stcrawcosts.cpp

这是一个用于立体匹配的程序。 可以用来进行立体匹配！

///////////////////////////////////////////////////////////////////////////
//
// NAME
//  StcRawCosts.cpp -- compute raw per-pixel matching score between a pair of frames
//
// DESCRIPTION
//  The CStereoMatcher class implements a number of popular stereo
//  correspondence algorithms (currently for 2-frame rectified image pairs).
//
// The definition of disparity is as follows:
//  Disparity is the (floating point) inter-frame (instantaneous) displacement
//  of a pixel between successive frames in a multi-frame stereo pair.
//
//  In building the disparity-space image (DSI) m_cost, we enumerate a set of
//  m_disp_n disparitites, k = 0..m_disp_n-1.
//
//  The mapping between the integral disparities k and the floating point
//  disparity d is given by:
//      d = disp_min + k * m_disp_num / m_disp_den
//
//  Because we may be matching frames that are not adjacent in the sequence
//  we also define the frame difference f_d and scaled disparity s_d as:
//      f_d = frame_match - frame_ref
//      s_d = f_d * d
//  The coordinate of a pixel x_m in the matching frame corresponding
//  to a reference pixel x_r in the reference frame is given by
//      x_m = x_r - s_d
//  (this is because input images are ordered left to right, so that pixel
//  motion is leftward).
//
//  When we store the floating point disparities in a gray_level image, we
//  use the formulas
//      g_d = (d - disp_min) * disp_scale
//        d =  disp_min  + g_d / disp_scale
//  to do the conversions.
//
// IMPLEMENTATION
//  For better efficiency, we first interpolate the matching line
//  up by a factor of m_disp_den.
//
// SEE ALSO
//  StereoMatcher.h         longer description of this class
//
// Copyright