package association;

import descriptors.HOG;
import parameters.tracking_arguments;
import particlefiltering.observation;
import particlefiltering.pHistogram;
import tracking.Atributes;

public class Probabilities {
	
	
	public double prob_app (Atributes A1, Atributes A2, tracking_arguments tracking_parameters, int frame, int boxi, int boxj, String image_name)
	{
		int new_height = -1;     
		
        int number_of_cells_x = 1;
		
        int number_of_cells_y = 1;
			
		int bins_per_cell = 9;
		
		boolean image_normalization = false; 
		
		String image_normalization_weight = "Gaussian";
		
		double image_normalization_weight_radius = 5.0;
		
		String gradient_option = "Simple";
		
		String histogram_normalization_metric = "L1";
		
		String weight_function = "Bernstein";
		
		boolean deformable_weights = false;
		
		pHistogram histo1 = new pHistogram(); 

		HOG hog1 = new HOG();
		
		double[] e1 = hog1.hog_descriptor ( 
				A1.image, /*Original image*/
		        new_height,        
		        number_of_cells_x,
		        number_of_cells_y,
		        bins_per_cell,
		        image_normalization,
		        image_normalization_weight, 
		        image_normalization_weight_radius,
		        gradient_option,
		        histogram_normalization_metric,
		        weight_function,
		        deformable_weights );
		
		histo1.alloc_vector(e1.length);
		
		for (int i = 0; i < e1.length; i++) {
		  histo1.set_data(i, e1[i]);
		}
		
		observation obs = new observation();
		
		obs.normalize_histogram (histo1);
		
		pHistogram histo2 = new pHistogram();

		HOG hog2 = new HOG();

		double[] e2 = hog2.hog_descriptor ( 
				A2.image, 
                new_height,        
                number_of_cells_x,
                number_of_cells_y,
                bins_per_cell,
                image_normalization,
                image_normalization_weight, 
                image_normalization_weight_radius,
                gradient_option,
                histogram_normalization_metric,
                weight_function,
                deformable_weights );
		
		
		histo2.alloc_vector(e2.length);
		
		for (int i = 0; i < e2.length; i++) {
		  histo2.set_data(i, e2[i]);
		}
	
		obs.normalize_histogram (histo2);

		/* compute likelihood as e^{\lambda D^2(h, h^*)} */
		double d_sq = obs.histo_dist_sq (histo1, histo2);

		return Math.exp (-tracking_parameters.sigma_app * d_sq);
	}
	
	public double prob_pos (Atributes A1, Atributes A2, tracking_arguments tracking_parameters)
	{	
		double d = Math.sqrt ( ( A2.get_x() - A1.get_x() ) * ( A2.get_x() - A1.get_x() ) + ( A2.get_y() - A1.get_y() ) * ( A2.get_y() - A1.get_y() ) );

	    double sigma = tracking_parameters.sigma_pos; 
	    
	    double prob = Math.exp (-(d*d)/(2*sigma*sigma));

	    return prob;
	}
	
	public double prob_size (Atributes A1, Atributes A2, tracking_arguments tracking_parameters)
	{
		int s1 = A1.get_width() * A1.get_height();
		
		int s2 = A2.get_width() * A2.get_height();
		
		double s = (s1-s2)/s1;
		
		double sigma = tracking_parameters.sigma_size;
	    
	    double prob = Math.exp (-(s*s)/(2*sigma*sigma));

	    return prob;
	}	   
}