Implemented 8bpp mode for ML output (b6b0e4d8) · Commits · Elphel / imagej-elphel

src/main/java/BiQuadParameters.java

+7 −2

Original line number	Diff line number	Diff line
		@@ -80,6 +80,7 @@ public class BiQuadParameters {
		public boolean ml_keep_hor_vert = true; // include combined horizontal and vertical pairs data in the ML output
		public boolean ml_keep_debug= true; // include debug layer(s) data in the ML output
		public boolean ml_8bit= true; // output in 8-bit format (default - 32-bit TIFF
		public double ml_limit_extrim = 0.00001; // ignore lowest and highest values when converting to 8 bpp
		public boolean ml_show_ml = true; // show each generated MLoutput file


		@@ -185,6 +186,8 @@ public class BiQuadParameters {
		"Keep additional (debug) layers that may change for different file versions");
		gd.addCheckbox ("Use 8 bpp TIFF (default - 32 bpp)", this.ml_8bit,
		"Reduce file size by lowering bpp");
		gd.addNumericField("When converting to 8bpp, limit fraction of extreme values", 1E6 * this.ml_limit_extrim, 1,8,"ppm",
		"Use values histogram to find min/max values, ignoring(limiting) this fraction (parts per million) of pixels at both extremes");
		gd.addCheckbox ("Show each generated ML file", this.ml_show_ml,
		"Use only for small number of generated files to reduce memory usage");

		@@ -241,6 +244,7 @@ public class BiQuadParameters {
		this.ml_keep_hor_vert= gd.getNextBoolean();
		this.ml_keep_debug= gd.getNextBoolean();
		this.ml_8bit= gd.getNextBoolean();
		this.ml_limit_extrim= gd.getNextNumber() * 1E-6;
		this.ml_show_ml= gd.getNextBoolean();
		}

		@@ -298,6 +302,7 @@ public class BiQuadParameters {
		properties.setProperty(prefix+"ml_keep_hor_vert", this.ml_keep_hor_vert+"");
		properties.setProperty(prefix+"ml_keep_debug", this.ml_keep_debug+"");
		properties.setProperty(prefix+"ml_8bit", this.ml_8bit+"");
		properties.setProperty(prefix+"ml_limit_extrim", this.ml_limit_extrim+"");
		properties.setProperty(prefix+"ml_show_ml", this.ml_show_ml+"");


		@@ -353,6 +358,7 @@ public class BiQuadParameters {
		if (properties.getProperty(prefix+"ml_keep_hor_vert")!=null) this.ml_keep_hor_vert=Boolean.parseBoolean(properties.getProperty(prefix+"ml_keep_hor_vert"));
		if (properties.getProperty(prefix+"ml_keep_debug")!=null) this.ml_keep_debug=Boolean.parseBoolean(properties.getProperty(prefix+"ml_keep_debug"));
		if (properties.getProperty(prefix+"ml_8bit")!=null) this.ml_8bit=Boolean.parseBoolean(properties.getProperty(prefix+"ml_8bit"));
		if (properties.getProperty(prefix+"ml_limit_extrim")!=null) this.ml_limit_extrim=Double.parseDouble(properties.getProperty(prefix+"ml_limit_extrim"));
		if (properties.getProperty(prefix+"ml_show_ml")!=null) this.ml_show_ml=Boolean.parseBoolean(properties.getProperty(prefix+"ml_show_ml"));
		}
		@Override
		@@ -408,9 +414,8 @@ public class BiQuadParameters {
		bqp.ml_keep_hor_vert= this.ml_keep_hor_vert;
		bqp.ml_keep_debug= this.ml_keep_debug;
		bqp.ml_8bit= this.ml_8bit;
		bqp.ml_limit_extrim= this.ml_limit_extrim;
		bqp.ml_show_ml= this.ml_show_ml;
		return bqp;


		}
		}

src/main/java/Correlation2d.java

+47 −0

Original line number	Diff line number	Diff line
		@@ -1162,7 +1162,54 @@ public class Correlation2d {
		int oindex = tileY tile_width full_width + tileX * tile_width + (ml_index/tile_width)*full_width + (ml_index%tile_width) ;
		ml_data[ml_layer][oindex] = ml_value;
		}
		/**
		* Get a single value from the combined multi-layer ML array, viewable as an image
		* @param tileX horizontal tile index
		* @param tileY vertical tile index
		* @param ml_hwidth half-width of the preserved 2d correlation (0 - single point, 1 -> 3x3, 2 -> 5x5, 7 - all data)
		* @param ml_data multi-layer array, each layer matches an image of ((2 * ml_hwidth + 1) * tilesX) by ((2 * ml_hwidth + 1) * tilesY) in scanline order
		* Each tile corresponds to (2 * ml_hwidth + 1) * (2 * ml_hwidth + 1) square in the image. Only selected tiles will be updated, so it is good to initialize array
		* with all Double.NaN values
		* @param ml_layer layer to save tile data
		* @param ml_index data index within tile
		* @param tilesX image width in tiles
		* @return value indexed by tileX, tileY, ml_layer and ml_index
		*/
		public double restoreMlTilePixel(
		int tileX,
		int tileY,
		int ml_hwidth,
		double [][] ml_data,
		int ml_layer,
		int ml_index,
		int tilesX) {
		int tile_width = 2 * ml_hwidth + 1;
		int full_width = tile_width * tilesX;
		int oindex = tileY tile_width full_width + tileX * tile_width + (ml_index/tile_width)*full_width + (ml_index%tile_width) ;
		return ml_data[ml_layer][oindex];
		}

		/**
		* Get an index of the selected tile+index in a ML array layer
		* @param tileX horizontal tile index
		* @param tileY vertical tile index
		* @param ml_hwidth half-width of the preserved 2d correlation (0 - single point, 1 -> 3x3, 2 -> 5x5, 7 - all data)
		* @param ml_index data index within tile
		* @param tilesX image width in tiles
		* @return index of teh selected pixel in thye whole image (specified by tileX, tileY, and ml_index)
		*/

		public int getMlTilePixelIndex(
		int tileX,
		int tileY,
		int ml_hwidth,
		int ml_index,
		int tilesX) {
		int tile_width = 2 * ml_hwidth + 1;
		int full_width = tile_width * tilesX;
		int oindex = tileY tile_width full_width + tileX * tile_width + (ml_index/tile_width)*full_width + (ml_index%tile_width) ;
		return oindex;
		}

src/main/java/TwoQuadCLT.java

+158 −7

Original line number	Diff line number	Diff line
		@@ -1400,6 +1400,11 @@ public class TwoQuadCLT {
		String ml_directory= quadCLT_main.correctionsParameters.selectMlDirectory(
		true, // smart,
		true); //newAllowed, // save
		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)
		(debugLevel > -1)); // boolean debug)

		for (int sweep_step = 0; sweep_step < clt_parameters.rig.ml_sweep_steps; sweep_step++){
		double disparity_offset = 0; // clt_parameters.rig.ml_disparity_sweep * (2.0 * sweep_step/(clt_parameters.rig.ml_sweep_steps - 1.0) -1.0);
		@@ -1422,6 +1427,9 @@ public class TwoQuadCLT {
		disparity_offset, // double disp_offset,
		quadCLT_main, // QuadCLT quadCLT_main,
		quadCLT_aux, // QuadCLT quadCLT_aux,
		corr2d, //Correlation2d corr2d, // to access "other" layer
		clt_parameters.rig.ml_8bit, // boolean use8bpp,
		clt_parameters.rig.ml_limit_extrim, // double limit_extrim,
		clt_parameters.rig.ml_keep_aux, // boolean keep_aux,
		clt_parameters.rig.ml_keep_inter, // boolean keep_inter,
		clt_parameters.rig.ml_keep_hor_vert, // boolean keep_hor_vert,
		@@ -1441,6 +1449,9 @@ public class TwoQuadCLT {
		double disp_offset,
		QuadCLT quadCLT_main,
		QuadCLT quadCLT_aux,
		Correlation2d corr2d, // to access "other" layer
		boolean use8bpp,
		double limit_extrim,
		boolean keep_aux,
		boolean keep_inter,
		boolean keep_hor_vert,
		@@ -1478,19 +1489,155 @@ public class TwoQuadCLT {
		int [] dbg_indices = {
		ImageDtt.ML_DBG1_INDEX //18 - just debug data (first - auto phase correlation)
		};
		int [] non_corr_indices = {
		ImageDtt.ML_OTHER_INDEX, //17 - other data: 0 (top left tile corner) - preset disparity of the tile, 1: (next element) - ground trouth data, 2:
		ImageDtt.ML_DBG1_INDEX //18 - just debug data (first - auto phase correlation)
		};

		boolean [] skip_layers = new boolean [ImageDtt.ML_TITLES.length];
		if (!keep_aux) for (int nl:aux_indices) skip_layers[nl] = true;
		if (!keep_inter) for (int nl:inter_indices) skip_layers[nl] = true;
		if (!keep_hor_vert) for (int nl:hor_vert_indices) skip_layers[nl] = true;
		if (!keep_debug) for (int nl:dbg_indices) skip_layers[nl] = true;

		float [] fpixels;
		ImageStack array_stack=new ImageStack(width,height);
		double soft_mn = Double.NaN,soft_mx = Double.NaN;
		if (use8bpp) {
		int num_bins = 256;
		boolean [] skip_histogram = skip_layers.clone();
		for (int nl:non_corr_indices) skip_histogram[nl] = true;
		double mn = 0.0, mx = 0.0; // data has both positive and negative values
		for (int nl = 0; nl < ml_data.length; nl++) if (!skip_histogram[nl]) {
		for (int i = 0; i < ml_data[nl].length; i++) if (!Double.isNaN(ml_data[nl][i])){
		if (ml_data[nl][i] > mx) mx = ml_data[nl][i];
		else if (ml_data[nl][i] < mn) mn = ml_data[nl][i];
		}
		}
		if (debugLevel > -2) {
		System.out.println("saveMlFile(): min="+mn+", max="+mx);
		}
		int [] histogram = new int [num_bins];
		int num_values = 0;
		for (int nl = 0; nl < ml_data.length; nl++) if (!skip_histogram[nl]) {
		for (int i = 0; i < ml_data[nl].length; i++) if (!Double.isNaN(ml_data[nl][i])){
		int bin = (int) Math.round(num_bins*(ml_data[nl][i] - mn)/(mx-mn));
		// rounding errors?
		if (bin < 0) bin = 0;
		else if (bin >= num_bins) bin = num_bins-1;
		histogram[bin]++;
		num_values++;
		}
		}
		double ignore_vals = limit_extrim*num_values;
		soft_mn = mn;
		soft_mx = mx;
		{
		double sl = 0.0;
		int i = 0;
		while (sl < ignore_vals) {
		i++;
		sl+= histogram[i];
		soft_mn += (mx-mn)/num_bins;
		}
		double f = (sl - ignore_vals)/histogram[i];
		soft_mn -= (mx-mn)/num_bins*(1.0 - f);

		sl = 0.0;
		i = num_bins-1;
		while (sl < ignore_vals) {
		i--;
		sl+= histogram[i];
		soft_mx -= (mx-mn)/num_bins;
		}
		f = (sl - ignore_vals)/histogram[i];
		soft_mn += (mx-mn)/num_bins*(1.0 - f);
		if (debugLevel > -2) {
		System.out.println("saveMlFile(): soft min="+soft_mn+", soft max="+soft_mx);
		}
		}
		// convert double data to byte, so v<=soft_mn -> 1; v>= soft_mx -> 255, NaN -> 0
		byte [][] iml_data = new byte [ml_data.length][];
		for (int nl = 0; nl < ml_data.length; nl++) if (!skip_layers[nl]) {
		iml_data[nl] = new byte [ml_data[nl].length];
		if (nl == ImageDtt.ML_OTHER_INDEX) {
		// special treatment - make 2 bytes of one disparity value
		for (int tileY = 0; tileY < tilesY; tileY++) {
		for (int tileX = 0; tileX < tilesX; tileX++) {
		int nTile = tileY * tilesX + tileX;
		double target_disparity = corr2d.restoreMlTilePixel(
		tileX, // int tileX,
		tileY, // int tileY,
		ml_hwidth, // int ml_hwidth,
		ml_data, // double [][] ml_data,
		ImageDtt.ML_OTHER_INDEX, // int ml_layer,
		ImageDtt.ML_OTHER_TARGET , // int ml_index,
		tilesX); // int tilesX);
		double gtruth_disparity = corr2d.restoreMlTilePixel(
		tileX, // int tileX,
		tileY, // int tileY,
		ml_hwidth, // int ml_hwidth,
		ml_data, // double [][] ml_data,
		ImageDtt.ML_OTHER_INDEX, // int ml_layer,
		ImageDtt.ML_OTHER_GTRUTH , // int ml_index,
		tilesX); // int tilesX);
		double gtruth_strength = corr2d.restoreMlTilePixel(
		tileX, // int tileX,
		tileY, // int tileY,
		ml_hwidth, // int ml_hwidth,
		ml_data, // double [][] ml_data,
		ImageDtt.ML_OTHER_INDEX, // int ml_layer,
		ImageDtt.ML_OTHER_GTRUTH_STRENGTH , // int ml_index,
		tilesX); // int tilesX);
		// converting disparity to 9.7 ( 1/128 pixel step, +/-256 pixels disparity range), 0x8000 - zero disparity
		// converting strength to 2 bytes 0.16 fixed point
		int itd = (int) Math.round(128 * target_disparity) + 0x8000;
		int [] itarget_disparity = {itd >> 8, itd & 0xff};
		int igt = (int) Math.round(128 * gtruth_disparity) + 0x8000;
		int [] igtruth_disparity = {igt >> 8, igt & 0xff};
		int igs = (int) Math.round(0x10000 * gtruth_strength);
		int [] igtruth_strength = {igs >> 8, igs & 0xff};
		for (int nb = 0; nb<2; nb++) {
		if (!Double.isNaN(target_disparity)) {
		int indx = corr2d.getMlTilePixelIndex(tileX,tileY, ml_hwidth, ImageDtt.ML_OTHER_TARGET + nb, tilesX);
		iml_data[nl][indx] = (byte) itarget_disparity[nb];
		}
		if (!Double.isNaN(gtruth_disparity)) {
		int indx = corr2d.getMlTilePixelIndex(tileX,tileY, ml_hwidth, ImageDtt.ML_OTHER_GTRUTH + nb, tilesX);
		iml_data[nl][indx] = (byte) igtruth_disparity[nb];
		}
		if (gtruth_strength > 0.0) {
		int indx = corr2d.getMlTilePixelIndex(tileX,tileY, ml_hwidth, ImageDtt.ML_OTHER_GTRUTH_STRENGTH + nb, tilesX);
		iml_data[nl][indx] = (byte) igtruth_strength[nb];
		}
		}
		}
		}
		} else {
		double k = 254.0/(soft_mx-soft_mn);
		for (int i = 0; i < ml_data[nl].length;i++) {
		if (Double.isNaN(ml_data[nl][i])){
		iml_data[nl][i] = 0; // -128;
		} else {
		int iv = (int) Math.round(k*(ml_data[nl][i]-soft_mn));
		if (iv < 0) iv = 0;
		else if (iv > 254) iv = 254;
		iml_data[nl][i] = (byte) iv; // (iv - 127); // NaN will stay 0;
		}
		}
		}
		}
		for (int nl = 0; nl< ml_data.length; nl++) if (!skip_layers[nl]) {
		array_stack.addSlice(ImageDtt.ML_TITLES[nl], iml_data[nl]);
		}
		} else {
		float [] fpixels;
		for (int nl = 0; nl< ml_data.length; nl++) if (!skip_layers[nl]) {
		fpixels=new float[ml_data[nl].length];
		for (int j=0;j<fpixels.length;j++) fpixels[j]=(float) ml_data[nl][j];
		array_stack.addSlice(ImageDtt.ML_TITLES[nl], fpixels);
		}
		}

		double disparityRadiusMain = quadCLT_main.geometryCorrection.getDisparityRadius();
		double disparityRadiusAux = quadCLT_aux.geometryCorrection.getDisparityRadius();
		double intercameraBaseline = quadCLT_aux.geometryCorrection.getBaseline();
		@@ -1505,12 +1652,16 @@ public class TwoQuadCLT {
		imp_ml.setProperty("disparityRadiusMain", ""+disparityRadiusMain);
		imp_ml.setProperty("disparityRadiusAux", ""+disparityRadiusAux);
		imp_ml.setProperty("intercameraBaseline", ""+intercameraBaseline);
		imp_ml.setProperty("data_min", ""+soft_mn);
		imp_ml.setProperty("data_max", ""+soft_mx);

		imp_ml.setProperty("comment_tileWidth", "Square tile size for each 2d correlation, always odd");
		imp_ml.setProperty("comment_dispOffset", "Tile target disparity minum ground truth disparity");
		imp_ml.setProperty("comment_ML_OTHER_TARGET", "Offset of the target disparity in the \"other\" layer tile");
		imp_ml.setProperty("comment_ML_OTHER_GTRUTH", "Offset of the ground truth disparity in the \"other\" layer tile");
		imp_ml.setProperty("comment_ML_OTHER_GTRUTH_STRENGTH", "Offset of the ground truth strength in the \"other\" layer tile");
		imp_ml.setProperty("comment_data_min", "Defined only for 8bpp mode - value, corresponding to -127 (-128 is NaN)");
		imp_ml.setProperty("comment_data_max", "Defined only for 8bpp mode - value, corresponding to +127 (-128 is NaN)");

		imp_ml.setProperty("comment_disparityRadiusMain", "Side of the square where 4 main camera subcameras are located (mm)");
		imp_ml.setProperty("comment_disparityRadiusAux", "Side of the square where 4 main camera subcameras are located (mm). Disparity is specified for the main camera");

Admin message