// Last edited on 2013-04-06 19:04:19 by stolfilocal

package main;

import java.awt.image.BufferedImage;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileReader;
import java.io.IOException;
import java.io.LineNumberReader;
import java.io.ObjectInputStream;
import java.util.ArrayList;
import java.util.LinkedList;

import javax.imageio.ImageIO;

import parameters.detection_arguments;
import parameters.io_arguments;
import preprocessing.Parser;
import preprocessing.Util;
import regrouping.Groupment;
import segmentation.Labels;
import segmentation.Regions;
import segmentation.Segmentation;
import classification.Classify;
import classification.Features;
import classification.SVM;
import classification.SVM_DATA;
import fr.lip6.classifier.Classifier;
import fr.lip6.classifier.DoublePegasosSVM;
import fr.lip6.type.TrainingSample;

public class DetectionThread extends Thread {
	
	/*Letters recognized by this thread, magnified for the scale {param.getScale()}*/
	Regions[] regions; 
	
	/*List of characters chains recognized by this thread in one image scale*/
	LinkedList<regrouping.Box> list; 
	
	/*Image to be analyzed*/
	BufferedImage input_image;
	
	/*Flag to control if the {input_image} is negated or not*/
	boolean inversion;
	
	/*Flag that stores the image pyramid level that is being processed*/
	int pyramid_level;
	
	/*Detection parameters used in the segmentation, classification and regrouping steps*/
	detection_arguments parameters; 
	
	ArrayList<SVM> svm_descriptors;
	
	ArrayList<SVM_DATA> svm_classifiers;
	
	/*Flag that controls the end of the thread*/
	boolean finished = false;
	
	/**/
	public Labels Labeling (int w, int h, int[] image) {
		
		Labels labels = new Labels (w, h);
		
		int number_of_labels = labels.create_labels_from_image (w, h, image);
		
		labels.setNumberOfRegions(number_of_labels);
	
        return labels;
	}
	
	void Write_Label_Image (Labels labels, String filename, int pyramid_level, detection_arguments detection_parameters) {

		BufferedImage image = labels.Get_Image_From_Labels ();

		if (detection_parameters.write_image) {
			try {
				ImageIO.write(image, "png",  new File(Util.make_name (filename, inversion, pyramid_level, ".png")));
			}
			catch (Exception e) {
				System.err.println("error: fail to save " + filename + " image");
			}
		}
	}
	
	/* Magnifies each region of {this.reg} from {level_current} to {level_base}.
	 * Than merges the elements of letter list {this.reg} which have {dec < 0} with {reg_base}. 
	 * Assumes that the regions in {reg_base} already magnified for {level_base}.
	 *If two boxes overlap significantly retains only the one with better {dec} score. */
	public ArrayList<Regions> Merge_Letter_List (int level_base, ArrayList<Regions> regions_base, int level_current) {
		
		ArrayList<Regions> regions_end = new ArrayList<Regions>();
		
		assert(level_current == pyramid_level);
		
		for (int i = 0; i < regions.length; i++) {
			Regions.rescale_size (regions[i], level_current, level_base);
		}
		
		/*Take every element of {reg} with {dec < 0} that is not strictly dominated by an element of {reg_base}*/
		for (int i = 0; i < regions.length; i++) {
			Regions regionsi = regions[i];
			if (regionsi.getDec() >= 0) {continue;}
			boolean dominated = false;
			/*for (int j = 0; j < regions_base.size(); j++) {
				Regions regionsj = regions_base.get(j);
				if ( (Regions.overlap(regionsj, regionsi) > 0.3) && (Regions.dominates(regionsj, regionsi)) ) {
					dominated = true;
				}
			}*/
			if (!dominated) {
				regions_end.add(regionsi);
			}
		}
		
		/*Take every element of {reg_base} that strictly dominates every element of {reg} that has {dec < 0} and overlaps it*/
		for (int i = 0; i < regions_base.size(); i++) {
			Regions regi = regions_base.get(i);
			//assert(regi.getDec() < 0);
			boolean dominates_all = true;
			/*for (int j = 0; j < regions.length; j++) {
				Regions regj = regions[j];
				if (regj.getDec() >= 0) {continue;}
				if ( (Regions.overlap(regj, regi) > 0.3) && (!Regions.dominates(regi, regj)) ) {
					dominates_all = false;
				}
			}*/
			if (dominates_all) {
				regions_end.add(regi);
			}
		}
		return regions_end;
	}
	
	public BufferedImage Segmentation (BufferedImage image, boolean inversion, int pyramid_level, detection_arguments detection_parameters) {
		 
		int w = image.getWidth(null);
		
		int h = image.getHeight(null);
		
		int size = w * h;
		
    	int[] min = new int[size];
		
		int[] max = new int[size];
		
		int[] f_h1 = new int[size];
		
		int[] f_h2 = new int[size];
		
		Util util = new Util();
		
		int[] array = util.getImageArray2 (image);

		Segmentation segmentation = new Segmentation();
		
		segmentation.Min_Max (w, h, array, min, max, detection_parameters.mask_size);
		
		segmentation.Double_Toggle_Mapping (w, h, array, min, max, f_h1, f_h2, detection_parameters);
		
		BufferedImage out = Util.getImage (w, h, f_h1);
	
		if (detection_parameters.write_image) {
	    	
	    	Util.writeArrayToPGMImage (w, h, min, Util.make_name ("min", inversion, pyramid_level, ".pgm"));
	    	
	    	Util.writeArrayToPGMImage (w, h, max, Util.make_name ("max", inversion, pyramid_level, ".pgm"));
	    	
	    	Util.writeArrayToPGMImage (w, h, f_h1, Util.make_name ("thresholdedA", inversion, pyramid_level, ".pgm"));
	    	
	    	Util.writeArrayToPGMImage (w, h, f_h2, Util.make_name ("thresholdedB", inversion, pyramid_level, ".pgm"));
  		}
		return out;
    } 

	
	Labels Refining_and_Labeling (BufferedImage image_segmented, int pyramid_level, detection_arguments detection_parameters) {

		int w = image_segmented.getWidth(null);
		
		int h = image_segmented.getHeight(null);
		
		Util util = new Util();
		
		int[] isegmented = util.getImageArray2(image_segmented);
		
		Labels labels = Labeling (w, h, isegmented);

		Write_Label_Image  (labels, "labels_raw", pyramid_level, detection_parameters);
		
		Segmentation segmentation = new Segmentation();
		
		segmentation.Remove_Unknown_Value (w, h, isegmented, labels, pyramid_level, detection_parameters);
	
		if (detection_parameters.write_image) {
			Util.writePGMImage(image_segmented, Util.make_name ("thresholded_wun", inversion, pyramid_level, ".pgm"));
		}
		
		labels = Labeling (w, h, isegmented);

		Write_Label_Image  (labels, "labels_wun", pyramid_level, detection_parameters);
		
		segmentation.Size_Check (w, h, isegmented, labels, pyramid_level, detection_parameters);
		
		labels = Labeling (w, h, isegmented);
		
		Write_Label_Image  (labels, "labels_refined", pyramid_level, detection_parameters); 

		regions = segmentation.Analyze_Regions (w, h, isegmented, labels, pyramid_level);
		
		return labels;
	}
	
	public void Classification2 (
			Labels labels, 
			boolean inversion, 
			detection_arguments detection_parameters ) {
			
		long start;	
		
		double fourier[] = new double[labels.getNumberOfRegions()-1];
		double polar[] = new double[labels.getNumberOfRegions()-1];
		double zernike[] = new double[labels.getNumberOfRegions()-1];
		
		boolean normalization = false;
		
		/*Fourier classification*/
		if (detection_parameters.descriptors[0]) {

			start = System.currentTimeMillis(); 
			
			Features fourier_descriptor = new Features();

			SVM_DATA fourier_obj = svm_classifiers.get(0);

			ArrayList<TrainingSample<double[]>> list = fourier_descriptor.Extracting_Fourier2 (labels, regions, fourier_obj.mean, fourier_obj.deviation, normalization, parameters);

			for(int i = 1 ; i < labels.getNumberOfRegions(); i++) {

				TrainingSample<double[]> e = list.get(i-1);

				/*Score found to each segmented image region*/
				double v = fourier_obj.cls.valueOf(e.sample);
				
				fourier[i-1] = v;

				regions[i].setDec(-v);
			}
			
			Util.Compute_Processing_Time (start, "Fourier SVM Classification");
			
		}
		/*Polar classification*/
		if (detection_parameters.descriptors[1]) {
			
			start = System.currentTimeMillis(); 

			Features polar_descriptor = new Features();

			SVM_DATA polar_obj = svm_classifiers.get(1);

			ArrayList<TrainingSample<double[]>> list = polar_descriptor.Extracting_Polar (labels, regions, polar_obj.mean, polar_obj.deviation, normalization, parameters);

			//System.err.printf("Polar List size : %d\n", list.size());
			
			for(int i = 1 ; i < labels.getNumberOfRegions(); i++) {

				TrainingSample<double[]> e = list.get(i-1);

				/*Score found to each segmented image region*/
				double v = polar_obj.cls.valueOf(e.sample);
				
				polar[i-1] = v;

				regions[i].setDec(-v);
			}
			
			Util.Compute_Processing_Time (start, "Polar SVM Classification");
		}
		/*Zernike classification*/
		if (detection_parameters.descriptors[2]) {
			
			start = System.currentTimeMillis(); 

			Features zernike_descriptor = new Features();

			SVM_DATA zernike_obj = svm_classifiers.get(2);

			ArrayList<TrainingSample<double[]>> list = zernike_descriptor.Extracting_Zernike (labels, regions, zernike_obj.mean, zernike_obj.deviation, normalization, parameters);

			//System.err.printf("Zernike List size : %d\n", list.size());
			
			for(int i = 1 ; i < labels.getNumberOfRegions(); i++) {

				TrainingSample<double[]> e = list.get(i-1);

				/*Score found to each segmented image region*/
				double v = zernike_obj.cls.valueOf(e.sample);
				
				zernike[i-1] = v;

				regions[i].setDec(-v);
			}
			
			Util.Compute_Processing_Time (start, "Zernike SVM Classification");
		}
		/*Combining all descriptors*/
		if (detection_parameters.descriptors[3]) {
			
			start = System.currentTimeMillis(); 

			/*for (int i = 0; i < detection_parameters.descriptors.length; i++) { 
				if (!detection_parameters.descriptors[i]) {
					System.err.printf("Is not possible to compute the combination of the descriptors some of them are false");
					System.exit(1);
				}
			}*/
			
			Features all_descriptors = new Features();

			SVM_DATA all_obj = svm_classifiers.get(3);

			ArrayList<TrainingSample<double[]>> list = all_descriptors.Combining_All (fourier, polar, zernike, all_obj.mean, all_obj.deviation, normalization, parameters); 

			//System.err.printf("All List size : %d\n", list.size());
			
			for(int i = 1 ; i < labels.getNumberOfRegions(); i++) {

				TrainingSample<double[]> e = list.get(i-1);

				/*Score found to each segmented image region*/
				double v = all_obj.cls.valueOf(e.sample);

				regions[i].setDec(-v);
			}
			Util.Compute_Processing_Time (start, "All SVM Classification");
		}
		
	}
	
	
	/*Given of result of the segmentation {lab_contour}, filtered by size, for one pyramid level {param.getScale()}
	 *applies the letter recognizer and grouping algorithm. The recognized letter boxes are left in {this.reg} and
	 *recognized text boxes are left in {this.list} both magnified fir the current pyramid level. */
	public void Classification1 (
			Labels labels, 
			boolean inversion, 
			detection_arguments detection_parameters ) {
		
		SVM fourier = svm_descriptors.get(0);
		SVM polar = svm_descriptors.get(1);
		SVM zernike = svm_descriptors.get(2);
		SVM all = svm_descriptors.get(3);
		
		long start;		
		
		Classify fourier_scores = new Classify();
		
		Classify zernike_scores = new Classify(); 
		
		Classify polar_scores = new Classify();

		/*Fourier classification*/
		if (detection_parameters.descriptors[0]) {
			
			start = System.currentTimeMillis(); 

			Features fourier_descriptor = new Features();

			fourier_descriptor.Extracting_Fourier (labels, regions, fourier.Get_Sample_Size(), detection_parameters); 

			fourier_scores.classify_rbf (fourier, fourier_descriptor);

			Util.Compute_Processing_Time (start, "Fourier Classification");
			
			for(int i = 1; i < labels.getNumberOfRegions(); i++) {
				regions[i].setDec(fourier_scores.Get_Score_Value(i-1));
				//System.out.printf("Score svm : %f\n", fourier_scores.Get_Score_Value(i-1));
			}
		}

		/*Zernike classification*/
		if (detection_parameters.descriptors[1]) {
			
			start = System.currentTimeMillis();

			Features zernike_descriptor = new Features();

			zernike_descriptor.Extracting_Zernike (labels, regions, zernike.Get_Sample_Size(), detection_parameters);

			zernike_scores.classify_rbf (zernike, zernike_descriptor);

			Util.Compute_Processing_Time (start, "Zernike Classification");

			for(int i = 1; i < labels.getNumberOfRegions(); i++) {
				regions[i].setDec(zernike_scores.Get_Score_Value(i-1));
			}
		}
		
		/*Polar classification*/
		if (detection_parameters.descriptors[2]) {
			
			start = System.currentTimeMillis();

			Features polar_descriptor = new Features();

			polar_descriptor.Extracting_Polar (labels, regions, polar.Get_Sample_Size(), detection_parameters);

			polar_scores.classify_rbf (polar, polar_descriptor);

			Util.Compute_Processing_Time (start, "Polar Classification");

			for(int i = 1; i < labels.getNumberOfRegions(); i++) {
				regions[i].setDec(polar_scores.Get_Score_Value(i-1));
			}
		}
		
		/*Combining all descriptors*/			
		if (detection_parameters.descriptors[3]) {
			
			for (int i = 0; i < detection_parameters.descriptors.length; i++) { 
				if (!detection_parameters.descriptors[i]) {
					System.err.printf("Is not possible to compute the combination of the descriptors some of them are false");
					System.exit(1);
				}
			}
			
			start = System.currentTimeMillis();

			Features all_descriptors = new Features();

			all_descriptors.Combining_All (fourier_scores, zernike_scores, polar_scores, detection_parameters);

			Classify final_scores = new Classify(); 

			final_scores.classify_rbf (all, all_descriptors);

			Util.Compute_Processing_Time (start, "Final Classification");

			/*Marking each region with the end-score of the classification*/

			for(int i = 1; i < labels.getNumberOfRegions(); i++) {
				regions[i].setDec(final_scores.Get_Score_Value(i-1));
			}
		}
	}
	
	void Compute_And_Draw_Regions_Barycenters (
			BufferedImage image, 
			BufferedImage image_segmented,
			Labels labels, 
			Groupment groupment,
			detection_arguments detection_parameters) {
		
		long start = System.currentTimeMillis();;

		int w = image.getWidth();
		
		int h = image.getHeight();
		
		int size = w * h;

		Util util = new Util();
		
		int[] array = util.getImageArray2 (image);

		int[] out = new int[size];

		for (int i = 0; i < size; i++) {
			
			int label = labels.getValue(i);
			
			double score = regions[label].getDec();
			
			if ( (label != 0) && (score < 0) ) {
				
				out[i] = Util.getPixelFromRGB (0, Util.Value(score), 0);
				
			} else if ( (label != 0) && (score > 0) ) {
				
				out[i] = Util.getPixelFromRGB (Util.Value(score), 0, 0);
				
			} else {
				if (inversion) {
					out[i] = Util.getPixelFromRGB (255-array[i], 255-array[i], 255-array[i]);
				} else {
					out[i] = Util.getPixelFromRGB (array[i], array[i], array[i]);
				}
			}
		}

		int startX = 0, startY = 0;

		BufferedImage outimg = new BufferedImage (w, h, BufferedImage.TYPE_INT_ARGB);

		outimg.setRGB (startX, startY, w, h, out, 0, w);

		groupment.Dispose_Big_Regions_Agroupments (outimg, list, detection_parameters);
		
		groupment.Dispose_Region_And_Barycenters (outimg, regions, labels.getNumberOfRegions());
		
		groupment.Dispose_Region_And_Barycenters (image_segmented, regions, labels.getNumberOfRegions());
		
		try {
			BufferedImage ilabels = labels.Get_Image_From_Labels (regions);
			groupment.Dispose_Region_And_Barycenters (ilabels, regions, labels.getNumberOfRegions());
			ImageIO.write(ilabels, "png", new File(Util.make_name ("recognized_letters_labels_orig", inversion, pyramid_level, ".png")));
			ImageIO.write(outimg, "png", new File(Util.make_name ("recognized_letters_orig", inversion, pyramid_level, ".png")));
			Util.writePGMImage(image_segmented, Util.make_name ("recognized_letters_thrs", inversion, pyramid_level, ".pgm"));
		}
		catch (Exception e) {
			System.err.println("cannot store image");
		}
		
		Util.Compute_Processing_Time (start, "Drawing barycenters");
	}
		
	void Regions_Groupment ( 
			BufferedImage image, 
			BufferedImage image_segmented,
			Labels labels, 
			detection_arguments detection_parameters) { 
		
		long start;
		
		start = System.currentTimeMillis();
				
		Groupment groupment = new Groupment();

		list = groupment.Grouping_Regions (labels, regions, detection_parameters); 
		
		if (parameters.write_image) {
			Compute_And_Draw_Regions_Barycenters (image, image_segmented, labels, groupment, detection_parameters);
		}
		
		Util.Compute_Processing_Time (start, "Regions_Groupment");
	}
	
	public DetectionThread ( 
			BufferedImage input_image, 
			boolean inversion,
			int pyramid_level,
			ArrayList<SVM> svm_descriptors,
			ArrayList<SVM_DATA> svm_classifiers,
			detection_arguments detection_parameters) 
	{
		this.input_image = input_image;
		this.inversion = inversion;
		this.pyramid_level = pyramid_level;
		this.parameters = detection_parameters;
		if (parameters.svm_option) {
			this.svm_descriptors = svm_descriptors;
		}
		else {
			this.svm_classifiers = svm_classifiers;
		}
	}
		
	public void run ()
	{ 
		long start = System.currentTimeMillis();
		
		BufferedImage image_segmented = Segmentation (input_image, inversion, pyramid_level, parameters);
		
		Labels labels = Refining_and_Labeling (image_segmented, pyramid_level, parameters);
		
		if (parameters.svm_option) {
			Classification1 (labels, inversion, parameters);
		}
		else {
			Classification2 (labels, inversion, parameters);
		}
		
		Regions_Groupment (input_image, image_segmented, labels, parameters);
		
		finished = true;

		Util.Compute_Processing_Time (start, "Segmentation, Classification and Grouping");
		
		Thread.yield();
	}
	
	public boolean Has_Finished()
	{
		return finished;
	}
	
	public LinkedList<regrouping.Box> Get_Text_List()
	{
		return list;
	}
	
}