Skip to content

I
imagej-elphel
Commit f04bcc82 authored by Andrey Filippov's avatar Andrey Filippov
Browse files
Options

working on GPU code

parent cd9b6096
Hide whitespace changes
Inline Side-by-side
Showing with 913 additions and 17 deletions
.gitignore
View file @ f04bcc82
...@@ -6,4 +6,5 @@ ...@@ -6,4 +6,5 @@
NC393I NC393I
attic attic
*.log *.log
FOCUS-PSF* FOCUS-PSF*
\ No newline at end of file src/main/resources/trained_model
\ No newline at end of file
src/main/java/GPUTileProcessor.java
View file @ f04bcc82
...@@ -117,7 +117,7 @@ public class GPUTileProcessor { ...@@ -117,7 +117,7 @@ public class GPUTileProcessor {
copyH2D.WidthInBytes = width_in_bytes; copyH2D.WidthInBytes = width_in_bytes;
copyH2D.Height = height; // /4; copyH2D.Height = height; // /4;
// for copying results back to host // for copying results to host
CUDA_MEMCPY2D copyD2H = new CUDA_MEMCPY2D(); CUDA_MEMCPY2D copyD2H = new CUDA_MEMCPY2D();
copyD2H.srcMemoryType = CUmemorytype.CU_MEMORYTYPE_DEVICE; copyD2H.srcMemoryType = CUmemorytype.CU_MEMORYTYPE_DEVICE;
copyD2H.srcDevice = dst_dpointer; // ((test & 1) ==0) ? src_dpointer : dst_dpointer; // copy same data copyD2H.srcDevice = dst_dpointer; // ((test & 1) ==0) ? src_dpointer : dst_dpointer; // copy same data
...@@ -130,8 +130,7 @@ public class GPUTileProcessor { ...@@ -130,8 +130,7 @@ public class GPUTileProcessor {
copyD2H.WidthInBytes = width_in_bytes; copyD2H.WidthInBytes = width_in_bytes;
copyD2H.Height = height; // /2; copyD2H.Height = height; // /2;
// Set up the kernel parameters: A pointer to an array // kernel parameters: pointer to pointers
// of pointers which point to the actual values.
Pointer kernelParameters = Pointer.to( Pointer kernelParameters = Pointer.to(
Pointer.to(dst_dpointer), Pointer.to(dst_dpointer),
Pointer.to(src_dpointer), Pointer.to(src_dpointer),
...@@ -152,16 +151,16 @@ public class GPUTileProcessor { ...@@ -152,16 +151,16 @@ public class GPUTileProcessor {
0, null, // Shared memory size and stream (shared - only dynamic, static is in code) 0, null, // Shared memory size and stream (shared - only dynamic, static is in code)
kernelParameters, null); // Kernel- and extra parameters kernelParameters, null); // Kernel- and extra parameters
// Copy the data from the device back to the host // Copy the data from the device to the host
cuMemcpy2D(copyD2H); cuMemcpy2D(copyD2H);
// clean up // clean up
cuMemFree(src_dpointer); cuMemFree(src_dpointer);
cuMemFree(dst_dpointer); cuMemFree(dst_dpointer);
} }
public int setup() throws IOException // String arg, ImagePlus imagePlus) public int setup() throws IOException
{ {
// From code by Marco Hutter - http://www.jcuda.org
// Enable exceptions and omit all subsequent error checks // Enable exceptions and omit all subsequent error checks
JCudaDriver.setExceptionsEnabled(true); JCudaDriver.setExceptionsEnabled(true);
JNvrtc.setExceptionsEnabled(true); JNvrtc.setExceptionsEnabled(true);
...@@ -174,7 +173,9 @@ public class GPUTileProcessor { ...@@ -174,7 +173,9 @@ public class GPUTileProcessor {
cuCtxCreate(context, 0, device); cuCtxCreate(context, 0, device);
// Obtain the CUDA source code from the CUDA file // Obtain the CUDA source code from the CUDA file
// Get absolute path to the file in resource foldder, then read it as a normal file.
// When using just Eclipse resources - it does not notice that the file
// was edited (happens frequently during kernel development).
ClassLoader classLoader = getClass().getClassLoader(); ClassLoader classLoader = getClass().getClassLoader();
File file = new File(classLoader.getResource(GPU_KERNEL_FILE).getFile()); File file = new File(classLoader.getResource(GPU_KERNEL_FILE).getFile());
System.out.println(file.getAbsolutePath()); System.out.println(file.getAbsolutePath());
...@@ -196,9 +197,7 @@ public class GPUTileProcessor { ...@@ -196,9 +197,7 @@ public class GPUTileProcessor {
} }
/** /**
* Create the CUDA function object for the kernel function with the * Create the kernel function by its name in the source code
* given name that is contained in the given source code
*
* @param sourceCode The source code * @param sourceCode The source code
* @param kernelName The kernel function name * @param kernelName The kernel function name
* @return * @return
......
src/main/java/ImageDtt.java
View file @ f04bcc82
...@@ -4696,6 +4696,7 @@ public class ImageDtt { ...@@ -4696,6 +4696,7 @@ public class ImageDtt {
else if (ktileY >= clt_kernels[chn].length) ktileY = clt_kernels[chn].length-1; else if (ktileY >= clt_kernels[chn].length) ktileY = clt_kernels[chn].length-1;
if (ktileX < 0) ktileX = 0; if (ktileX < 0) ktileX = 0;
else if (ktileX >= clt_kernels[chn][ktileY].length) ktileX = clt_kernels[chn][ktileY].length-1; else if (ktileX >= clt_kernels[chn][ktileY].length) ktileX = clt_kernels[chn][ktileY].length-1;
// extract center offset data stored with each kernel tile
CltExtra ce = new CltExtra (clt_kernels[chn][ktileY][ktileX][4]); CltExtra ce = new CltExtra (clt_kernels[chn][ktileY][ktileX][4]);
// 2. calculate correction for center of the kernel offset // 2. calculate correction for center of the kernel offset
double kdx = centerX - (ktileX -1 +0.5) * kernel_step; // difference in pixel double kdx = centerX - (ktileX -1 +0.5) * kernel_step; // difference in pixel
...@@ -5005,6 +5006,7 @@ public class ImageDtt { ...@@ -5005,6 +5006,7 @@ public class ImageDtt {
clt_tile[dct_mode] = dtt.fold_tile (tile_in, transform_size, dct_mode); // DCCT, DSCT, DCST, DSST clt_tile[dct_mode] = dtt.fold_tile (tile_in, transform_size, dct_mode); // DCCT, DSCT, DCST, DSST
} }
// clt_tile[dct_mode] = dtt.dttt_iv (clt_tile[dct_mode], dct_mode, transform_size); // clt_tile[dct_mode] = dtt.dttt_iv (clt_tile[dct_mode], dct_mode, transform_size);
/*
if (bdebug) { if (bdebug) {
double [] clt_tile_dbg = clt_tile[dct_mode].clone(); double [] clt_tile_dbg = clt_tile[dct_mode].clone();
double [] clt_tile_dbg1 = clt_tile[dct_mode].clone(); double [] clt_tile_dbg1 = clt_tile[dct_mode].clone();
...@@ -5019,8 +5021,11 @@ public class ImageDtt { ...@@ -5019,8 +5021,11 @@ public class ImageDtt {
} }
System.out.println("done debug"); System.out.println("done debug");
} else { } else {
*/
clt_tile[dct_mode] = dtt.dttt_iv (clt_tile[dct_mode], dct_mode, transform_size); clt_tile[dct_mode] = dtt.dttt_iv (clt_tile[dct_mode], dct_mode, transform_size);
/*
} }
*/
} }
...@@ -7593,6 +7598,787 @@ public class ImageDtt { ...@@ -7593,6 +7598,787 @@ public class ImageDtt {
// public double [][][][][][] clt_bi_quad( // public double [][][][][][] clt_bi_quad(
public double [][][][][][][] clt_bi_quad_dbg(
final EyesisCorrectionParameters.CLTParameters clt_parameters,
final double fatzero, // May use correlation fat zero from 2 different parameters - fat_zero and rig.ml_fatzero
final boolean notch_mode, // use notch filter for inter-camera correlation to detect poles
final int lt_rad, // low texture mode - inter-correlation is averaged between the neighbors before argmax-ing, using (2*notch_mode+1)^2 square
final boolean no_int_x0, // do not offset window to integer maximum - used when averaging low textures to avoid "jumps" for very wide
final int [][] tile_op, // [tilesY][tilesX] - what to do - 0 - nothing for this tile
final double [][] disparity_array, // [tilesY][tilesX] - individual per-tile expected disparity
final double [][][] image_data_main, // first index - number of image in a quad
final double [][][] image_data_aux, // first index - number of image in a quad
final boolean [][] saturation_main, // (near) saturated pixels or null
final boolean [][] saturation_aux, // (near) saturated pixels or null
// correlation results - combo will be for the correlation between two quad cameras
final double [][][][] clt_corr_combo, // [type][tilesY][tilesX][(2*transform_size-1)*(2*transform_size-1)] // if null - will not calculate
// [type][tilesY][tilesX] should be set by caller
// types: 0 - selected correlation (product+offset), 1 - sum
final double [][] disparity_bimap, // [23][tilesY][tilesX], only [6][] is needed on input or null - do not calculate
// last 2 - contrast, avg/ "geometric average)
final double [][] ml_data, // data for ML - 18 layers - 4 center areas (3x3, 5x5,..) per camera-per direction, 1 - composite, and 1 with just 1 data (target disparity)
final double [][][][] texture_tiles_main, // [tilesY][tilesX]["RGBA".length()][]; null - will skip images combining
final double [][][][] texture_tiles_aux, // [tilesY][tilesX]["RGBA".length()][]; null - will skip images combining
final int width, // may be not multiple of 8, same for the height
final GeometryCorrection geometryCorrection_main,
final GeometryCorrection geometryCorrection_aux,
final double [][][][][][] clt_kernels_main, // [channel_in_quad][color][tileY][tileX][band][pixel] , size should match image (have 1 tile around)
final double [][][][][][] clt_kernels_aux, // [channel_in_quad][color][tileY][tileX][band][pixel] , size should match image (have 1 tile around)
final double corr_magic_scale, // still not understood coefficient that reduces reported disparity value. Seems to be around 0.85
final boolean keep_clt_data,
// final int [][] woi_tops,
final double [][][] ers_delay, // if not null - fill with tile center acquisition delay
final int threadsMax, // maximal number of threads to launch
final int debugLevel,
final double [][][] port_xy_main_dbg, // for each tile/port save x,y pixel coordinates (gpu code development)
final double [][][] port_xy_aux_dbg) // for each tile/port save x,y pixel coordinates (gpu code development)
{
final int globalDebugLevel = clt_parameters.rig.rig_mode_debug?debugLevel:-2;
final int debug_tileX = clt_parameters.tileX;
final int debug_tileY = clt_parameters.tileY;
final int quad_main = image_data_main.length; // number of subcameras
final int quad_aux = image_data_aux.length; // number of subcameras
final int numcol = 3; // number of colors
final int nChn = image_data_main[0].length;
final int height=image_data_main[0][0].length/width;
final int tilesX=width/clt_parameters.transform_size;
final int tilesY=height/clt_parameters.transform_size;
final int nTilesInChn=tilesX*tilesY;
// clt_data does not need to be for the whole image (no, it is used for textures)
final double [][][][][][][] clt_bidata = (keep_clt_data)? (new double[2][][][][][][]):null;
if (clt_bidata != null) {
clt_bidata[0] = new double[quad_main][nChn][tilesY][tilesX][][];
clt_bidata[1] = new double[quad_aux][nChn][tilesY][tilesX][][];
}
final double [][] lt_corr = (lt_rad > 0)? (new double [nTilesInChn][]):null; // will keep inter-camera combo correlation, later combined in a separate multi-thread run
final Thread[] threads = newThreadArray(threadsMax);
final AtomicInteger ai = new AtomicInteger(0);
final double [] col_weights= new double [numcol]; // colors are RBG
col_weights[2] = 1.0/(1.0 + clt_parameters.corr_red + clt_parameters.corr_blue); // green color
col_weights[0] = clt_parameters.corr_red * col_weights[2];
col_weights[1] = clt_parameters.corr_blue * col_weights[2];
final int corr_size = clt_parameters.transform_size * 2 -1;
final int [][] transpose_indices = new int [corr_size*(corr_size-1)/2][2];
int indx = 0;
for (int i =0; i < corr_size-1; i++){
for (int j = i+1; j < corr_size; j++){
transpose_indices[indx ][0] = i * corr_size + j;
transpose_indices[indx++][1] = j * corr_size + i;
}
}
// get ml_data half width
final int ml_hwidth = (ml_data != null)?(((int) Math.round(Math.sqrt(ml_data[0].length/nTilesInChn)) - 1) / 2):0;
// Create window to select center correlation strip using
// ortho_height - full width of non-zero elements
// ortho_eff_height - effective height (ration of the weighted column sum to the center value)
int wcenter = clt_parameters.transform_size - 1;
final double [] ortho_weights = new double [corr_size]; // [15]
for (int i = 0; i < corr_size; i++){
if ((i >= wcenter - clt_parameters.img_dtt.ortho_height/2) && (i <= wcenter + clt_parameters.img_dtt.ortho_height/2)) {
double dx = 1.0*(i-wcenter)/(clt_parameters.img_dtt.ortho_height/2 + 1);
ortho_weights[i] = 0.5*(1.0+Math.cos(Math.PI*dx))/clt_parameters.img_dtt.ortho_eff_height;
}
}
if (globalDebugLevel > 0){
System.out.println("ortho_height="+ clt_parameters.img_dtt.ortho_height+" ortho_eff_height="+ clt_parameters.img_dtt.ortho_eff_height);
for (int i = 0; i < corr_size; i++){
System.out.println(" ortho_weights["+i+"]="+ ortho_weights[i]);
}
}
if (globalDebugLevel > -2) {
System.out.println("clt_aberrations_quad_corr(): width="+width+" height="+height+" transform_size="+clt_parameters.transform_size+
" debug_tileX="+debug_tileX+" debug_tileY="+debug_tileY+" globalDebugLevel="+globalDebugLevel);
}
final int [][] corr_pairs ={ // {first, second, rot} rot: 0 - as is, 1 - swap y,x
{0,1,0},
{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 = clt_parameters.transform_size * clt_parameters.transform_size;
final double [] filter_direct= new double[transform_len];
if (clt_parameters.corr_sigma == 0) {
filter_direct[0] = 1.0;
for (int i= 1; i<filter_direct.length;i++) filter_direct[i] =0;
} else {
for (int i = 0; i < clt_parameters.transform_size; i++){
for (int j = 0; j < clt_parameters.transform_size; j++){
filter_direct[i * clt_parameters.transform_size+j] = Math.exp(-(i*i+j*j)/(2*clt_parameters.corr_sigma)); // FIXME: should be sigma*sigma !
}
}
}
// normalize
double sum = 0;
for (int i = 0; i < clt_parameters.transform_size; i++){
for (int j = 0; j < clt_parameters.transform_size; j++){
double d = filter_direct[i*clt_parameters.transform_size+j];
d*=Math.cos(Math.PI*i/(2*clt_parameters.transform_size))*Math.cos(Math.PI*j/(2*clt_parameters.transform_size));
if (i > 0) d*= 2.0;
if (j > 0) d*= 2.0;
sum +=d;
}
}
for (int i = 0; i<filter_direct.length; i++){
filter_direct[i] /= sum;
}
DttRad2 dtt = new DttRad2(clt_parameters.transform_size);
final double [] filter= dtt.dttt_iiie(filter_direct);
for (int i=0; i < filter.length;i++) filter[i] *= 2*clt_parameters.transform_size;
// prepare disparity maps and weights
final int max_search_radius = (int) Math.abs(clt_parameters.max_corr_radius); // use negative max_corr_radius for squares instead of circles?
final int max_search_radius_poly = 1;
if (globalDebugLevel > 0){
System.out.println("max_corr_radius= "+clt_parameters.max_corr_radius);
System.out.println("max_search_radius= "+max_search_radius);
System.out.println("max_search_radius_poly="+max_search_radius_poly);
System.out.println("corr_fat_zero= "+fatzero);
System.out.println("disparity_array[0][0]= "+disparity_array[0][0]);
}
// add optional initialization of debug layers here
if (disparity_bimap != null){
for (int i = 0; i < disparity_bimap.length;i++){
disparity_bimap[i] = new double [tilesY*tilesX];
}
}
if (ers_delay != null) {
ers_delay[0] = new double [quad_main][];
for (int i = 0; i < quad_main; i++) ers_delay[0][i] = new double [tilesX*tilesY];
ers_delay[1] = new double [quad_aux][];
for (int i = 0; i < quad_aux; i++) ers_delay[1][i] = new double [tilesX*tilesY];
}
dtt.set_window(clt_parameters.clt_window);
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*clt_parameters.transform_size, 2*clt_parameters.transform_size, "lt_window");
}
final Matrix [] corr_rots_main = geometryCorrection_main.getCorrVector().getRotMatrices(); // get array of per-sensor rotation matrices
final Matrix rigMatrix = geometryCorrection_aux.getRotMatrix(true);
final Matrix [] corr_rots_aux = geometryCorrection_aux.getCorrVector().getRotMatrices(rigMatrix); // get array of per-sensor rotation matrices
for (int ithread = 0; ithread < threads.length; ithread++) {
threads[ithread] = new Thread() {
@Override
public void run() {
DttRad2 dtt = new DttRad2(clt_parameters.transform_size);
dtt.set_window(clt_parameters.clt_window);
int tileY,tileX,tIndex; // , chn;
// showDoubleFloatArrays sdfa_instance = new showDoubleFloatArrays(); // just for debugging?
double centerX; // center of aberration-corrected (common model) tile, X
double centerY; //
double [][] fract_shiftsXY_main = new double[quad_main][];
double [][] fract_shiftsXY_aux = new double[quad_aux][];
double [][] tcorr_combo = null; // [15*15] pixel space
double [][][][] clt_data_main = new double[quad_main][nChn][][];
double [][][][] clt_data_aux = new double[quad_aux][nChn][][];
double [][] ml_data_main = (ml_data != null)? new double [ML_TOP_AUX_INDEX][(2*ml_hwidth +1)*(2*ml_hwidth +1)]:null;
double [][] ml_data_aux = (ml_data != null)? new double [ML_TOP_AUX_INDEX][(2*ml_hwidth +1)*(2*ml_hwidth +1)]:null;
double [] ml_data_inter = (ml_data != null)? new double [(2*ml_hwidth +1)*(2*ml_hwidth +1)]:null;
// double [] ml_data_other = (ml_data != null)? new double [(2*ml_hwidth +1)*(2*ml_hwidth +1)]:null;
double [] ml_data_dbg1 = (ml_data != null)? new double [(2*ml_hwidth +1)*(2*ml_hwidth +1)]:null;
Correlation2d corr2d = new Correlation2d(
clt_parameters.img_dtt, // ImageDttParameters imgdtt_params,
clt_parameters.transform_size, // int transform_size,
2.0, // double wndx_scale, // (wndy scale is always 1.0)
(globalDebugLevel > -1)); // boolean debug)
for (int nTile = ai.getAndIncrement(); nTile < nTilesInChn; nTile = ai.getAndIncrement()) {
tileY = nTile /tilesX;
tileX = nTile % tilesX;
tIndex = tileY * tilesX + tileX;
if (tile_op[tileY][tileX] == 0) {
if (disparity_bimap != null){
disparity_bimap[BI_TARGET_INDEX][tIndex] = Double.NaN;
}
continue; // nothing to do for this tile
}
int img_mask = getImgMask(tile_op[tileY][tileX]); // which images to use
int corr_mask = getPairMask(tile_op[tileY][tileX]); // which pairs to combine in the combo: 1 - top, 2 bottom, 4 - left, 8 - right
// mask out pairs that use missing channels
// Is it currently used with diagonals?
// TODO: use masks from tile task
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);
final int [] overexp_main = (saturation_main != null) ? ( new int [2]): null;
final int [] overexp_aux = (saturation_aux != null) ? ( new int [2]): null;
// Moved from inside chn loop
centerX = tileX * clt_parameters.transform_size + clt_parameters.transform_size/2; // - shiftX;
centerY = tileY * clt_parameters.transform_size + clt_parameters.transform_size/2; // - shiftY;
// TODO: move port coordinates out of color channel loop
double [][] centersXY_main;
double [][] centersXY_aux;
double disparity_main = disparity_array[tileY][tileX];
double disparity_aux = disparity_main * geometryCorrection_aux.getDisparityRadius()/geometryCorrection_main.getDisparityRadius();
if (disparity_bimap != null){
disparity_bimap[BI_TARGET_INDEX][tIndex] = disparity_main;
}
centersXY_main = geometryCorrection_main.getPortsCoordinatesAndDerivatives(
geometryCorrection_main, // GeometryCorrection gc_main,
false, // boolean use_rig_offsets,
corr_rots_main, // Matrix [] rots,
null, // Matrix [][] deriv_rots,
null, // double [][] pXYderiv, // if not null, should be double[8][]
centerX,
centerY,
disparity_main); // + disparity_corr);
centersXY_aux = geometryCorrection_aux.getPortsCoordinatesAndDerivatives(
geometryCorrection_main, // GeometryCorrection gc_main,
true, // boolean use_rig_offsets,
corr_rots_aux, // Matrix [] rots,
null, // Matrix [][] deriv_rots,
null, // double [][] pXYderiv, // if not null, should be double[8][]
centerX,
centerY,
disparity_aux); // + disparity_corr);
// acquisition time of the tiles centers in scanline times
if (ers_delay != null) {
for (int i = 0; i < quad_main; i++) ers_delay[0][i][nTile] = centersXY_main[i][1]-geometryCorrection_main.woi_tops[i];
for (int i = 0; i < quad_aux; i++) ers_delay[1][i][nTile] = centersXY_aux[i][1]- geometryCorrection_aux.woi_tops[i];
}
if ((globalDebugLevel > 0) && (tileX == debug_tileX) && (tileY == debug_tileY)) {
for (int i = 0; i < quad_main; i++) {
System.out.println("clt_aberrations_quad_corr(): tileX="+tileX+", tileY="+tileY+
" centerX="+centerX+" centerY="+centerY+" disparity="+disparity_array[tileY][tileX]+
" centersXY_main["+i+"][0]="+centersXY_main[i][0]+" centersXY_main["+i+"][1]="+centersXY_main[i][1]);
}
for (int i = 0; i < quad_aux; i++) {
System.out.println("clt_aberrations_quad_corr(): tileX="+tileX+", tileY="+tileY+
" centerX="+centerX+" centerY="+centerY+" disparity="+disparity_array[tileY][tileX]+
" centersXY_aux["+i+"][0]="+centersXY_aux[i][0]+" centersXY_aux["+i+"][1]="+centersXY_aux[i][1]);
}
}
if (port_xy_main_dbg != null) {
port_xy_main_dbg[nTile] = centersXY_main;
}
if (port_xy_aux_dbg != null) {
port_xy_aux_dbg [nTile] = centersXY_main;
}
if ((globalDebugLevel > -1) && (tileX == debug_tileX) && (tileY == debug_tileY)) { // before correction
System.out.print(disparity_array[tileY][tileX]+"\t"+
centersXY_main[0][0]+"\t"+centersXY_main[0][1]+"\t"+
centersXY_main[1][0]+"\t"+centersXY_main[1][1]+"\t"+
centersXY_main[2][0]+"\t"+centersXY_main[2][1]+"\t"+
centersXY_main[3][0]+"\t"+centersXY_main[3][1]+"\t");
System.out.print(disparity_array[tileY][tileX]+"\t"+
centersXY_aux[0][0]+"\t"+centersXY_aux[0][1]+"\t"+
centersXY_aux[1][0]+"\t"+centersXY_aux[1][1]+"\t"+
centersXY_aux[2][0]+"\t"+centersXY_aux[2][1]+"\t"+
centersXY_aux[3][0]+"\t"+centersXY_aux[3][1]+"\t");
}
for (int chn = 0; chn <numcol; chn++) {
if ((globalDebugLevel > -1) && (tileX == debug_tileX) && (tileY == debug_tileY) && (chn == 2)) {
System.out.println("\nMain camera, centerX="+centerX+", centerY="+centerY);
System.out.println(disparity_array[tileY][tileX]+"\t"+
centersXY_main[0][0]+"\t"+centersXY_main[0][1]+"\t"+
centersXY_main[1][0]+"\t"+centersXY_main[1][1]+"\t"+
centersXY_main[2][0]+"\t"+centersXY_main[2][1]+"\t"+
centersXY_main[3][0]+"\t"+centersXY_main[3][1]+"\t");
System.out.println("\nAux camera, centerX="+centerX+", centerY="+centerY);
System.out.println(disparity_array[tileY][tileX]+"\t"+
centersXY_aux[0][0]+"\t"+centersXY_aux[0][1]+"\t"+
centersXY_aux[1][0]+"\t"+centersXY_aux[1][1]+"\t"+
centersXY_aux[2][0]+"\t"+centersXY_aux[2][1]+"\t"+
centersXY_aux[3][0]+"\t"+centersXY_aux[3][1]+"\t");
}
for (int i = 0; i < quad_main; i++) {
clt_data_main[i][chn] = new double [4][];
fract_shiftsXY_main[i] = extract_correct_tile( // return a pair of residual offsets
image_data_main[i],
width, // image width
(clt_kernels_main == null) ? null : clt_kernels_main[i], // [color][tileY][tileX][band][pixel]
clt_data_main[i][chn], //double [][] clt_tile, // should be double [4][];
clt_parameters.kernel_step,
clt_parameters.transform_size,
dtt,
chn,
centersXY_main[i][0], // centerX, // center of aberration-corrected (common model) tile, X
centersXY_main[i][1], // centerY, //
(globalDebugLevel > -2) && (tileX == debug_tileX) && (tileY == debug_tileY)? 2:0, // external tile compare
false,// no_deconvolution,
false, // ); // transpose);
((saturation_main != null) ? saturation_main[i] : null), //final boolean [][] saturation_imp, // (near) saturated pixels or null
((saturation_main != null) ? overexp_main: null)); // final double [] overexposed)
}
for (int i = 0; i < quad_aux; i++) {
clt_data_aux[i][chn] = new double [4][];
fract_shiftsXY_aux[i] = extract_correct_tile( // return a pair of residual offsets
image_data_aux[i],
width, // image width
(clt_kernels_aux == null) ? null : clt_kernels_aux[i], // [color][tileY][tileX][band][pixel]
clt_data_aux[i][chn], //double [][] clt_tile, // should be double [4][];
clt_parameters.kernel_step,
clt_parameters.transform_size,
dtt,
chn,
centersXY_aux[i][0], // centerX, // center of aberration-corrected (common model) tile, X
centersXY_aux[i][1], // centerY, //
0, // external tile compare
false,// no_deconvolution,
false, // ); // transpose);
((saturation_aux != null) ? saturation_aux[i] : null), //final boolean [][] saturation_imp, // (near) saturated pixels or null
((saturation_aux != null) ? overexp_aux: null)); // final double [] overexposed)
}
if ((globalDebugLevel > -1) && (tileX == debug_tileX) && (tileY == debug_tileY) && (chn == 2)) {
System.out.println();
}
if ((globalDebugLevel > 0) && (debug_tileX == tileX) && (debug_tileY == tileY) && (chn == 2)) {
showDoubleFloatArrays sdfa_instance = new showDoubleFloatArrays(); // just for debugging?
String [] titles = {"CC0","SC0","CS0","SS0","CC1","SC1","CS1","SS1","CC2","SC2","CS2","SS2","CC3","SC3","CS3","SS3"};
double [][] dbg_tile = new double [16][];
for (int i = 0; i < 16; i++) dbg_tile[i]=clt_data_main[i>>2][chn][i & 3];
sdfa_instance.showArrays(dbg_tile, clt_parameters.transform_size, clt_parameters.transform_size, true, "MAIN_pre-shifted_x"+tileX+"_y"+tileY, titles);
double [][] dbg_tile_aux = new double [16][];
for (int i = 0; i < 16; i++) dbg_tile[i]=clt_data_aux[i>>2][chn][i & 3];
sdfa_instance.showArrays(dbg_tile_aux, clt_parameters.transform_size, clt_parameters.transform_size, true, "AUX_pre-shifted_x"+tileX+"_y"+tileY, titles);
}
if ((globalDebugLevel > 0) && (tileX >= debug_tileX - 2) && (tileX <= debug_tileX + 2) &&
(tileY >= debug_tileY - 2) && (tileY <= debug_tileY+2)) {
for (int i = 0; i < quad_main; i++) {
System.out.println("clt_bi_quad(): color="+chn+", tileX="+tileX+", tileY="+tileY+
" fract_shiftsXY_main["+i+"][0]="+fract_shiftsXY_main[i][0]+" fract_shiftsXY_main["+i+"][1]="+fract_shiftsXY_main[i][1]);
}
for (int i = 0; i < quad_aux; i++) {
System.out.println("clt_bi_quad(): color="+chn+", tileX="+tileX+", tileY="+tileY+
" fract_shiftsXY_aux["+i+"][0]="+fract_shiftsXY_aux[i][0]+" fract_shiftsXY_aux["+i+"][1]="+fract_shiftsXY_aux[i][1]);
}
}
// apply residual shift
for (int i = 0; i < quad_main; i++) {
fract_shift( // fractional shift in transform domain. Currently uses sin/cos - change to tables with 2? rotations
clt_data_main[i][chn], // double [][] clt_tile,
clt_parameters.transform_size,
fract_shiftsXY_main[i][0], // double shiftX,
fract_shiftsXY_main[i][1], // double shiftY,
((globalDebugLevel > 1) && (chn==0) && (tileX >= debug_tileX - 2) && (tileX <= debug_tileX + 2) &&
(tileY >= debug_tileY - 2) && (tileY <= debug_tileY+2)));
}
for (int i = 0; i < quad_aux; i++) {
fract_shift( // fractional shift in transform domain. Currently uses sin/cos - change to tables with 2? rotations
clt_data_aux[i][chn], // double [][] clt_tile,
clt_parameters.transform_size,
fract_shiftsXY_aux[i][0], // double shiftX,
fract_shiftsXY_aux[i][1], // double shiftY,
((globalDebugLevel > 1) && (chn==0) && (tileX >= debug_tileX - 2) && (tileX <= debug_tileX + 2) &&
(tileY >= debug_tileY - 2) && (tileY <= debug_tileY+2)));
}
if ((globalDebugLevel > 0) && (debug_tileX == tileX) && (debug_tileY == tileY)) {
showDoubleFloatArrays sdfa_instance = new showDoubleFloatArrays(); // just for debugging?
String [] titles = {"CC0","SC0","CS0","SS0","CC1","SC1","CS1","SS1","CC2","SC2","CS2","SS2","CC3","SC3","CS3","SS3"};
double [][] dbg_tile = new double [16][];
for (int i = 0; i < 16; i++) dbg_tile[i]=clt_data_main[i>>2][chn][i & 3];
sdfa_instance.showArrays(dbg_tile, clt_parameters.transform_size, clt_parameters.transform_size, true, "MAIN_shifted_x"+tileX+"_y"+tileY+"-z", titles);
double [][] dbg_tile_aux = new double [16][];
for (int i = 0; i < 16; i++) dbg_tile[i]=clt_data_aux[i>>2][chn][i & 3];
sdfa_instance.showArrays(dbg_tile_aux, clt_parameters.transform_size, clt_parameters.transform_size, true, "AUX_shifted_x"+tileX+"_y"+tileY+"-z", titles);
}
}
int tile_lma_debug_level = ((tileX == debug_tileX) && (tileY == debug_tileY))? clt_parameters.img_dtt.lma_debug_level : -1;
// all color channels are done here
double extra_disparity_main = 0.0; // used for textures: if allowed, shift images extra before trying to combine
double extra_disparity_aux = 0.0; // used for textures: if allowed, shift images extra before trying to combine
// fill clt_corr_combo if it exists
if (disparity_bimap != null){ // not null - calculate correlations
// calculate overexposed fraction - remove ?
// if (saturation_imp != null){
// disparity_map[OVEREXPOSED][nTile] = (1.0 * overexp_all[0]) / overexp_all[1];
// }
double [] tile_corrs_main = tileCorrs(
clt_parameters, // final EyesisCorrectionParameters.CLTParameters clt_parameters,
fatzero, // final double fatzero, // May use correlation fat zero from 2 different parameters - fat_zero and rig.ml_fatzero
true, // final boolean get4dirs, // calculate disparity/strength for each of the 4 directions
corr2d, // final Correlation2d corr2d,
clt_data_main, // final double [][][][] clt_data,
filter, // final double [] filter,
col_weights, // final double [] col_weights,
ml_hwidth, // final int ml_hwidth,
ml_data_main, // final double [][] ml_center_corr,
tileX, // final int tileX, // only used in debug output
tileY, // final int tileY,
tile_lma_debug_level); // final int debugLevel)
double [] tile_corrs_aux = tileCorrs(
clt_parameters, // final EyesisCorrectionParameters.CLTParameters clt_parameters,
fatzero, // final double fatzero, // May use correlation fat zero from 2 different parameters - fat_zero and rig.ml_fatzero
true, // final boolean get4dirs, // calculate disparity/strength for each of the 4 directions
corr2d, // final Correlation2d corr2d,
clt_data_aux, // final double [][][][] clt_data,
filter, // final double [] filter,
col_weights, // final double [] col_weights,
ml_hwidth, // final int ml_hwidth,
ml_data_aux, // final double [][] ml_center_corr,
tileX, // final int tileX, // only used in debug output
tileY, // final int tileY,
tile_lma_debug_level); // final int debugLevel)
extra_disparity_main = tile_corrs_main[DISP_FULL_INDEX];
extra_disparity_aux = tile_corrs_aux[DISP_FULL_INDEX];
if (Double.isNaN(extra_disparity_main)) extra_disparity_main = 0;
if (Double.isNaN(extra_disparity_aux)) extra_disparity_aux = 0;
for (int i = 0; i < tile_corrs_main.length; i++) {
int dest = SNGL_TO_BI[0][i];
if (disparity_bimap[dest] != null) disparity_bimap[dest][nTile] = tile_corrs_main[i];
}
for (int i = 0; i < tile_corrs_aux.length; i++) {
int dest = SNGL_TO_BI[1][i];
if (disparity_bimap[dest] != null) disparity_bimap[dest][nTile] = tile_corrs_aux[i];
}
if (Double.isNaN(disparity_bimap[BI_STR_FULL_INDEX][tIndex])) {
System.out.println("BUG: 3. disparity_map[BI_STR_FULL_INDEX][tIndex] should not be NaN");
}
if (Double.isNaN(disparity_bimap[BI_ASTR_FULL_INDEX][tIndex])) {
System.out.println("BUG: 3a. disparity_map[BI_ASTR_FULL_INDEX][tIndex] should not be NaN");
}
if (clt_corr_combo != null) { // [type][tilesY][tilesX][(2*transform_size-1)*(2*transform_size-1)] // if null - will not calculate
tcorr_combo = new double [TCORR_TITLES.length][corr_size * corr_size];
}
double [] inter_cam_corr = corr2d.correlateInterCamerasFD(
clt_data_main, // double [][][][] clt_data_tile_main,
clt_data_aux, // double [][][][] clt_data_tile_aux,
filter, // double [] lpf,
col_weights, // double [] col_weights,
fatzero); // double fat_zero)
double [] inter_corrs_dxy = tileInterCamCorrs(
no_int_x0, // final boolean no_int_x0, // do not offset window to integer - used when averaging low textures to avoid "jumps" for very wide
clt_parameters, // final EyesisCorrectionParameters.CLTParameters clt_parameters,
inter_cam_corr, // final double [] inter_cam_corr,
corr2d, // final Correlation2d corr2d,
ml_hwidth, // final int ml_hwidth,
ml_data_inter, // final double [] ml_center_corr,
tcorr_combo, // double [][] tcorr_combo,
notch_mode, // final boolean notch_mode,
tileX, // final int tileX, // only used in debug output
tileY, // final int tileY,
tile_lma_debug_level); // final int debugLevel)
if (lt_corr != null) {
lt_corr[tIndex] = inter_cam_corr;
}
if (inter_corrs_dxy != null) {
disparity_bimap[BI_DISP_CROSS_INDEX][nTile] = inter_corrs_dxy[INDEX_DISP];
disparity_bimap[BI_STR_CROSS_INDEX][nTile] = inter_corrs_dxy[INDEX_STRENGTH];
disparity_bimap[BI_DISP_CROSS_DX_INDEX][nTile] = inter_corrs_dxy[INDEX_DX];
disparity_bimap[BI_DISP_CROSS_DY_INDEX][nTile] = inter_corrs_dxy[INDEX_DY];
// TODO: Use strength for the same residual disparity
disparity_bimap[BI_STR_ALL_INDEX][nTile] =
Math.pow(inter_corrs_dxy[INDEX_STRENGTH]*
disparity_bimap[BI_STR_FULL_INDEX][nTile]*
disparity_bimap[BI_ASTR_FULL_INDEX][nTile], 1.0/3);
}
// finalize ML stuff
if (ml_data != null) {
// save data for the main camera
for (int nlayer = 0; nlayer < ML_TOP_AUX_INDEX; nlayer++) {
// save main camera data
corr2d.saveMlTile(
tileX, // int tileX,
tileY, // int tileY,
ml_hwidth, // int ml_hwidth,
ml_data, // double [][] ml_data,
nlayer + 0, // int ml_layer,
ml_data_main[nlayer], // double [] ml_tile,
tilesX); // int tilesX);
// save aux_camera data
corr2d.saveMlTile(
tileX, // int tileX,
tileY, // int tileY,
ml_hwidth, // int ml_hwidth,
ml_data, // double [][] ml_data,
nlayer + ML_TOP_AUX_INDEX, // int ml_layer,
ml_data_aux[nlayer], // double [] ml_tile,
tilesX); // int tilesX);
}
// save inter-camera correlation
corr2d.saveMlTile(
tileX, // int tileX,
tileY, // int tileY,
ml_hwidth, // int ml_hwidth,
ml_data, // double [][] ml_data,
ML_INTER_INDEX, // int ml_layer,
ml_data_inter, // double [] ml_tile,
tilesX); // int tilesX);
// save other data (just 1 value)
/*
corr2d.saveMlTile(
tileX, // int tileX,
tileY, // int tileY,
ml_hwidth, // int ml_hwidth,
ml_data, // double [][] ml_data,
ML_OTHER_INDEX, // int ml_layer,
ml_data_other, // double [] ml_tile,
tilesX); // int tilesX);
*/
corr2d.saveMlTilePixel(
tileX, // int tileX,
tileY, // int tileY,
ml_hwidth, // int ml_hwidth,
ml_data, // double [][] ml_data,
ML_OTHER_INDEX, // int ml_layer,
ML_OTHER_TARGET , // int ml_index,
disparity_main, // target disparitydouble ml_value,
tilesX); // int tilesX);
if (ml_data_dbg1 != null) {
tileInterCamCorrs(
false, // final boolean no_int_x0, // do not offset window to integer - used when averaging low textures to avoid "jumps" for very wide
clt_parameters, // final EyesisCorrectionParameters.CLTParameters clt_parameters,
fatzero, // final double fatzero, // May use correlation fat zero from 2 different parameters - fat_zero and rig.ml_fatzero
corr2d, // final Correlation2d corr2d,
clt_data_main, // double [][][][] clt_data_tile_main,
clt_data_main, // double [][][][] clt_data_tile_aux,
filter, // final double [] filter,
col_weights, // final double [] col_weights,
ml_hwidth, // final int ml_hwidth,
ml_data_dbg1, // final double [] ml_center_corr,
null, // double [][] tcorr_combo,
false, // final boolean notch_mode,
tileX, // final int tileX, // only used in debug output
tileY, // final int tileY,
tile_lma_debug_level); // final int debugLevel)
corr2d.saveMlTile(
tileX, // int tileX,
tileY, // int tileY,
ml_hwidth, // int ml_hwidth,
ml_data, // double [][] ml_data,
ML_DBG1_INDEX, // int ml_layer,
ml_data_dbg1, // double [] ml_tile,
tilesX); // int tilesX);
}
}
} // if (disparity_map != null){ // not null - calculate correlations
if (tcorr_combo != null) { // [type][tilesY][tilesX][(2*transform_size-1)*(2*transform_size-1)] // if null - will not calculate
for (int i = 0; i < tcorr_combo.length; i++) {
clt_corr_combo[i][tileY][tileX] = tcorr_combo[i];
}
}
if (texture_tiles_main !=null) {
generateTextureTiles (
clt_parameters, // final EyesisCorrectionParameters.CLTParameters clt_parameters,
extra_disparity_main, // final double extra_disparity,
quad_main, // final int quad, // number of subcameras
numcol, // final int numcol, // number of colors
img_mask, // _main, // int img_mask,
tile_op, // _main, // final int [][] tile_op, // [tilesY][tilesX] - what to do - 0 - nothing for this tile
clt_data_main, // final double [][][][] clt_data,
texture_tiles_main, // final double [][][][] texture_tiles, // [tilesY][tilesX]["RGBA".length()][]; null - will skip images combining
filter, // final double [] filter,
lt_window2, // final double [] lt_window2,
port_offsets, // final double[][] port_offsets,
col_weights, // final double [] col_weights,
dtt, // final DttRad2 dtt,
tileX, // final int tileX, // only used in debug output
tileY, // final int tileY,
tile_lma_debug_level); // final int debugLevel);
}
if (texture_tiles_aux !=null) {
generateTextureTiles (
clt_parameters, // final EyesisCorrectionParameters.CLTParameters clt_parameters,
extra_disparity_aux, // final double extra_disparity,
quad_aux, // final int quad, // number of subcameras
numcol, // final int numcol, // number of colors
img_mask, // _aux, // int img_mask,
tile_op, // _aux, // final int [][] tile_op, // [tilesY][tilesX] - what to do - 0 - nothing for this tile
clt_data_aux, // final double [][][][] clt_data,
texture_tiles_aux, // final double [][][][] texture_tiles, // [tilesY][tilesX]["RGBA".length()][]; null - will skip images combining
filter, // final double [] filter,
lt_window2, // final double [] lt_window2,
port_offsets, // final double[][] port_offsets,
col_weights, // final double [] col_weights,
dtt, // final DttRad2 dtt,
tileX, // final int tileX, // only used in debug output
tileY, // final int tileY,
tile_lma_debug_level); // final int debugLevel);
}
// Save channel tiles to result
if (clt_bidata != null) {
for (int i = 0; i < quad_main; i++) for (int j = 0; j < numcol; j++){
clt_bidata[0][i][j][tileY][tileX] = clt_data_main[i][j];
}
for (int i = 0; i < quad_aux; i++) for (int j = 0; j < numcol; j++){
clt_bidata[1][i][j][tileY][tileX] = clt_data_aux[i][j];
}
}
}
}
};
}
startAndJoin(threads);
// If it was low-texture mode, use lt_corr to average bi-quad inter-correlation between neighbor tiles and then calculate disparity/strength
if (lt_corr != null) {
//notch_mode
// prepare weights for neighbors
final int lt_rad_x = notch_mode? 0: lt_rad;
final double [][] neib_weights = new double[lt_rad+1][lt_rad_x+1];
for (int i = 0; i <= lt_rad; i++) {
for (int j = 0; j <= lt_rad_x; j++) {
neib_weights[i][j] = Math.cos(Math.PI * i /(2 * lt_rad + 1)) * Math.cos(Math.PI * j /(2 * lt_rad + 1)); // no need to normalize - it will need to skip empty tiles anyway
}
}
// final int corr_size = clt_parameters.transform_size * 2 -1;
final TileNeibs tnImage = new TileNeibs(tilesX, tilesY);
ai.set(0);
for (int ithread = 0; ithread < threads.length; ithread++) {
threads[ithread] = new Thread() {
@Override
public void run() {
double [] ml_data_inter = (ml_data != null)? new double [(2*ml_hwidth +1)*(2*ml_hwidth +1)]:null;
double [][] tcorr_combo = null; // [15*15] pixel space
Correlation2d corr2d = new Correlation2d(
clt_parameters.img_dtt, // ImageDttParameters imgdtt_params,
clt_parameters.transform_size, // int transform_size,
2.0, // double wndx_scale, // (wndy scale is always 1.0)
(globalDebugLevel > -1)); // boolean debug)
for (int nTile = ai.getAndIncrement(); nTile < nTilesInChn; nTile = ai.getAndIncrement()) if (lt_corr[nTile] != null){ // center must be non-null (from tile_op)
int tileX = nTile % tilesX;
int tileY = nTile / tilesX;
double [] tile_corrs = new double [corr_size * corr_size];
double sw = 0.0;
for (int dy = -lt_rad; dy <= lt_rad; dy++) {
int ady = (dy > 0)? dy:(-dy);
for (int dx = -lt_rad_x; dx <= lt_rad_x; dx++) {
int nTile1 = tnImage.getNeibIndex(nTile, dx, dy);
if (nTile1 >= 0) {
double [] ot_corr = lt_corr[nTile1];
if (ot_corr != null) {
int adx = (dx > 0)? dx:(-dx);
double nw = neib_weights[ady][adx];
for (int i = 0; i < tile_corrs.length; i++) {
tile_corrs[i] += nw * ot_corr[i];
}
sw+=nw;
}
}
}
}
if (sw > 0.0) { // with the current window should always be so, as te center tile is non-null
double s = 1.0/sw;
for (int i = 0; i < tile_corrs.length; i++) {
tile_corrs[i] *= s;
}
if (clt_corr_combo != null) { // [type][tilesY][tilesX][(2*transform_size-1)*(2*transform_size-1)] // if null - will not calculate
tcorr_combo = new double [TCORR_TITLES.length][corr_size * corr_size];
}
int tile_lma_debug_level = ((tileX == debug_tileX) && (tileY == debug_tileY))? clt_parameters.img_dtt.lma_debug_level : -1;
double [] inter_corrs_dxy = tileInterCamCorrs(
true, // final boolean no_int_x0, // do not offset window to integer - used when averaging low textures to avoid "jumps" for very wide
clt_parameters, // final EyesisCorrectionParameters.CLTParameters clt_parameters,
tile_corrs, // final double [] inter_cam_corr,
corr2d, // final Correlation2d corr2d,
ml_hwidth, // final int ml_hwidth,
ml_data_inter, // final double [] ml_center_corr,
tcorr_combo, // double [][] tcorr_combo,
notch_mode, // final boolean notch_mode,
tileX, // final int tileX, // only used in debug output
tileY, // final int tileY,
tile_lma_debug_level); // final int debugLevel)
if (inter_corrs_dxy != null) {
disparity_bimap[BI_DISP_CROSS_INDEX][nTile] = inter_corrs_dxy[INDEX_DISP];
disparity_bimap[BI_STR_CROSS_INDEX][nTile] = inter_corrs_dxy[INDEX_STRENGTH];
disparity_bimap[BI_DISP_CROSS_DX_INDEX][nTile] = inter_corrs_dxy[INDEX_DX];
disparity_bimap[BI_DISP_CROSS_DY_INDEX][nTile] = inter_corrs_dxy[INDEX_DY];
// TODO: Use strength for the same residual disparity
disparity_bimap[BI_STR_ALL_INDEX][nTile] =
Math.pow(inter_corrs_dxy[INDEX_STRENGTH]*
disparity_bimap[BI_STR_FULL_INDEX][nTile]*
disparity_bimap[BI_ASTR_FULL_INDEX][nTile], 1.0/3);
}
if (tcorr_combo != null) { // [type][tilesY][tilesX][(2*transform_size-1)*(2*transform_size-1)] // if null - will not calculate
for (int i = 0; i < tcorr_combo.length; i++) {
clt_corr_combo[i][tileY][tileX] = tcorr_combo[i];
}
}
if (ml_data != null) {
// save inter-camera correlation
corr2d.saveMlTile(
tileX, // int tileX,
tileY, // int tileY,
ml_hwidth, // int ml_hwidth,
ml_data, // double [][] ml_data,
ML_INTER_INDEX, // int ml_layer,
ml_data_inter, // double [] ml_tile,
tilesX); // int tilesX);
}
// TODO: save ml_data_inter, tcorr_combo
}
}
}
};
}
startAndJoin(threads);
}
return clt_bidata;
}
public double [][][][][][][] clt_bi_quad( public double [][][][][][][] clt_bi_quad(
final EyesisCorrectionParameters.CLTParameters clt_parameters, final EyesisCorrectionParameters.CLTParameters clt_parameters,
final double fatzero, // May use correlation fat zero from 2 different parameters - fat_zero and rig.ml_fatzero final double fatzero, // May use correlation fat zero from 2 different parameters - fat_zero and rig.ml_fatzero
...@@ -8367,7 +9153,6 @@ public class ImageDtt { ...@@ -8367,7 +9153,6 @@ public class ImageDtt {
return clt_bidata; return clt_bidata;
} }
public void clt_bi_macro( // not yet operational public void clt_bi_macro( // not yet operational
final EyesisCorrectionParameters.CLTParameters clt_parameters, final EyesisCorrectionParameters.CLTParameters clt_parameters,
final double fatzero, // May use correlation fat zero from 2 different parameters - fat_zero and rig.ml_fatzero final double fatzero, // May use correlation fat zero from 2 different parameters - fat_zero and rig.ml_fatzero
......
src/main/java/TwoQuadCLT.java
View file @ f04bcc82
...@@ -20,6 +20,7 @@ ...@@ -20,6 +20,7 @@
** -----------------------------------------------------------------------------** ** -----------------------------------------------------------------------------**
** **
*/ */
import java.io.DataOutputStream;
import java.io.File; import java.io.File;
import java.io.FileNotFoundException; import java.io.FileNotFoundException;
import java.io.FileOutputStream; import java.io.FileOutputStream;
...@@ -907,7 +908,80 @@ public class TwoQuadCLT { ...@@ -907,7 +908,80 @@ public class TwoQuadCLT {
return results; return results;
} }
public void saveFloatKernels(String file_prefix,
double [][][][][][] clt_kernels,
double [][][] image_data,
double [][][] port_xy,
boolean transpose) throws IOException {
if (clt_kernels != null) {
for (int chn = 0; chn < clt_kernels.length; chn++) {
String kern_path = file_prefix+"_chn"+chn+(transpose?"_transposed":"")+".kernel";
String offs_path = file_prefix+"_chn"+chn+(transpose?"_transposed":"")+".kernel_offsets";
FileOutputStream fos = new FileOutputStream(kern_path);
DataOutputStream dos = new DataOutputStream(fos);
for (int ty = 0; ty < clt_kernels[chn][0].length; ty++) {
for (int tx = 0; tx < clt_kernels[chn][0][ty].length; tx++) {
for (int col = 0; col < clt_kernels[chn].length; col++) {
for (int p = 0; p < 4; p++) {
double [] pa = clt_kernels[chn][col][ty][tx][p];
for (int i0 = 0; i0 < 64; i0++) {
int i;
if (transpose) {
i = ((i0 & 7) << 3) + ((i0 >>3) & 7);
} else {
i = i0;
}
dos.writeFloat((float)pa[i]);
}
}
}
}
}
dos.close();
fos = new FileOutputStream(offs_path);
dos = new DataOutputStream(fos);
for (int ty = 0; ty < clt_kernels[chn][0].length; ty++) {
for (int tx = 0; tx < clt_kernels[chn][0][ty].length; tx++) {
for (int col = 0; col < clt_kernels[chn].length; col++) {
double [] pa = clt_kernels[chn][col][ty][tx][4];
for (int i = 0; i < pa.length; i++) {
dos.writeFloat((float)pa[i]);
}
}
}
}
dos.close();
}
}
if (image_data != null) {
for (int chn = 0; chn < image_data.length; chn++) {
String img_path = file_prefix+"_chn"+chn+".bayer";
FileOutputStream fos = new FileOutputStream(img_path);
DataOutputStream dos = new DataOutputStream(fos);
for (int i = 0; i < image_data[chn][0].length; i++) {
dos.writeFloat((float) (image_data[chn][0][i] + image_data[chn][1][i] + image_data[chn][2][i]));
}
dos.close();
}
}
if (port_xy != null) {
for (int chn = 0; chn < port_xy[0].length; chn++) {
String img_path = file_prefix+"_chn"+chn+".portsxy";
FileOutputStream fos = new FileOutputStream(img_path);
DataOutputStream dos = new DataOutputStream(fos);
for (int i = 0; i < port_xy.length; i++) {
dos.writeFloat((float) (port_xy[i][chn][0])); // x-offset
dos.writeFloat((float) (port_xy[i][chn][1])); // y-offset
}
dos.close();
}
}
}
public ImagePlus [] processCLTQuadCorrPairGpu( public ImagePlus [] processCLTQuadCorrPairGpu(
GPUTileProcessor gPUTileProcessor, GPUTileProcessor gPUTileProcessor,
...@@ -1003,6 +1077,8 @@ public class TwoQuadCLT { ...@@ -1003,6 +1077,8 @@ public class TwoQuadCLT {
double [][][][][][] clt_kernels_main = quadCLT_main.getCLTKernels(); // [4][3][123][164]{[64],[64],[64],[64],[8]} double [][][][][][] clt_kernels_main = quadCLT_main.getCLTKernels(); // [4][3][123][164]{[64],[64],[64],[64],[8]}
double [][][][][][] clt_kernels_aux = quadCLT_aux.getCLTKernels(); double [][][][][][] clt_kernels_aux = quadCLT_aux.getCLTKernels();
//[4][3][123][164][5][]
double [][] dbg_kern = clt_kernels_main[0][0][0][0]; double [][] dbg_kern = clt_kernels_main[0][0][0][0];
// here all data is ready (images, kernels) to try GPU code // here all data is ready (images, kernels) to try GPU code
...@@ -1033,13 +1109,12 @@ public class TwoQuadCLT { ...@@ -1033,13 +1109,12 @@ public class TwoQuadCLT {
"converted", "converted",
dbg_titles); dbg_titles);
if (debugLevel < 1000) {
return null;
}
double [][][] port_xy_main_dbg = new double [tilesX*tilesY][][];
double [][][] port_xy_aux_dbg = new double [tilesX*tilesY][][];
final double [][][][][][][] clt_bidata = // new double[2][quad][nChn][tilesY][tilesX][][]; // first index - main/aux final double [][][][][][][] clt_bidata = // new double[2][quad][nChn][tilesY][tilesX][][]; // first index - main/aux
image_dtt.clt_bi_quad ( image_dtt.clt_bi_quad_dbg (
clt_parameters, // final EyesisCorrectionParameters.CLTParameters clt_parameters, clt_parameters, // final EyesisCorrectionParameters.CLTParameters clt_parameters,
clt_parameters.fat_zero, // final double fatzero, // May use correlation fat zero from 2 different parameters - fat_zero and rig.ml_fatzero clt_parameters.fat_zero, // final double fatzero, // May use correlation fat zero from 2 different parameters - fat_zero and rig.ml_fatzero
notch_mode, // final boolean notch_mode, // use notch filter for inter-camera correlation to detect poles notch_mode, // final boolean notch_mode, // use notch filter for inter-camera correlation to detect poles
...@@ -1070,8 +1145,43 @@ public class TwoQuadCLT { ...@@ -1070,8 +1145,43 @@ public class TwoQuadCLT {
// woi_tops, // final int [][] woi_tops, // woi_tops, // final int [][] woi_tops,
ers_delay, // final double [][][] ers_delay, // if not null - fill with tile center acquisition delay ers_delay, // final double [][][] ers_delay, // if not null - fill with tile center acquisition delay
threadsMax, // final int threadsMax, // maximal number of threads to launch threadsMax, // final int threadsMax, // maximal number of threads to launch
debugLevel); // final int globalDebugLevel); debugLevel, // final int globalDebugLevel);
port_xy_main_dbg, // final double [][][] port_xy_main_dbg, // for each tile/port save x,y pixel coordinates (gpu code development)
port_xy_aux_dbg); // final double [][][] port_xy_aux_dbg) // for each tile/port save x,y pixel coordinates (gpu code development)
String kernel_dir = "/home/eyesis/workspace-python3/nvidia_dct8x8/clt/";
boolean [][] what_to_save = {{false,false,true}, {false,false,true}};
try {
saveFloatKernels(
kernel_dir +"main", // String file_prefix,
(what_to_save[0][0]?clt_kernels_main:null), // double [][][][][][] clt_kernels, // null
(what_to_save[0][1]?quadCLT_main.image_data:null),
(what_to_save[0][2]?port_xy_main_dbg:null), // double [][][] port_xy,
true);
} catch (IOException e) {
System.out.println("Failed to save flattened kernels tp "+kernel_dir);
// TODO Auto-generated catch block
e.printStackTrace();
} // boolean transpose);
try {
saveFloatKernels(
kernel_dir +"aux", // String file_prefix,
(what_to_save[1][0]?clt_kernels_aux:null), // double [][][][][][] clt_kernels, // null
(what_to_save[1][1]?quadCLT_aux.image_data:null),
(what_to_save[1][2]?port_xy_aux_dbg:null), // double [][][] port_xy,
true);
} catch (IOException e) {
System.out.println("Failed to save flattened kernels tp "+kernel_dir);
// TODO Auto-generated catch block
e.printStackTrace();
} // boolean transpose);
if (debugLevel < 1000) {
return null;
}
if (ers_delay !=null) { if (ers_delay !=null) {
showERSDelay(ers_delay); showERSDelay(ers_delay);
......
src/main/resources/dtt8x8.cuh
View file @ f04bcc82
...@@ -43,6 +43,7 @@ ...@@ -43,6 +43,7 @@
#define DTTTEST_BLK_STRIDE (DTTTEST_BLOCK_WIDTH+1) #define DTTTEST_BLK_STRIDE (DTTTEST_BLOCK_WIDTH+1)
#define DTT_SIZE 8 #define DTT_SIZE 8
//#define CUDART_INF_F __int_as_float(0x7f800000)
/* /*
Python code to generate constant coefficients: Python code to generate constant coefficients:
def dct_constants(): def dct_constants():
......
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