First run of TD neighbors

src/main/java/com/elphel/imagej/gpu/GpuQuad.java

@@ -126,6 +126,7 @@ public class GpuQuad{ // quad camera description
     private int sensor_mask_inter = -1;
     private int [] corr_mask_indices = null;
     private int [] sensor_mask_inter_indices = null;
+    private int [] sensor_mask_inter_indices_neibs = null;
     
     
 	public boolean [] getCorrMask() {
@@ -513,13 +514,17 @@ public class GpuQuad{ // quad camera description
 	public void setSensorMaskInter(int mask) {
 		sensor_mask_inter = mask & ((1 << num_cams) -1);
 		sensor_mask_inter_indices = new int [getNumCamsInter()+1];
+	    sensor_mask_inter_indices_neibs = new int [getNumCamsInter()+2];
+
 		int indx = 0;
 		for (int i = 0; i < num_cams; i++) {
 			if ((sensor_mask_inter & (1 << (i & 31))) != 0) {
 				sensor_mask_inter_indices[indx++] = i;
 			}
 		}
-		sensor_mask_inter_indices[indx++] = 0xff; // sum
+		sensor_mask_inter_indices[indx] = 0xff; // sum
+		sensor_mask_inter_indices_neibs[indx++] = 0xff; // sum
+		sensor_mask_inter_indices_neibs[indx++] = 0xfe; // sumof neibs
 	}
 	public int getNumCamsInter() {
 		return Integer.bitCount(sensor_mask_inter);
@@ -530,7 +535,24 @@ public class GpuQuad{ // quad camera description
 	public int [] getSensInter() {
 		return sensor_mask_inter_indices;
 	}
-
+	/*
+	public int getSensInterNeib(int indx) { // not used
+		return sensor_mask_inter_indices_neibs[indx];
+	}
+	*/
+	public int [] getSensInterNeib() {
+		return sensor_mask_inter_indices_neibs;
+	}
+	public int [] getSensInterNeib(boolean full) {
+		return full? sensor_mask_inter_indices_neibs : (new int [] {
+				sensor_mask_inter_indices_neibs[sensor_mask_inter_indices_neibs.length -2],
+				sensor_mask_inter_indices_neibs[sensor_mask_inter_indices_neibs.length -1]});
+	}
+	/*
+	public int getSensInterNeib(int indx, boolean full) {
+		return full? sensor_mask_inter_indices_neibs[indx] : getSensInterNeib(false)[indx];
+	}
+    */
 	
 	public void setGeometryCorrection() { // will reset geometry_correction_set when running GPU kernel
 		//    		if (geometry_correction_set) return;
@@ -3029,6 +3051,37 @@ public class GpuQuad{ // quad camera description
 	}
 
 
+	public float [][][] getCorrTilesLayerTD(
+			int []         indices,
+			float []       fdata,
+			boolean        inter,
+			int            pair) {
+		int tilesX =     img_width / GPUTileProcessor.DTT_SIZE;
+		int tilesY =     img_height / GPUTileProcessor.DTT_SIZE;
+        int num_pairs = inter? (getNumCamsInter() + 1 + 0): getNumUsedPairs();// Number of used pairs		num_pairs = num_pairs_in;
+		final int corr_size_td = 4 * GPUTileProcessor.DTT_SIZE * GPUTileProcessor.DTT_SIZE;
+//		int [] indices = getCorrIndices(); // also sets num_corr_tiles
+//		float [] fdata = getCorrTdData();
+		int num_tiles = num_corr_tiles / num_pairs;
+		float [][][] corr_tiles = new float [tilesY][tilesX][];
+
+		for (int nt = 0; nt < num_tiles; nt++ ) {
+			int nTile = (indices[nt * num_pairs] >> GPUTileProcessor.CORR_NTILE_SHIFT);
+			int ty = nTile / tilesX;
+			int tx = nTile % tilesX;
+
+			corr_tiles[ty][tx] = new float [corr_size_td];
+			System.arraycopy(
+					fdata,
+					(nt * num_pairs + pair) * corr_size_td,
+					corr_tiles[ty][tx],
+					0,
+					corr_size_td);
+		}
+		return corr_tiles;
+	}
+	
+	
 
 	public int [] setCorrTilesComboTd( // not used?
 			final float [][][] corr_tiles, // [tileY][tileX][4*64]

src/main/java/com/elphel/imagej/tileprocessor/OpticalFlow.java

@@ -13152,16 +13152,17 @@ public class OpticalFlow {
 		}
 		
 		boolean use_neibs =        clt_parameters.imp.use_neibs;                  // false; // true;
-		
+		boolean use_neibs_pd =     true;		
 		boolean neibs_nofpn_only = clt_parameters.imp.neibs_nofpn_only |
 				(initial_adjust && clt_parameters.imp.neibs_nofpn_init);           // consolidate neighbors fot non-fpn tiles only!
 		boolean redo_both =        clt_parameters.imp.redo_both;                  // use average of neighbors for both pd,td if any of the center tile tests (td, pd) fails
 		int min_num_neibs =        clt_parameters.imp.min_num_neibs;              // plus center, total number >= (min_num_neibs+1)
-		double scale_neibs_pd = use_neibs? clt_parameters.imp.scale_neibs_pd : 0; // scale threshold for the pixel-domain average maximums		
-		double scale_neibs_td = use_neibs? clt_parameters.imp.scale_neibs_td : 0; // scale threshold for the transform-domain average maximums
+		double scale_neibs_pd = use_neibs_pd? clt_parameters.imp.scale_neibs_pd : 0; // scale threshold for the pixel-domain average maximums		
+		double scale_neibs_td = use_neibs_pd? clt_parameters.imp.scale_neibs_td : 0; // scale threshold for the transform-domain average maximums
 		double scale_avg_weight =  clt_parameters.imp.scale_avg_weight;           // reduce influence of the averaged correlations compared to the single-tile ones
 		
 		int [] corr_indices_dbg = show_2d_correlations? image_dtt.getGPU().getCorrIndices() : null;
+		boolean use_partial = clt_parameters.imp.use_partial;
 		coord_motion = image_dtt.clt_process_tl_interscene(       // convert to pixel domain and process correlations already prepared in fcorr_td and/or fcorr_combo_td
 				clt_parameters.img_dtt,            // final ImageDttParameters  imgdtt_params,   // Now just extra correlation parameters, later will include, most others
 				// only used here to keep extra array element for disparity difference
@@ -13184,7 +13185,7 @@ public class OpticalFlow {
 				fpn_ignore_border,                 // final boolean             fpn_ignore_border, // only if fpn_mask != null - ignore tile if maximum touches fpn_mask			
 				motion_vectors,                    // final double [][][]       motion_vectors,  // [tilesY*tilesX][][] -> [][][num_sel_sensors+1][2]
 				clt_parameters.imp.run_poly,       // final boolean             run_poly,        // polynomial max, if false - centroid
-				clt_parameters.imp.use_partial,    // final boolean             use_partial,     // find motion vectors for individual pairs, false - for sum only
+				use_partial,                       // final boolean             use_partial,     // find motion vectors for individual pairs, false - for sum only
 				clt_parameters.imp.centroid_radius,// final double              centroid_radius, // 0 - use all tile, >0 - cosine window around local max
 				clt_parameters.imp.n_recenter,     // final int                 n_recenter,      // when cosine window, re-center window this many times
 				clt_parameters.imp.td_weight,      // final double  td_weight,    // mix correlations accumulated in TD with 
@@ -15131,7 +15132,7 @@ public class OpticalFlow {
 		double disparity_weight = use3D? clt_parameters.ilp.ilma_disparity_weight : 0.0;
 		int        margin = clt_parameters.imp.margin;
 		int        sensor_mask_inter = clt_parameters.imp.sensor_mask_inter ; //-1;
-		float [][][] facc_2d_img = new float [1][][];
+		float [][][] facc_2d_img = new float [1][][]; // set it to null?
 		IntersceneLma intersceneLma = new IntersceneLma(
 				clt_parameters.ilp.ilma_thread_invariant,
 				disparity_weight);