Logo Search packages:      
Sourcecode: qtpfsgui version File versions

tmo_pattanaik00.cpp

Go to the documentation of this file.
/**
 * @file tmo_pattanaik00.cpp
 * @brief Time-dependent Visual Adaptation Model, [Pattanaik2000]
 *
 * Time-Dependent Visual Adaptation for Realistic Image Display
 * S.N. Pattanaik, J. Tumblin, H. Yee, and D.P. Greenberg
 * In Proceedings of ACM SIGGRAPH 2000
 *
 * 
 * This file is a part of Qtpfsgui package.
 * ---------------------------------------------------------------------- 
 * Copyright (C) 2003,2004 Grzegorz Krawczyk
 * 
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 * ---------------------------------------------------------------------- 
 * 
 * @author Grzegorz Krawczyk, <krawczyk@mpi-sb.mpg.de>
 * @author Giuseppe Rota <grota@users.sourceforge.net>
 *
 * $Id: tmo_pattanaik00.cpp,v 1.5 2004/12/14 15:34:04 krawczyk Exp $
 */

#include <math.h>
#include "../../Libpfs/pfs.h"
#include "tmo_pattanaik00.h"


/// sensitivity of human visual system
00041 float n = 0.73f;

void calculateLocalAdaptation(pfs::Array2D* Y, int x, int y, float& Acone, float& Arod);
float sigma_response_rod(float I);
float sigma_response_cone(float I);
float model_response(float I, float sigma);



// tone mapping operator code
00051 void tmo_pattanaik00( pfs::Array2D* R, pfs::Array2D* G, pfs::Array2D* B, 
  pfs::Array2D* Y, VisualAdaptationModel* am, bool local )
{  
  ///--- initialization of parameters

  /// cones level of adaptation
  float Acone = am->getAcone();
  /// rods level of adaptation
  float Arod = am->getArod();
  /// cones bleaching term
  float Bcone = am->getBcone();
  /// rods bleaching term
  float Brod = am->getBrod();

  /// fraction of adaptation luminance that represents white luminance
  float white_factor = 5.0f;
  /// fraction of adaptation luminance that represents dark luminance
  float dark_factor = 32.0f/5.0f;

  /// goal adaptation luminance while watching display
  float G_display=25.0f;
  /// display luminance of white
  float display_white = G_display*white_factor;
  /// display luminance of dark
  float display_dark = G_display/dark_factor;
  /// half-saturation constant for display
  float display_sigma = sigma_response_cone(G_display);

  /// reference white for display
  float REF_Wht_display = model_response(display_white, display_sigma);
  /// reference black for display
  float REF_Blk_display = model_response(display_dark, display_sigma);

  /// display color saturation
  float S_d = (REF_Wht_display - REF_Blk_display)
    / (log10(display_white)-log10(display_dark));

  ///--- precalculated parameters

  /// half-saturation constant for cones
  float sigma_cone = sigma_response_cone(Acone);
  /// half-saturation constant for rods
  float sigma_rod = sigma_response_rod(Arod);

  // appearance model
  float scene_white_cone = Acone*white_factor;
  float scene_dark_cone = Acone/dark_factor;
  float scene_white_rod = Arod*white_factor;
  float scene_dark_rod = Arod/dark_factor;

  float REF_Wht_scene = Bcone*model_response(scene_white_cone, sigma_cone)
    + Brod*model_response(scene_white_rod, sigma_rod);

  float REF_Blk_scene = Bcone*model_response(scene_dark_cone, sigma_cone)
    + Brod*model_response(scene_dark_rod, sigma_rod);

  /// scene luminance minus shift
  float disp_x = 0.0f;
  /// scene luminance to display luminance scale factor
  float disp_y = 1.0f;
  /// display luminance plus shift
  float disp_z = 0.0f;

  float scale = (REF_Wht_display-REF_Blk_display)
    / (REF_Wht_scene-REF_Blk_scene);

  if( scale < 1.0f )
  {
    if( REF_Wht_scene>REF_Wht_display )
    {
      disp_x = REF_Wht_scene;
      disp_y = scale;
      disp_z = REF_Wht_display;
    }
    else if( REF_Blk_scene<REF_Blk_display )
    {
      disp_x = REF_Blk_scene;
      disp_y = scale;
      disp_z = REF_Blk_display;
    }
  }
  else if(scale > 1.0f )
  {
    if( REF_Wht_scene>REF_Wht_display )
    {
      disp_x = REF_Wht_scene;
      disp_y = 1.0f;
      disp_z = REF_Wht_display;
    }
    else if( REF_Blk_scene<REF_Blk_display )
    {
      disp_x = REF_Blk_scene;
      disp_y = 1.0f;
      disp_z = REF_Blk_display;
    }
  }

  ///--- tone map image

  int im_width = Y->getCols();
  int im_height = Y->getRows();
  for( int x=0 ; x<im_width ; x++ )
    for( int y=0 ; y<im_height ; y++ )
    {
      float l = (*Y)(x,y);
      float r = (*R)(x,y)/l;
      float g = (*G)(x,y)/l;
      float b = (*B)(x,y)/l;

      if( local )
      {
        calculateLocalAdaptation(Y,x,y,Acone,Arod);
        Bcone = 2e6/(2e6+Acone);
        Brod = 0.04f/(0.04f+Arod);
        sigma_cone = sigma_response_cone(Acone);
        sigma_rod = sigma_response_rod(Arod);
      }

      // receptor responses
      float Rrod = Brod*model_response(l, sigma_rod );
      float Rcone = Bcone*model_response(l, sigma_cone );
      float Rlum = Rrod + Rcone;
      if( Rlum>0.0f )
      {
      Rrod /= Rlum;
      Rcone /= Rlum;
      }

      float Scolor = (Bcone*pow(sigma_cone,n)*n*pow(l,n))
      / pow( pow(l,n)+pow(sigma_cone,n), 2 );
      Scolor /= S_d;

      // appearance model
      float Ra = (Rlum - disp_x)*disp_y+disp_z;
      Ra = (Ra<1.0f) ? ((Ra>0.0f) ? Ra : 0.0f ) : 0.9999999f;

      // inverse display model
      float I = display_sigma * pow(Ra/(1.0f-Ra), 1.0f/n) / display_white;

      // apply new luminance
      r = pow( r, Scolor )*I*Rcone + I*Rrod;
      g = pow( g, Scolor )*I*Rcone + I*Rrod;
      b = pow( b, Scolor )*I*Rcone + I*Rrod;

      (*R)(x,y) = (r<1.0f) ? ((r>0.0f) ? r : 0.0f) : 1.0f;
      (*G)(x,y) = (g<1.0f) ? ((g>0.0f) ? g : 0.0f) : 1.0f;
      (*B)(x,y) = (b<1.0f) ? ((b>0.0f) ? b : 0.0f) : 1.0f;
    }

}

///////////////////////////////////////////////////////////

00204 float sigma_response_rod(float I)
{
//!! INFO: this is essentially the same as below (formulas come from
// R.W.G.Hunt book)
//  float j = 0.00001f / (5.0f*I + 0.00001f);
//   float fls = 3800*pow(j,2.0f)*5*I+0.2*pow(1-pow(j,2.0f),4.0f)*pow(5*I,1.0f/6.0f);
//   float sigma_rod = pow(2,1.0f/n) / fls * I;
  
  float j = 1.0f / (5*1e4*I+1);
  float j2 = j*j;
  
  float sigma_rod = (2.5874f*I)/(19000.0f*j2*I+0.2615f*pow(1.0f-j2,4)*pow(I,1.0f/6.0f));
  return sigma_rod;
}

00219 float sigma_response_cone(float I)
{
//!! INFO: this is essentially the same as below (formulas come from
// R.W.G.Hunt book)
//   float k = 1.0f/(5.0f*I+1);
//   float fl = 0.2f*5.0f*pow(k,4.0f)*I
//     + 0.1f*pow(1.0f-pow(k,4.0f),2.0f)*pow(5*I,1.0f/3.0f);
//   float sigma_cone = pow(2.0f,1.0f/n) / fl * (5.0f*I);
  
  float k = 1.0f/(5.0f*I+1);
  float k4 = pow(k,4.0f);
  float sigma_cone = (12.9223f*I) / ( k4*I+0.171 * pow(1.0f-k4,2) * powf(I,1.0f/3.0f));
  return sigma_cone;
}

float model_response(float I, float sigma)
{
  return pow(I,n)/(pow(I,n)+pow(sigma,n));
}


///////////////////////////////////////////////////////////

/**
 * @brief Calculate local adaptation for specified pixel coordinates
 *
 * Calculation based on article "Adaptive Gain Control" by Pattanaik
 * 2002.
 *
 * @param Y luminance map
 * @param x x-coordinate of pixel
 * @param y x-coordinate of pixel
 * @param Acone [out] calculated adaptation for cones
 * @param Arod [out] calculated adaptation for rods
 */
00254 void calculateLocalAdaptation(pfs::Array2D* Y, int x, int y, float& Acone, float& Arod)
{
  int width = Y->getCols();
  int height = Y->getRows();
  
  int kernel_size = 4;
  static float LOG5 = log(5);
  float logLc = log((*Y)(x,y))/LOG5;

  float pix_num = 0.0;
  float pix_sum = 0.0;
  for( int kx = -kernel_size ; kx <= kernel_size ; kx++ )
    for( int ky = -kernel_size ; ky <= kernel_size ; ky++ )
      if( (kx*kx+ky*ky)<=(kernel_size*kernel_size) &&
        (x+kx)>0 && (x+kx)<width && (y+ky)>0 && (y+ky)<height )
      {
      float L = (*Y)(x+kx,y+ky);
      float w = exp( -pow(fabs(log(L)/LOG5-logLc),6.0) );
      pix_sum +=  w*L;
      pix_num +=  w;
      }

  if( pix_num>0.0 )
    Acone = Arod = (pix_sum / pix_num);
  else
    Acone = Arod = (*Y)(x,y);
}



///////////////////////////////////////////////////////////

00286 VisualAdaptationModel::VisualAdaptationModel()
{
  setAdaptation(60.0f, 60.0f);
}

00291 void VisualAdaptationModel::calculateAdaptation(float Gcone, float Grod, float dt)
{
  // Visual adaptation model for cones
  const float t0cone=0.080f;
  const float t0rod=0.150f;
  const float tau_cone=110.0f;
  const float tau_rod=400.0f;

  float in,out;
  float delta_out;
  float f,j,k;

  // Calculation of Acone & Arod
  // Acone
  in = Gcone;
  out = Acone;
  delta_out= in-out;
  f = (1.0f - exp( -dt/t0cone ) );
  Acone += f*delta_out;

  // Arod
  in = Grod;
  out = Arod;
  delta_out= in-out;
  f = (1.0f - exp( -dt/t0rod ) );
  Arod += f*delta_out;

  // Calculation of Bcone & Brod
  // Bcone
  in = Gcone;
  out = Bcone;
  j = dt * in * out / 2.2e8;
  k = dt * (1.0f - out ) / tau_cone;
  f = k-j;
  Bcone += f;

  // Brod
  in = Grod;
  out = Brod;
  j = dt * in * out / 16;
  k = dt * (1.0f - out ) / tau_rod;
  f = k-j;
  Brod += f;
}

00336 void VisualAdaptationModel::calculateAdaptation(pfs::Array2D *Y, float dt)
{
  float Acone = calculateLogAvgLuminance(Y);
  calculateAdaptation(Acone, Acone, dt);
}

00342 void VisualAdaptationModel::setAdaptation(float Gcone, float Grod)
{
  Acone = Gcone;
  Arod = Grod;
  Bcone = 2e6/(2e6+Acone);
  Brod = 0.04f/(0.04f+Arod);
}

00350 void VisualAdaptationModel::setAdaptation(pfs::Array2D *Y)
{
  float Acone = calculateLogAvgLuminance(Y) * 5.0f;
  setAdaptation(Acone, Acone);
}

00356 float VisualAdaptationModel::calculateLogAvgLuminance( pfs::Array2D* Y )
{
  float avLum = 0.0f;

  int size = Y->getCols() * Y->getRows();
  for( int i=0 ; i<size; i++ )
  {
    avLum += log( (*Y)(i) + 1e-4 );
  }
  return (exp(avLum/size ) - 1e-4);
}

Generated by  Doxygen 1.6.0   Back to index