hello,

i'm trying to find the correct algorythm to convert a 2d quad source to a final 2d quad using a linear matrix transformation.

My final quad is the screen coordinates and the source quad is the four points of a camera image corners.
I find and implement this function, but this is for the opposite effect. Source is the screen and final is the internal coordinates. this is the algorithm.

Can you help me?

maybe it exist a library to do this. I now it exist in opencv, but i dont know if is possible to implement in processing.

thanks a lot,

alex

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


  //this is the startx and endx of my src quad
  // top left corner and bottom right corner
  float sx = 0.0;
  float sy = 0.0;
  float ex = 1.0;
  float ey = 1.0;
 
  //this is an array of four points
  //that describe the four corners
  //of the warped quad -
  // currently they are ordered like:
  //
  //  pt0      pt1
  //
  //  pt3      pt2
  //
  //but I have also tried doing top-left, top-right, bottom-left, bottom-right
 
  float wx[4];
  float wy[4];
 
  //corners are in 0.0 - 1.0 range
  for(int i = 0; i < 4; i++){
  wx[i] = corners[i].x ;
  wy[i] = corners[i].y ;
  }
 
  //this is code from Myler and Weeks:
  //calculates the warp coeffecients that describe
  //the transformations between the src and dst quad
  float matrix[8];
 
  matrix[0] = (-wx[0] + wx[1]) / (ey-sy); //a
  matrix[1] = (-wx[0] + wx[3]) / (ex-sx); //b
  matrix[2] = (wx[0] - wx[1] + wx[2] - wx[3]) / ( (ey-sy)  * (ex-sx) ); //c
  matrix[3] =  wx[0]; //d
 
  matrix[4] = (-wy[0] + wy[1]) / (ex-sx); //e
  matrix[5] = (-wy[0] + wy[3]) / (ey-sy); //f
  matrix[6] =  (wy[0] - wy[1] + wy[2] - wy[3]) / ( (ey-sy)  * (ex-sx) ); //i
  matrix[7] = wy[0]; //j
 
  //myler and weeks describes then how to translate a new destination X and Y based
  //on these coeffecients
 
  //  X = a*x + b*y +c*x*y + d
  // Y = e*x + f*y + i*x*y + j;