Impala Documentation

std::vector<Array::Array2dScalarReal64*> Impala::Core::Feature::HaarResponse ( Array::Array2dScalarReal64 * image )

returns a vector of four images:

haar esponse in x direction
haar response in y direction
abs of haar esponse in x direction
abs of haar response in y direction

Please note that because the haar response basically soes a convolution with a kernel of size 2 and we don't have a border, the resulting images are 1 pixel smaller in both directions than the source image. We construct both x and y response at the same time in the functions HaarFirstLine and HaarSecondLine. Because we need to read two lines in the source image to write one line in the destination image HaarFirstLine initialises the pixels in the destination image. HaarSecondLine modifies the values set by HaarFirstLine.

Definition at line 75 of file HaarResponse.h.

References Impala::Core::Array::Abs(), Impala::Core::Array::Array2dTem< StorT, elemSize, ArithT >::H(), HaarFirstLine(), HaarSecondLine(), and Impala::Core::Array::Array2dTem< StorT, elemSize, ArithT >::W().

Referenced by DSurf(), and Impala::Core::Feature::TestFunctions::testHaar().

00076 {
00077     std::vector<Array::Array2dScalarReal64*> response;
00078     int w = image->W()-1;
00079     int h = image->H()-1;
00080     response.push_back(new Array::Array2dScalarReal64(w, h, 0, 0));
00081     response.push_back(new Array::Array2dScalarReal64(w, h, 0, 0));
00082     for(int y=0 ; y<image->H()-1 ; ++y)
00083     {
00084         HaarFirstLine(response, image, y);
00085         HaarSecondLine(response, image, y+1);
00086     }
00087     response.push_back(new Array::Array2dScalarReal64(w,h,0,0));
00088     response.push_back(new Array::Array2dScalarReal64(w,h,0,0));
00089     Abs(response[2],response[0]);
00090     Abs(response[3],response[1]);
00091     return response;
00092 }

Here is the call graph for this function: