrastAverageSigma",this.contrastAverageSigma+"");
properties.setProperty(prefix+"minimalPatternCluster",this.minimalPatternCluster+"");
properties.setProperty(prefix+"scaleMinimalInitialContrast",this.scaleMinimalInitialContrast+"");
properties.setProperty(prefix+"searchOverlap",this.searchOverlap+"");
......@@ -9377,6 +9565,14 @@ error=Sum(W(x,y)*(F^2 + 2*F*(A*x^2+B*y^2+C*x*y+D*x+E*y-Z(x,y)) +(...) )
this.correlationMinContrast=Double.parseDouble(properties.getProperty(prefix+"correlationMinContrast"));
if (properties.getProperty(prefix+"correlationMinInitialContrast")!=null)
this.correlationMinInitialContrast=Double.parseDouble(properties.getProperty(prefix+"correlationMinInitialContrast"));
if (properties.getProperty(prefix+"scaleFirstPassContrast")!=null)
this.scaleFirstPassContrast=Double.parseDouble(properties.getProperty(prefix+"scaleFirstPassContrast"));
if (properties.getProperty(prefix+"contrastSelectSigma")!=null)
this.contrastSelectSigma=Double.parseDouble(properties.getProperty(prefix+"contrastSelectSigma"));
if (properties.getProperty(prefix+"contrastAverageSigma")!=null)
this.contrastAverageSigma=Double.parseDouble(properties.getProperty(prefix+"contrastAverageSigma"));
if (properties.getProperty(prefix+"minimalPatternCluster")!=null)
this.minimalPatternCluster=Integer.parseInt(properties.getProperty(prefix+"minimalPatternCluster"));
if (properties.getProperty(prefix+"scaleMinimalInitialContrast")!=null)
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment