Skip to content

I
imagej-elphel
Commit 75365299 authored by Andrey Filippov's avatar Andrey Filippov
Browse files
Options

matching images

parent 238bf728
Show whitespace changes
Inline Side-by-side
Showing with 682 additions and 127 deletions
src/main/java/DttRad2.java
View file @ 75365299
......@@ -489,6 +489,26 @@ public class DttRad2 {
set_unfold_2d(len);
}
// get current LT window as a 2d tile (2*size * 2*size)
public double [] getWin2d(){
int size = this.hwindow.length;
int size2 = 2*size;
double [] rslt = new double [4*size*size];
for (int i=0; i< size; i++){
int ia0 = i * size2;
int ia1 = (size2 - i -1) * size2;
for (int j=0; j< size; j++){
double d = this.hwindow[i] * this.hwindow[j];
rslt [ia0 + j] = d;
rslt [ia1 + j] = d;
rslt [ia0 + size2 - j -1] = d;
rslt [ia1 + size2 - j -1] = d;
}
}
return rslt;
}
// Convert 2nx2n overlapping tile to n*n for dct-iv
public double [] fold_tile(double [] x, int mode) { // x should be 2n*2n
return fold_tile(x, 1 << (ilog2(x.length/4)/2));
......
src/main/java/EyesisCorrectionParameters.java
View file @ 75365299
......@@ -1909,20 +1909,27 @@ public class EyesisCorrectionParameters {
public double max_corr_sigma = 1.5; // weights of points around global max to find fractional
// pixel location by quadratic approximation
public double max_corr_radius = 2.5; // maximal distance from int max to consider
public int corr_mode = 2; // which correlation mode to use: 0 - integer max, 1 - center of mass, 2 - polynomial
// pixel location by quadratic approximation
public double corr_border_contrast = 0.01; // contrast of dotted border on correlation results
public int tile_task_op = 0xff; // bitmask of operation modes applied to tiles (0 - nothing), bits TBD later
// +(0..f) - images, +(00.f0) - process pairs +256 - force disparity when combining images
// +(0..f) - images, +(00.f0) - process pairs + 256 - force disparity when combining images
// window to process tiles (later arbitrary masks will be generated to follow particular stages);
public int tile_task_wl = 0; //
public int tile_task_wt = 0; //
public int tile_task_ww = 324; //
public int tile_task_wh = 242; //
public double min_shot = 10.0; // Do not adjust for shot noise if lower than
public double scale_shot = 3.0; // scale when dividing by sqrt
public double diff_thershold = 5.0; // RMS difference from average to discard channel (~ 1.0 - 1/255 full scale image)
public double diff_sigma = 5.0; // RMS difference from average to reduce weights (~ 1.0 - 1/255 full scale image)
public double diff_threshold = 1.5; // RMS difference from average to discard channel (~ 1.0 - 1/255 full scale image)
public boolean diff_gauss = true; // when averaging images, use gaussian around average as weight (false - sharp all/nothing)
public double min_agree = 3.0; // minimal number of channels to agree on a point (real number to work with fuzzy averages)
public boolean keep_weights = true; // add port weights to RGBA stack (debug feature)
public boolean sharp_alpha = false; // combining mode for alpha channel: false - treat as RGB, true - apply center 8x8 only
public boolean gen_chn_img = false; // generate shifted channel images
public boolean show_nonoverlap = true; // show result RGBA before overlap combined (first channels, then RGBA combined?)
......@@ -1970,14 +1977,20 @@ public class EyesisCorrectionParameters {
properties.setProperty(prefix+"min_corr_normalized",this.min_corr_normalized +"");
properties.setProperty(prefix+"max_corr_sigma", this.max_corr_sigma +"");
properties.setProperty(prefix+"max_corr_radius", this.max_corr_radius +"");
properties.setProperty(prefix+"corr_mode", this.corr_mode+"");
properties.setProperty(prefix+"corr_border_contrast", this.corr_border_contrast +"");
properties.setProperty(prefix+"tile_task_op", this.tile_task_op+"");
properties.setProperty(prefix+"tile_task_wl", this.tile_task_wl+"");
properties.setProperty(prefix+"tile_task_wt", this.tile_task_wt+"");
properties.setProperty(prefix+"tile_task_ww", this.tile_task_ww+"");
properties.setProperty(prefix+"tile_task_wh", this.tile_task_wh+"");
properties.setProperty(prefix+"diff_thershold", this.diff_thershold +"");
properties.setProperty(prefix+"min_shot", this.min_shot +"");
properties.setProperty(prefix+"scale_shot", this.scale_shot +"");
properties.setProperty(prefix+"diff_sigma", this.diff_sigma +"");
properties.setProperty(prefix+"diff_threshold", this.diff_threshold +"");
properties.setProperty(prefix+"diff_gauss", this.diff_gauss+"");
properties.setProperty(prefix+"min_agree", this.min_agree +"");
properties.setProperty(prefix+"keep_weights", this.keep_weights+"");
properties.setProperty(prefix+"sharp_alpha", this.sharp_alpha+"");
properties.setProperty(prefix+"gen_chn_img", this.gen_chn_img+"");
properties.setProperty(prefix+"show_nonoverlap", this.show_nonoverlap+"");
......@@ -2022,14 +2035,20 @@ public class EyesisCorrectionParameters {
if (properties.getProperty(prefix+"min_corr_normalized")!=null)this.min_corr_normalized=Double.parseDouble(properties.getProperty(prefix+"min_corr_normalized"));
if (properties.getProperty(prefix+"max_corr_sigma")!=null) this.max_corr_sigma=Double.parseDouble(properties.getProperty(prefix+"max_corr_sigma"));
if (properties.getProperty(prefix+"max_corr_radius")!=null) this.max_corr_radius=Double.parseDouble(properties.getProperty(prefix+"max_corr_radius"));
if (properties.getProperty(prefix+"corr_mode")!=null) this.corr_mode=Integer.parseInt(properties.getProperty(prefix+"corr_mode"));
if (properties.getProperty(prefix+"corr_border_contrast")!=null) this.corr_border_contrast=Double.parseDouble(properties.getProperty(prefix+"corr_border_contrast"));
if (properties.getProperty(prefix+"tile_task_op")!=null) this.tile_task_op=Integer.parseInt(properties.getProperty(prefix+"tile_task_op"));
if (properties.getProperty(prefix+"tile_task_wl")!=null) this.tile_task_wl=Integer.parseInt(properties.getProperty(prefix+"tile_task_wl"));
if (properties.getProperty(prefix+"tile_task_wt")!=null) this.tile_task_wt=Integer.parseInt(properties.getProperty(prefix+"tile_task_wt"));
if (properties.getProperty(prefix+"tile_task_ww")!=null) this.tile_task_ww=Integer.parseInt(properties.getProperty(prefix+"tile_task_ww"));
if (properties.getProperty(prefix+"tile_task_wh")!=null) this.tile_task_wh=Integer.parseInt(properties.getProperty(prefix+"tile_task_wh"));
if (properties.getProperty(prefix+"diff_thershold")!=null) this.diff_thershold=Double.parseDouble(properties.getProperty(prefix+"diff_thershold"));
if (properties.getProperty(prefix+"diff_gauss")!=null) this.diff_gauss=Boolean.parseBoolean(properties.getProperty(prefix+"v"));
if (properties.getProperty(prefix+"min_shot")!=null) this.min_shot=Double.parseDouble(properties.getProperty(prefix+"min_shot"));
if (properties.getProperty(prefix+"scale_shot")!=null) this.scale_shot=Double.parseDouble(properties.getProperty(prefix+"scale_shot"));
if (properties.getProperty(prefix+"diff_sigma")!=null) this.diff_sigma=Double.parseDouble(properties.getProperty(prefix+"diff_sigma"));
if (properties.getProperty(prefix+"diff_threshold")!=null) this.diff_threshold=Double.parseDouble(properties.getProperty(prefix+"diff_threshold"));
if (properties.getProperty(prefix+"diff_gauss")!=null) this.diff_gauss=Boolean.parseBoolean(properties.getProperty(prefix+"diff_gauss"));
if (properties.getProperty(prefix+"min_agree")!=null) this.min_agree=Double.parseDouble(properties.getProperty(prefix+"min_agree"));
if (properties.getProperty(prefix+"keep_weights")!=null) this.keep_weights=Boolean.parseBoolean(properties.getProperty(prefix+"keep_weights"));
if (properties.getProperty(prefix+"sharp_alpha")!=null) this.sharp_alpha=Boolean.parseBoolean(properties.getProperty(prefix+"sharp_alpha"));
if (properties.getProperty(prefix+"gen_chn_img")!=null) this.gen_chn_img=Boolean.parseBoolean(properties.getProperty(prefix+"gen_chn_img"));
if (properties.getProperty(prefix+"show_nonoverlap")!=null)this.show_nonoverlap=Boolean.parseBoolean(properties.getProperty(prefix+"show_nonoverlap"));
......@@ -2076,6 +2095,7 @@ public class EyesisCorrectionParameters {
gd.addNumericField("Minimal correlation value to consider valid when normalizing results", this.min_corr_normalized, 6);
gd.addNumericField("Sigma for weights of points around global max to find fractional", this.max_corr_sigma, 3);
gd.addNumericField("Maximal distance from int max to consider", this.max_corr_radius, 3);
gd.addNumericField("Correlation mode: 0 - integer max, 1 - center of mass, 2 - polynomial", this.corr_mode, 0);
gd.addNumericField("Contrast of dotted border on correlation results", this.corr_border_contrast, 6);
gd.addMessage("--- tiles tasks ---");
gd.addNumericField("Tile operations bits: +(0..f) - images, +(00.f0) - process pairs +256, ... ", this.tile_task_op, 0);
......@@ -2083,8 +2103,15 @@ public class EyesisCorrectionParameters {
gd.addNumericField("Tile operations window top", this.tile_task_wt, 0);
gd.addNumericField("Tile operations window width", this.tile_task_ww, 0);
gd.addNumericField("Tile operations window height", this.tile_task_wh, 0);
gd.addNumericField("RMS difference from average to discard channel (255 full scale image)", this.diff_thershold, 4);
gd.addNumericField("Do not adjust for shot noise (~sqrt) if lower than this", this.min_shot, 4);
gd.addNumericField("Scale when dividing by sqrt for shot noise compensation of pixel differences (<0 - disable)", this.scale_shot, 4);
gd.addNumericField("RMS difference from average to reduce weights (255 full scale image)", this.diff_sigma, 4);
gd.addNumericField("RMS difference from average in sigmas to discard channel", this.diff_threshold, 4);
gd.addCheckbox ("Gaussian as weight when averaging images (false - sharp all/nothing)", this.diff_gauss);
gd.addNumericField("Minimal number of channels to agree on a point (real number to work with fuzzy averages)", this.min_agree, 2);
gd.addCheckbox ("Add port weights to RGBA stack (debug feature)", this.keep_weights);
gd.addCheckbox ("Alpha channel: use center 8x8 (unchecked - treat same as RGB)", this.sharp_alpha);
gd.addCheckbox ("Generate shifted channel images", this.gen_chn_img);
gd.addCheckbox ("Show result RGBA before overlap combined", this.show_nonoverlap);
......@@ -2131,14 +2158,20 @@ public class EyesisCorrectionParameters {
this.min_corr_normalized= gd.getNextNumber();
this.max_corr_sigma= gd.getNextNumber();
this.max_corr_radius= gd.getNextNumber();
this.corr_mode= (int) gd.getNextNumber();
this.corr_border_contrast= gd.getNextNumber();
this.tile_task_op= (int) gd.getNextNumber();
this.tile_task_wl= (int) gd.getNextNumber();
this.tile_task_wt= (int) gd.getNextNumber();
this.tile_task_ww= (int) gd.getNextNumber();
this.tile_task_wh= (int) gd.getNextNumber();
this.diff_thershold= gd.getNextNumber();
this.min_shot= gd.getNextNumber();
this.scale_shot= gd.getNextNumber();
this.diff_sigma= gd.getNextNumber();
this.diff_threshold= gd.getNextNumber();
this.diff_gauss= gd.getNextBoolean();
this.min_agree= gd.getNextNumber();
this.keep_weights= gd.getNextBoolean();
this.sharp_alpha= gd.getNextBoolean();
this.gen_chn_img= gd.getNextBoolean();
this.show_nonoverlap= gd.getNextBoolean();
......
src/main/java/EyesisDCT.java
View file @ 75365299
......@@ -4821,8 +4821,9 @@ public class EyesisDCT {
// for testing defined for a window, later the tiles to process will be calculated based on previous passes results
int [][] tile_op = new int [tilesY][tilesX]; // all zero
int txl = clt_parameters.tile_task_wl;
int txr = txl + clt_parameters.tile_task_wl;
int tyt = clt_parameters.tile_task_wl;
int txr = txl + clt_parameters.tile_task_ww;
int tyt = clt_parameters.tile_task_wt;
int tyb = tyt + clt_parameters.tile_task_wh;
if (txl < 0) txl = 0;
else if (txl >= tilesX) txl = tilesX - 1;
......@@ -4841,6 +4842,14 @@ public class EyesisDCT {
tile_op[i][j] = clt_parameters.tile_task_op;
}
}
if (debugLevel > -1){
System.out.println("clt_parameters.tile_task_wl="+clt_parameters.tile_task_wl );
System.out.println("clt_parameters.tile_task_wt="+clt_parameters.tile_task_wt );
System.out.println("clt_parameters.tile_task_ww="+clt_parameters.tile_task_ww );
System.out.println("clt_parameters.tile_task_wh="+clt_parameters.tile_task_wh );
System.out.println("tyt="+tyt+" tyb="+tyb+" txl="+txl+" txr="+txr );
}
//TODO: Add array of default disparity - use for combining images in force disparity mode (no correlation), when disparity is predicted from other tiles
double [][][][] clt_corr_combo = null;
......@@ -4849,7 +4858,7 @@ public class EyesisDCT {
// undecided, so 2 modes of combining alpha - same as rgb, or use center tile only
if (clt_parameters.correlate){
clt_corr_combo = new double [2][tilesY][tilesX][];
texture_tiles = new double [tilesY][tilesX]["RGBA".length()][];
texture_tiles = new double [tilesY][tilesX][][]; // ["RGBA".length()][];
for (int i = 0; i < tilesY; i++){
for (int j = 0; j < tilesX; j++){
clt_corr_combo[0][i][j] = null;
......@@ -4867,6 +4876,7 @@ public class EyesisDCT {
}
}
double [][] disparity_map = new double [8][]; //[0] -residual disparity, [1] - orthogonal (just for debugging)
double min_corr_selected = clt_parameters.corr_normalize? clt_parameters.min_corr_normalized: clt_parameters.min_corr;
double [][][][][][] clt_data = image_dtt.clt_aberrations_quad_corr(
tile_op, // per-tile operation bit codes
clt_parameters.disparity, // final double disparity,
......@@ -4883,11 +4893,19 @@ public class EyesisDCT {
clt_parameters.corr_red,
clt_parameters.corr_blue,
clt_parameters.corr_sigma,
clt_parameters.corr_mask,
// clt_parameters.corr_mask,
clt_parameters.corr_normalize, // normalize correlation results by rms
clt_parameters.corr_normalize? clt_parameters.min_corr_normalized: clt_parameters.min_corr, // 0.0001; // minimal correlation value to consider valid
min_corr_selected, // 0.0001; // minimal correlation value to consider valid
clt_parameters.max_corr_sigma,// 1.5; // weights of points around global max to find fractional
clt_parameters.max_corr_radius,
clt_parameters.corr_mode, // Correlation mode: 0 - integer max, 1 - center of mass, 2 - polynomial
clt_parameters.min_shot, // 10.0; // Do not adjust for shot noise if lower than
clt_parameters.scale_shot, // 3.0; // scale when dividing by sqrt ( <0 - disable correction)
clt_parameters.diff_sigma, // 5.0;//RMS difference from average to reduce weights (~ 1.0 - 1/255 full scale image)
clt_parameters.diff_threshold, // 5.0; // RMS difference from average to discard channel (~ 1.0 - 1/255 full scale image)
clt_parameters.diff_gauss, // true; // when averaging images, use gaussian around average as weight (false - sharp all/nothing)
clt_parameters.min_agree, // 3.0; // minimal number of channels to agree on a point (real number to work with fuzzy averages)
clt_parameters.keep_weights, // Add port weights to RGBA stack (debug feature)
geometryCorrection, // final GeometryCorrection geometryCorrection,
clt_kernels, // final double [][][][][][] clt_kernels, // [channel_in_quad][color][tileY][tileX][band][pixel] , size should match image (have 1 tile around)
clt_parameters.kernel_step,
......@@ -4902,16 +4920,18 @@ public class EyesisDCT {
// (clt_parameters.dbg_mode & 256) != 0, // transpose convolve
threadsMax,
debugLevel);
if (debugLevel > -1){
System.out.println("clt_data.length="+clt_data.length+" clt_data[0].length="+clt_data[0].length
+" clt_data[0][0].length="+clt_data[0][0].length+" clt_data[0][0][0].length="+
clt_data[0][0][0].length);
}
// +" clt_data[0][0][0][0].length="+clt_data[0][0][0][0].length+
// " clt_data[0][0][0][0][0].length="+clt_data[0][0][0][0][0].length);
// visualize texture tiles as RGBA slices
double [][] texture_nonoverlap = null;
double [][] texture_overlap = null;
String [] rgba_titles = {"red","green","blue","alpha"};
String [] rgba_titles = {"red","blue","green","alpha"};
String [] rgba_weights_titles = {"red","blue","green","alpha","port0","port1","port2","port3","r-rms","b-rms","g-rms","w-rms"};
if (texture_tiles != null){
if (clt_parameters.show_nonoverlap){
texture_nonoverlap = image_dtt.combineRGBATiles(
......@@ -4927,7 +4947,7 @@ public class EyesisDCT {
tilesY * (2 * clt_parameters.transform_size),
true,
name + "-TXTNOL-D"+clt_parameters.disparity,
rgba_titles );
(clt_parameters.keep_weights?rgba_weights_titles:rgba_titles));
}
if (clt_parameters.show_overlap){
......@@ -4944,8 +4964,7 @@ public class EyesisDCT {
tilesY * clt_parameters.transform_size,
true,
name + "-TXTOL-D"+clt_parameters.disparity,
rgba_titles );
(clt_parameters.keep_weights?rgba_weights_titles:rgba_titles));
}
}
......@@ -5037,7 +5056,7 @@ public class EyesisDCT {
System.out.println("--tilesX="+tilesX);
System.out.println("--tilesY="+tilesY);
}
if (debugLevel > 1){
if (debugLevel > 0){
double [][] clt = new double [clt_data[iQuad].length*4][];
for (int chn = 0; chn < clt_data[iQuad].length; chn++) {
double [][] clt_set = image_dtt.clt_dbg(
......@@ -5067,13 +5086,7 @@ public class EyesisDCT {
debugLevel);
}
if (debugLevel > 0) sdfa_instance.showArrays( // -1
iclt_data,
(tilesX + 0) * clt_parameters.transform_size,
(tilesY + 0) * clt_parameters.transform_size,
true,
results[iQuad].getTitle()+"-rbg_sigma");
if (debugLevel > 0) sdfa_instance.showArrays(iclt_data,
if (debugLevel > -1) sdfa_instance.showArrays(iclt_data,
(tilesX + 0) * clt_parameters.transform_size,
(tilesY + 0) * clt_parameters.transform_size,
true,
......
src/main/java/ImageDtt.java
View file @ 75365299
import java.util.concurrent.atomic.AtomicInteger;
import ij.IJ;
import ij.ImageStack;
/**
......@@ -28,6 +27,27 @@ import ij.ImageStack;
*/
public class ImageDtt {
static double [] kern_g={
0.0, 0.125, 0.0 ,
0.125, 0.5, 0.125,
0.0, 0.125, 0.0 };
static double [] kern_rb={
0.0625, 0.125, 0.0625,
0.125, 0.25, 0.125,
0.0625, 0.125, 0.0625};
// static double [][] kerns = {kern_rb,kern_rb,kern_g};
static int [][] corn_side_indices = { // of 012/345/678 3x3 square kernel
{4,5,7,8}, // top left corner
{3,4,5,6,7,8}, // top middle
{3,4,6,7}, // top right
{1,2,4,5,7,8}, // middle left
{0,1,2,3,4,5,6,7,8}, // middle
{0,1,3,4,6,7}, // middle right
{1,2,4,5}, // bottom left
{0,1,2,3,4,5}, // mottom middle
{0,1,3,4}}; // bottom right
public ImageDtt(){
}
......@@ -893,6 +913,11 @@ public class ImageDtt {
return clt_data;
}
/*
*
*/
public double [][][][][][] clt_aberrations_quad_corr(
final int [][] tile_op, // [tilesY][tilesX] - what to do - 0 - nothing for this tile
final double disparity,
......@@ -915,11 +940,19 @@ public class ImageDtt {
final double corr_red,
final double corr_blue,
final double corr_sigma,
final int corr_mask, // which pairs to combine in the combo: 1 - top, 2 bottom, 4 - left, 8 - right
// final int corr_mask, // which pairs to combine in the combo: 1 - top, 2 bottom, 4 - left, 8 - right
final boolean corr_normalize, // normalize correlation results by rms
final double min_corr, // 0.0001; // minimal correlation value to consider valid
final double max_corr_sigma, // = 1.5; // weights of points around global max to find fractional
final double max_corr_radius, // = 2.5;
final int corr_mode, // Correlation mode: 0 - integer max, 1 - center of mass, 2 - polynomial
final double min_shot, // 10.0; // Do not adjust for shot noise if lower than
final double scale_shot, // 3.0; // scale when dividing by sqrt ( <0 - disable correction)
final double diff_sigma, // 5.0;//RMS difference from average to reduce weights (~ 1.0 - 1/255 full scale image)
final double diff_threshold, // 5.0; // RMS difference from average to discard channel (~ 1.0 - 1/255 full scale image)
final boolean diff_gauss, // true; // when averaging images, use gaussian around average as weight (false - sharp all/nothing)
final double min_agree, // 3.0; // minimal number of channels to agree on a point (real number to work with fuzzy averages)
final boolean keep_weights, // Add port weights to RGBA stack (debug feature)
final GeometryCorrection geometryCorrection,
final double [][][][][][] clt_kernels, // [channel_in_quad][color][tileY][tileX][band][pixel] , size should match image (have 1 tile around)
final int kernel_step,
......@@ -937,6 +970,7 @@ public class ImageDtt {
{
final int quad = 4; // number of subcameras
final int numcol = 3; // number of colors
final int force_disparity_bit = 8; // move to parameters?
final int nChn = image_data[0].length;
final int height=image_data[0][0].length/width;
final int tilesX=width/transform_size;
......@@ -974,6 +1008,12 @@ public class ImageDtt {
{2,3,0},
{0,2,1},
{1,3,1}};
final double[][] port_offsets = {
{-0.5, -0.5},
{ 0.5, -0.5},
{-0.5, 0.5},
{ 0.5, 0.5}};
final int transform_len = transform_size * transform_size;
......@@ -985,7 +1025,7 @@ public class ImageDtt {
} else {
for (int i = 0; i < transform_size; i++){
for (int j = 0; j < transform_size; j++){
filter_direct[i*transform_size+j] = Math.exp(-(i*i+j*j)/(2*corr_sigma));
filter_direct[i*transform_size+j] = Math.exp(-(i*i+j*j)/(2*corr_sigma)); // FIXME: should be sigma*sigma !
}
}
}
......@@ -1008,19 +1048,14 @@ public class ImageDtt {
final double [] filter= dtt.dttt_iiie(filter_direct);
for (int i=0; i < filter.length;i++) filter[i] *= 2*transform_size;
// prepare disparity maps and weighs
// prepare disparity maps and weights
final int max_search_radius = (int) Math.abs(max_corr_radius); // use negative max_corr_radius for squares instead of circles?
final int max_search_radius_poly = 1;
if (globalDebugLevel > -1){
int numPairs = 0;
for (int pair = 0; pair < corr_pairs.length; pair++) if (((corr_mask >> pair) & 1) != 0){
numPairs++;
}
System.out.println("max_corr_radius= "+max_corr_radius);
System.out.println("max_search_radius="+max_search_radius);
System.out.println("corr_mask= "+corr_mask);
System.out.println("max_search_radius_poly="+max_search_radius_poly);
System.out.println("corr_fat_zero= "+corr_fat_zero);
System.out.println("numPairs= "+numPairs);
}
if (disparity_map != null){
......@@ -1028,8 +1063,22 @@ public class ImageDtt {
disparity_map[i] = new double [tilesY*tilesX];
}
}
final double [] corr_max_weights =((max_corr_sigma > 0) && (disparity_map != null))?
setMaxXYWeights(max_corr_sigma,max_search_radius): null; // here use square anyway
// final double [] corr_max_weights =(((max_corr_sigma > 0) && (disparity_map != null))?
// setMaxXYWeights(max_corr_sigma,max_search_radius): null); // here use square anyway
final double [] corr_max_weights_poly =(((max_corr_sigma > 0) && (disparity_map != null))?
setMaxXYWeights(max_corr_sigma,max_search_radius_poly): null); // here use square anyway
dtt.set_window(window_type);
final double [] lt_window = dtt.getWin2d(); // [256]
final double [] lt_window2 = new double [lt_window.length]; // squared
for (int i = 0; i < lt_window.length; i++) lt_window2[i] = lt_window[i] * lt_window[i];
if (globalDebugLevel > 1) {
showDoubleFloatArrays sdfa_instance = new showDoubleFloatArrays(); // just for debugging?
sdfa_instance.showArrays(lt_window, 2*transform_size, 2*transform_size, "lt_window");
}
for (int ithread = 0; ithread < threads.length; ithread++) {
......@@ -1047,14 +1096,21 @@ public class ImageDtt {
double [][][] tcorr_partial = null; // [quad][numcol+1][15*15]
double [][][][] tcorr_tpartial = null; // [quad][numcol+1][4][8*8]
PolynomialApproximation pa = null;
if (corr_max_weights !=null) pa = new PolynomialApproximation(0); // debug level
if (corr_max_weights_poly !=null) pa = new PolynomialApproximation(0); // debug level
for (int nTile = ai.getAndIncrement(); nTile < nTilesInChn; nTile = ai.getAndIncrement()) {
tileY = nTile /tilesX;
tileX = nTile % tilesX;
if (tile_op[tileY][tileX] == 0) continue; // nothing to do for this tile
int img_mask = (tile_op[tileY][tileX] >> 0) & 0xf; // which images to use
int corr_mask = (tile_op[tileY][tileX] >> 4) & 0xf; // which pairs to combine in the combo: 1 - top, 2 bottom, 4 - left, 8 - right
// mask out pairs that use missing channels
for (int i = 0; i< corr_pairs.length; i++){
if ((((1 << corr_pairs[i][0]) & img_mask) == 0) || (((1 << corr_pairs[i][1]) & img_mask) == 0)) {
corr_mask &= ~ (1 << i);
}
}
boolean debugTile =(tileX == debug_tileX) && (tileY == debug_tileY);
for (int chn = 0; chn <numcol; chn++) {
centerX = tileX * transform_size + transform_size/2 - shiftX;
centerY = tileY * transform_size + transform_size/2 - shiftY;
double [][] centersXY = geometryCorrection.getPortsCoordinates(
......@@ -1125,6 +1181,7 @@ public class ImageDtt {
}
}
// all color channels are done here
double extra_disparity = 0.0; // if allowed, shift images extra before trying to combine
if (clt_corr_combo != null){ // not null - calculate correlations
tcorr_tpartial=new double[corr_pairs.length][numcol+1][4][transform_len];
......@@ -1304,26 +1361,24 @@ public class ImageDtt {
min_corr, // minimal value to consider (at integer location, not interpolated)
debugMax);
if (icorr_max != null){
double [] corr_max_XYi = {icorr_max[0],icorr_max[1]};
disparity_map[0][tileY*tilesX + tileX] = transform_size - 1 -corr_max_XYi[0];
disparity_map[1][tileY*tilesX + tileX] = transform_size - 1 -corr_max_XYi[1];
} else {
if (icorr_max == null){
disparity_map[0][tileY*tilesX + tileX] = Double.NaN;
disparity_map[1][tileY*tilesX + tileX] = Double.NaN;
disparity_map[2][tileY*tilesX + tileX] = Double.NaN;
disparity_map[3][tileY*tilesX + tileX] = Double.NaN;
disparity_map[4][tileY*tilesX + tileX] = Double.NaN;
disparity_map[5][tileY*tilesX + tileX] = Double.NaN;
continue;
}
} else {
double [] corr_max_XYi = {icorr_max[0],icorr_max[1]};
disparity_map[0][tileY*tilesX + tileX] = transform_size - 1 -corr_max_XYi[0];
disparity_map[1][tileY*tilesX + tileX] = transform_size - 1 -corr_max_XYi[1];
// for the integer maximum provide contrast and variety
int max_index = icorr_max[1]*corr_size + icorr_max[0];
disparity_map[6][tileY*tilesX + tileX] = tcorr_combo[0][max_index]; // correlation combo value at the integer maximum
// undo scaling caused by optional normalization
disparity_map[7][tileY*tilesX + tileX] = (rms[1]*tcorr_combo[1][max_index])/(rms[0]*tcorr_combo[0][max_index]); // correlation combo value at the integer maximum
// Calculate "center of mass" coordinates
// Calculate "center of mass" coordinates
double [] corr_max_XYm = getMaxXYCm( // get fractiona center as a "center of mass" inside circle/square from the integer max
tcorr_combo[0], // [data_size * data_size]
corr_size,
......@@ -1338,15 +1393,15 @@ public class ImageDtt {
disparity_map[3][tileY*tilesX + tileX] = Double.NaN;
}
// Calculate polynomial interpolated maximum coordinates
// Calculate polynomial interpolated maximum coordinates
double [] corr_max_XY = getMaxXYPoly( // get interpolated maximum coordinates using 2-nd degree polynomial
pa,
tcorr_combo[0], // [data_size * data_size]
corr_size,
icorr_max, // integer center coordinates (relative to top left)
corr_max_weights, // [(radius+1) * (radius+1)]
max_search_radius,
corr_max_weights_poly, // [(radius+1) * (radius+1)]
max_search_radius_poly, // max_search_radius, for polynomial - always use 1
debugMax);
if (corr_max_XY != null){
disparity_map[4][tileY*tilesX + tileX] = transform_size - 1 -corr_max_XY[0];
......@@ -1355,10 +1410,134 @@ public class ImageDtt {
disparity_map[4][tileY*tilesX + tileX] = Double.NaN;
disparity_map[5][tileY*tilesX + tileX] = Double.NaN;
}
if (corr_mode == 0) extra_disparity = disparity_map[0][tileY*tilesX + tileX];
else if (corr_mode == 1) extra_disparity = disparity_map[2][tileY*tilesX + tileX];
else if (corr_mode == 2) extra_disparity = disparity_map[4][tileY*tilesX + tileX];
if (Double.isNaN(extra_disparity)) extra_disparity = 0;
}
}
}
} // end of if (clt_corr_combo != null)
if ((extra_disparity != 0) && (((1 << force_disparity_bit) & tile_op[tileY][tileX]) == 0)){ // 0 - adjust disparity, 1 - uase provided
// shift images by 0.5 * extra disparity in the diagonal direction
for (int chn = 0; chn <numcol; chn++) { // color
for (int i = 0; i < quad; i++) {
fract_shift( // fractional shift in transform domain. Currently uses sin/cos - change to tables with 2? rotations
clt_data[i][chn][tileY][tileX], // double [][] clt_tile,
transform_size,
extra_disparity * port_offsets[i][0], // double shiftX,
extra_disparity * port_offsets[i][1], // double shiftY,
// (globalDebugLevel > 0) && (tileX == debug_tileX) && (tileY == debug_tileY)); // external tile compare
((globalDebugLevel > 0) && (chn==0) && (tileX >= debug_tileX - 2) && (tileX <= debug_tileX + 2) &&
(tileY >= debug_tileY - 2) && (tileY <= debug_tileY+2)));
}
}
}
// lpf tiles (same as images before)
// iclt tiles
double [][][] iclt_tile = new double [quad][numcol][];
double [] clt_tile;
double scale = 0.25; // matching iclt_2d
for (int i = 0; i < quad; i++) {
for (int chn = 0; chn <numcol; chn++) { // color
// double [] clt_tile = new double [transform_size*transform_size];
for (int dct_mode = 0; dct_mode < 4; dct_mode++){
clt_tile = clt_data[i][chn][tileY][tileX][dct_mode].clone();
// lpf each of the 4 quadrants before idct
for (int j = 0; j < filter.length; j++){
clt_tile[j] *= scale*filter[j];
}
// IDCT-IV should be in reversed order: CC->CC, SC->CS, CS->SC, SS->SS
int idct_mode = ((dct_mode << 1) & 2) | ((dct_mode >> 1) & 1);
clt_tile = dtt.dttt_iv (clt_tile, idct_mode, transform_size);
// iclt_tile[i][chn] = dtt.dttt_iv (clt_data[i][chn][tileY][tileX][dct_mode], idct_mode, transform_size);
double [] tile_mdct = dtt.unfold_tile(clt_tile, transform_size, dct_mode); // mode=0 - DCCT 16x16
// accumulate partial mdct results
if (dct_mode == 0){
iclt_tile[i][chn] = tile_mdct;
} else{
for (int j = 0; j<tile_mdct.length; j++){
iclt_tile[i][chn][j] += tile_mdct[j]; // matching iclt_2d
}
}
}
}
}
if ((globalDebugLevel > 0) && debugTile) {
showDoubleFloatArrays sdfa_instance = new showDoubleFloatArrays(); // just for debugging?
String [] titles = {"red0","blue0","green0","red1","blue1","green1","red2","blue2","green2","red3","blue3","green3"};
double [][] dbg_tile = new double [quad*numcol][];
for (int i = 0; i < quad; i++) {
for (int chn = 0; chn <numcol; chn++) { // color
dbg_tile[i * numcol + chn] = iclt_tile[i][chn];
}
}
sdfa_instance.showArrays(dbg_tile, 2* transform_size, 2* transform_size, true, "iclt_x"+tileX+"_y"+tileY, titles);
}
// "de-bayer" tiles for matching, use original data for output
double [][][] tiles_debayered = new double [quad][numcol][];
for (int i =0; i<quad; i++){
for (int chn = 0; chn < numcol; chn++){
// tiles_debayered[i][chn] = tile_debayer(
// (chn != 2), // red or blue (flase - green)
// iclt_tile[i][chn],
// 2 * transform_size);
tiles_debayered[i][chn] = tile_debayer_shot_corr(
(chn != 2), // red or blue (flase - green)
iclt_tile[i][chn],
2 * transform_size,
lt_window2, // squared lapping window
min_shot, // 10.0; // Do not adjust for shot noise if lower than
scale_shot, //3.0; // scale when dividing by sqrt
lt_window2); // re-apply window to the result
}
}
if ((globalDebugLevel > 0) && debugTile) {
showDoubleFloatArrays sdfa_instance = new showDoubleFloatArrays(); // just for debugging?
String [] titles = {"red0","blue0","green0","red1","blue1","green1","red2","blue2","green2","red3","blue3","green3"};
double [][] dbg_tile = new double [quad*numcol][];
for (int i = 0; i < quad; i++) {
for (int chn = 0; chn <numcol; chn++) { // color
dbg_tile[i * numcol + chn] = tiles_debayered[i][chn];
}
}
sdfa_instance.showArrays(dbg_tile, 2* transform_size, 2* transform_size, true, "tiles_debayered_x"+tileX+"_y"+tileY, titles);
}
texture_tiles[tileY][tileX] = tile_combine_rgba(
tiles_debayered, // iclt_tile, // [port][numcol][256]
lt_window2, // [256]
port_offsets, // [port]{x_off, y_off}
img_mask, // which port to use, 0xf - all 4 (will modify as local variable)
diff_sigma, // pixel value/pixel change
diff_threshold, // pixel value/pixel change
diff_gauss, // when averaging images, use gaussian around average as weight (false - sharp all/nothing)
min_agree, // minimal number of channels to agree on a point (real number to work with fuzzy averages)
col_weights, // color channel weights, sum == 1.0
keep_weights, // keep_weights); // return channel weights after A in RGBA
(globalDebugLevel > 0) && debugTile);
// mix RGB from iclt_tile, mix alpha with - what? correlation strength or 'don't care'? good correlation or all > min?
for (int i = 0; i < iclt_tile[0][0].length; i++ ) {
double sw = 0.0;
for (int ip = 0; ip < quad; ip++) {
sw += texture_tiles[tileY][tileX][numcol+1+ip][i];
}
if (sw != 0 ) sw = 1.0/sw;
for (int chn = 0; chn <numcol; chn++) { // color
texture_tiles[tileY][tileX][chn][i] = 0.0; //iclt[tileY][tileX][chn]
for (int ip = 0; ip < quad; ip++) {
texture_tiles[tileY][tileX][chn][i] += sw * texture_tiles[tileY][tileX][numcol+1+ip][i] * iclt_tile[ip][chn][i];
}
}
}
}
}
};
......@@ -1366,6 +1545,317 @@ public class ImageDtt {
startAndJoin(threads);
return clt_data;
}
public double [][] tile_combine_rgba(
double [][][] iclt_tile, // [port][numcol][256]
double [] lt_window, // [256]
double [][] port_offsets, // [port]{x_off, y_off}
int port_mask, // which port to use, 0xf - all 4 (will modify as local variable)
double diff_sigma, // pixel value/pixel change
double diff_threshold, // pixel value/pixel change
boolean diff_gauss, // when averaging images, use gaussian around average as weight (false - sharp all/nothing)
double min_agree, // minimal number of channels to agree on a point (real number to work with fuzzy averages)
double [] chn_weights, // color channel weights, sum == 1.0
boolean keep_weights, // return channel weights after A in RGBA
boolean debug)
{
int ports = iclt_tile.length;
int numcol = iclt_tile[0].length;
int tile_len = iclt_tile[0][0].length;
int usedPorts = ((port_mask >> 0) & 1) + ((port_mask >> 1) & 1) + ((port_mask >> 2) & 1) + ((port_mask >> 3) & 1);
double [][] port_weights = new double[ports][tile_len];
double [][] color_avg = new double[numcol][tile_len];
double [][] rgba = new double[numcol + 1 + (keep_weights?(ports + 4):0)][];
int rms_start = numcol + 1 + ports;
if (keep_weights){
for (int chn = 0; chn <= numcol ; chn++){
rgba[rms_start + chn] = new double [tile_len]; // rms for each color, then - weighted
}
for (int i = 0; i <tile_len; i++){
double sw = 0.0;
for (int chn = 0; chn < numcol; chn ++ ){
double s0 = 0, s1 = 0, s2 = 0;
for (int ip = 0; ip < ports; ip++)if ((port_mask & ( 1 << ip)) != 0){
s0 += 1;
s1 += iclt_tile[ip][chn][i];
s2 += iclt_tile[ip][chn][i]*iclt_tile[ip][chn][i];
}
rgba[rms_start+chn][i] = Math.sqrt(s0*s2 - s1*s1) / s0;
sw += chn_weights[chn]*rgba[rms_start+chn][i]*rgba[rms_start+chn][i];
}
rgba[rms_start+numcol][i] = Math.sqrt(sw); // will fade as window
}
}
double [] alpha = new double[tile_len];
double threshold2 = diff_sigma * diff_threshold;
threshold2 *= threshold2; // squared to compare with diff^2
if (usedPorts > 1) {
double [] pair_dist2r = new double [ports*(ports-1)/2]; // reversed squared distance between images - to be used with gaussian
int [][] pair_ports = new int [ports*(ports-1)/2][2];
int indx = 0;
double ksigma = 1.0/(2.0*diff_sigma*diff_sigma); // multiply by a weighted sum of squares of the differences
for (int i = 0; i < ports; i++) if ((port_mask & ( 1 << i)) != 0){
for (int j = i+1; j < ports; j++) if ((port_mask & ( 1 << j)) != 0){
double dx = port_offsets[j][0] - port_offsets[i][0];
double dy = port_offsets[j][1] - port_offsets[i][1];
pair_ports[indx][0] = i;
pair_ports[indx][1] = j;
pair_dist2r[indx++] = ksigma/(dx*dx+dy*dy); // 2*sigma^2 * r^2
}
}
// there will be no pairs for a single used port
for (int i = 0; i < tile_len; i++){
for (int ip = 0; ip < ports; ip++) port_weights[ip][i] = 0.0;
for (int ip = 0; ip<pair_ports.length; ip++){
double d = 0;
for (int chn = 0; chn < numcol; chn++){
double dc = iclt_tile[pair_ports[ip][0]][chn][i] - iclt_tile[pair_ports[ip][1]][chn][i];
dc /= lt_window[i]; // to compensate fading near the edges
d+= chn_weights[chn]*dc*dc;
}
d = Math.exp(-pair_dist2r[ip]*d); // 0.5 for exact match, lower for mismatch. Add this weight to both ports involved
// Add weight to both channels in a pair
port_weights[pair_ports[ip][0]][i] +=d;
port_weights[pair_ports[ip][1]][i] +=d;
}
// find 2 best ports (resolving 2 pairs of close values)
int bestPort1=0;
for (int ip = bestPort1+1; ip < ports; ip++) if (port_weights[ip][i] > port_weights[bestPort1][i]) bestPort1 = ip;
int bestPort2 = (bestPort1 == 0)?1:0;
for (int ip = bestPort2+1; ip < ports; ip++) if ((ip != bestPort1) && (port_weights[ip][i] > port_weights[bestPort2][i])) bestPort2 = ip;
// find weighted average between these 2 ports
double w1 = port_weights[bestPort1][i]/(port_weights[bestPort1][i]+port_weights[bestPort2][i]);
double w2 = 1.0 - w1;
for (int chn = 0; chn < numcol; chn++){
color_avg[chn][i] = w1 * iclt_tile[bestPort1][chn][i] + w2 * iclt_tile[bestPort2][chn][i];
}
// recalculate all weights using difference from this average of the best pair
double [] d2 = new double [ports]; //weighted squared differences
for (int ip = 0; ip < ports; ip++) if ((port_mask & ( 1 << ip)) != 0){
d2[ip] = 0;
for (int chn = 0; chn < numcol; chn++){
double dc = iclt_tile[ip][chn][i] - color_avg[chn][i];
dc /= lt_window[i]; // to compensate fading near the edges
d2[ip]+= chn_weights[chn]*dc*dc;
}
port_weights[ip][i] = Math.exp(-ksigma * d2[ip]);
}
// and now make a new average with those weights
double k = 0.0;
for (int ip = 0; ip < ports; ip++) k+=port_weights[ip][i];
k = 1.0/k;
for (int chn = 0; chn < numcol; chn++){
color_avg[chn][i] = 0;
for (int ip = 0; ip < ports; ip++) {
color_avg[chn][i] += k* port_weights[ip][i] * iclt_tile[ip][chn][i];
}
}
/*
// remove outlayers (if any) here completely? diff_threshold
// threshold2
while ((d2[0] > threshold2) || (d2[1] > threshold2) || (d2[2] > threshold2) || (d2[3] > threshold2)){ // assuming ports==4!
// find worst channel
int iworst = 0;
for (int ip = 1; ip < ports; ip++) if (d2[ip] > d2[iworst]) iworst = ip;
port_mask &= ~ (1 << iworst); // remove worst port
port_weights[iworst][i] = 0.0;
// recalculate new average after worst is removed
k = 0.0;
for (int ip = 0; ip < ports; ip++) k+=port_weights[ip][i];
k = 1.0/k;
for (int chn = 0; chn < numcol; chn++){
color_avg[chn][i] = 0;
for (int ip = 0; ip < ports; ip++) {
color_avg[chn][i] += k* port_weights[ip][i] * iclt_tile[ip][chn][i];
}
}
// recalculate all weights using difference from this average of the best pair
d2 = new double [ports]; //weighted squared differences
for (int ip = 0; ip < ports; ip++) if ((port_mask & ( 1 << ip)) != 0){
d2[ip] = 0;
for (int chn = 0; chn < numcol; chn++){
double dc = iclt_tile[ip][chn][i] - color_avg[chn][i];
dc /= lt_window[i]; // to compensate fading near the edges
d2[ip]+= chn_weights[chn]*dc*dc;
}
port_weights[ip][i] = Math.exp(-ksigma * d2[ip]);
}
}
// one last time re-average? (weights will change)
k = 0.0;
for (int ip = 0; ip < ports; ip++) k+=port_weights[ip][i];
k = 1.0/k;
for (int chn = 0; chn < numcol; chn++){
color_avg[chn][i] = 0;
for (int ip = 0; ip < ports; ip++) {
color_avg[chn][i] += k* port_weights[ip][i] * iclt_tile[ip][chn][i];
}
}
*/
} // or (int i = 0; i < tile_len; i++){
} else if (usedPorts > 0){ // just copy from a single channel
for (int ip = 0; ip < ports; ip++) if ((port_mask & ( 1 << ip)) != 0){
for (int i = 0; i < tile_len; i++){
for (int chn = 0; chn < numcol; chn++){
color_avg[chn][i] = iclt_tile[ip][chn][i];
}
port_weights[ip][i] = 1.0; // lt_window[i]; // or use 1.0?
}
}
}
// calculate alpha from channel weights. Start with just a sum of weights?
for (int i = 0; i < tile_len; i++){
alpha[i] = 0.0;
for (int ip = 0; ip < ports; ip++) if ((port_mask & ( 1 << ip)) != 0){
alpha[i]+= port_weights[ip][i];
}
alpha[i] *= lt_window[i]; // make it configurable?
}
for (int i = 0; i < numcol; i++) rgba[i] = color_avg[i];
rgba[numcol] = alpha;
for (int i = 0; i < ports; i++) rgba[numcol + 1 + i] = port_weights[i];
return rgba;
}
public double [] tile_debayer_shot_corr(
boolean rb,
double [] tile,
int tile_size,
double [] window2, // squared lapping window
double min_shot, // 10.0; // Do not adjust for shot noise if lower than
double scale_shot, //3.0; // scale when dividing by sqrt
double [] window_back) // re-apply window to the result
{
double [] tile_nw = new double [tile.length];
for (int i = 0; i < tile.length; i++) tile_nw[i] = tile[i]/window2[i]; //unapply squared window
double [] tile_db = tile_debayer(
rb,
tile_nw,
tile_size);
if (scale_shot > 0){
double k = 1.0/Math.sqrt(min_shot);
for (int i = 0; i < tile.length; i++) tile_db[i] = scale_shot* ((tile_db[i] > min_shot)? Math.sqrt(tile_db[i]) : (k*tile_db[i]));
}
if (window_back != null) {
for (int i = 0; i < tile.length; i++) tile_db[i] = tile_db[i] * window_back[i]; // optionally re-apply window (may bve a different one)
}
return tile_db;
}
public double [] tile_debayer(
boolean rb,
double [] tile,
int tile_size)
{
int [] neib_indices = {-tile_size - 1, -tile_size, -tile_size + 1, -1, 0, 1, tile_size - 1, tile_size, tile_size + 1};
int tsm1 = tile_size - 1;
double [] rslt = new double [tile_size*tile_size]; // assuming cleared to 0.0;
double [] kern = rb ? kern_rb : kern_g;
double k_corn = rb? (16.0/9.0):(4.0/3.0);
double k_side = rb? (4.0/3.0):(8.0/7.0);
// top left
int indx = 0;
int side_type = 0;
for (int dri = 0; dri < corn_side_indices[side_type].length; dri++) {
int dr = corn_side_indices[side_type][dri]; // middle left
rslt[indx]+=tile[indx+neib_indices[dr]]*kern[dr];
}
rslt[indx] *= k_corn;
// top middle
side_type = 1;
for (int j = 1; j < tsm1; j++){
indx = j;
for (int dri = 0; dri < corn_side_indices[side_type].length; dri++) {
int dr = corn_side_indices[side_type][dri]; // middle left
rslt[indx]+=tile[indx + neib_indices[dr]] * kern[dr];
}
rslt[indx] *= k_side;
}
// top right
indx = tsm1;
side_type = 2;
for (int dri = 0; dri < corn_side_indices[side_type].length; dri++) {
int dr = corn_side_indices[side_type][dri]; // middle left
rslt[indx]+=tile[indx+neib_indices[dr]]*kern[dr];
}
rslt[indx] *= k_corn;
// middle left
side_type = 3;
for (int i = 1; i < tsm1; i++){
indx = i* tile_size;
for (int dri = 0; dri < corn_side_indices[side_type].length; dri++) {
int dr = corn_side_indices[side_type][dri]; // middle left
rslt[indx]+=tile[indx + neib_indices[dr]] * kern[dr];
}
rslt[indx] *= k_side;
}
// middle middle
side_type = 4;
for (int i = 1; i < tsm1; i++){
for (int j = 1; j < tsm1; j++){
indx = i*tile_size+j;
rslt[indx] = 0.0;
for (int dr = 0; dr < neib_indices.length; dr++){
rslt[indx]+=tile[indx+neib_indices[dr]]*kern[dr];
}
}
}
// middle right
side_type = 5;
for (int i = 1; i < tsm1; i++){
indx = i* tile_size + tsm1;
for (int dri = 0; dri < corn_side_indices[side_type].length; dri++) {
int dr = corn_side_indices[side_type][dri]; // middle left
rslt[indx]+=tile[indx + neib_indices[dr]] * kern[dr];
}
rslt[indx] *= k_side;
}
// bottom left
indx = tsm1*tile_size;
side_type = 6;
for (int dri = 0; dri < corn_side_indices[side_type].length; dri++) {
int dr = corn_side_indices[side_type][dri]; // middle left
rslt[indx]+=tile[indx+neib_indices[dr]]*kern[dr];
}
rslt[indx] *= k_corn;
// bottom middle
side_type = 7;
// tsm1*tile_size;
for (int j = 1; j < tsm1; j++){
indx++;
for (int dri = 0; dri < corn_side_indices[side_type].length; dri++) {
int dr = corn_side_indices[side_type][dri]; // middle left
rslt[indx]+=tile[indx + neib_indices[dr]] * kern[dr];
}
rslt[indx] *= k_side;
}
// bottom right
indx++; // = tile_size*tile_size-1;
side_type = 8;
for (int dri = 0; dri < corn_side_indices[side_type].length; dri++) {
int dr = corn_side_indices[side_type][dri]; // middle left
rslt[indx]+=tile[indx+neib_indices[dr]]*kern[dr];
}
rslt[indx] *= k_corn;
return rslt;
}
//final int [] neib_indices = {-width-1,-width,-width+1,-1,0,1,width-1,width,width+1};
// return weights for positive x,y, [(radius+a)*(radius+1)]
public double [] setMaxXYWeights(
double sigma,
......@@ -1375,7 +1865,7 @@ public class ImageDtt {
int indx = 0;
for (int i = 0; i <= radius; i ++){
for (int j = 0; j <= radius; j ++){
weights[indx++] = Math.exp(-(i*i+j*j)/(2*sigma));
weights[indx++] = Math.exp(-(i*i+j*j)/(2*sigma*sigma));
}
}
return weights;
......@@ -1729,7 +2219,19 @@ public class ImageDtt {
System.out.println("iclt_2d():overlap= "+overlap);
System.out.println("iclt_2d():sharp_alpha= "+sharp_alpha);
}
final double [][] dpixels = new double["RGBA".length()][width*height]; // assuming java initializes them to 0
boolean has_weights = false;
boolean set_has_weight = false;
for (int i = 0; (i < tilesY) && !set_has_weight; i++){
for (int j = 0; (j < tilesX) && !set_has_weight; j++){
if (texture_tiles[i][j] != null) {
set_has_weight = true;
has_weights = texture_tiles[i][j].length > 4;
}
}
}
// final double [][] dpixels = new double["RGBA".length()+(has_weights? 4: 0)][width*height]; // assuming java initializes them to 0
final double [][] dpixels = new double["RGBA".length()+(has_weights? 8: 0)][width*height]; // assuming java initializes them to 0
final Thread[] threads = newThreadArray(threadsMax);
final AtomicInteger ai = new AtomicInteger(0);
final AtomicInteger nser = new AtomicInteger(0);
......@@ -1766,7 +2268,7 @@ public class ImageDtt {
for (int i = 0; i < n2;i++){
int start_line = ((tileY*transform_size + i) * tilesX + tileX)*transform_size - offset;
for (int chn = 0; chn < texture_tile.length; chn++) {
if ((chn < 3) || !sharp_alpha) {
if ((chn != 3) || !sharp_alpha) {
for (int j = 0; j<n2;j++) {
dpixels[chn][start_line + j] += texture_tile[chn][n2 * i + j];
}
......@@ -1784,7 +2286,7 @@ public class ImageDtt {
((tileY == lastY) && (i < (n2 - n_half)))) {
int start_line = ((tileY*transform_size + i) * tilesX + tileX)*transform_size - offset;
for (int chn = 0; chn < texture_tile.length; chn++) {
if ((chn < 3) || !sharp_alpha) {
if ((chn != 3) || !sharp_alpha) {
for (int j = 0; j<n2;j++) {
if ( ((tileX > 0) && (tileX < lastX)) ||
((tileX == 0) && (j >= n_half)) ||
......@@ -2252,8 +2754,6 @@ public class ImageDtt {
final int tilesY=corr_data.length;
final int tilesX=corr_data[0].length;
final int nTiles=tilesX*tilesY;
// final int corr_size = (int) Math.round(Math.sqrt(corr_data[0][0].length));
final int tile_size = corr_size+1;
final int corr_len = corr_size*corr_size;
......@@ -2272,16 +2772,9 @@ public class ImageDtt {
for (int nTile = ai.getAndIncrement(); nTile < nTiles; nTile = ai.getAndIncrement()) {
tileY = nTile/tilesX;
tileX = nTile - tileY * tilesX;
if ((corr_data[tileY][tileX] != null) && (corr_data[tileY][tileX].length > 0)) {
if (corr_data[tileY][tileX] != null) {
for (int i = 0; i < corr_size;i++){
try {
System.arraycopy(corr_data[tileY][tileX], corr_size* i, corr_data_out, ((tileY*tile_size + i) *tilesX + tileX)*tile_size , corr_size);
} catch (Exception e){
System.out.println("corr_data[tileY][tileX].length = "+corr_data[tileY][tileX].length+
" corr_size = "+corr_size+" i ="+i+" tile_size="+tile_size);
continue;
}
corr_data_out[((tileY*tile_size + i) *tilesX + tileX)*tile_size+corr_size] = border_contrast*((i & 1) - 0.5);
}
for (int i = 0; i < tile_size; i++){
......@@ -2312,10 +2805,6 @@ public class ImageDtt {
final int tilesY=corr_data.length;
final int tilesX=corr_data[0].length;
final int nTiles=tilesX*tilesY;
// final int corr_size = (int) Math.round(Math.sqrt(corr_data[0][0][0][0].length));
// final int pairs= corr_data[0][0].length;
// final int colors=corr_data[0][0][0].length;
final int tile_size = corr_size+1;
final int corr_len = corr_size*corr_size;
......@@ -2340,7 +2829,7 @@ public class ImageDtt {
for (int nTile = ai.getAndIncrement(); nTile < nTiles; nTile = ai.getAndIncrement()) {
tileY = nTile/tilesX;
tileX = nTile - tileY * tilesX;
if ((corr_data[tileY][tileX] != null) && (corr_data[tileY][tileX].length > 0)) {
if (corr_data[tileY][tileX] != null) {
for (int pair = 0; pair< pairs; pair++) {
for (int nColor = 0; nColor < colors; nColor++) {
int indx = pair*colors+nColor;
......
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