updated normalization routine in the matchTemplate to avoid division by zero on black images (ticket #798), added test

Added print of IPP info in test log files

added OpenCVFindIPP.cmake script, which will look for IPP installation at CMake configuration time. First, IPPROOT environment variable will be tested, if not found script will look at default install places.
The script should support IPP from 5.3 up to 7.x versions (although tested on Windows for IPP 6.1 and IPP 7.0 versions only)

Preliminary optimization of HOG with IPP added too. Not yet quite efficient, code for cpu branch should be redesigned in order to have better performance.

added OpenCVFindIPP.cmake script, which will look for IPP installation at CMake configuration time. First, IPPROOT environment variable will be tested, if not found script will look at default install places.
The script should support IPP from 5.3 up to 7.x versions (although tested on Windows for IPP 6.1 and IPP 7.0 versions only)

Preliminary optimization of HOG with IPP added too. Not yet quite efficient, code for cpu branch should be redesigned in order to have better performance.

fixed incorrect sign of the result of the convolution with normalized asymmetric kernels (ticket #779)

android-opencv.mk file.

simplified gpu::columnSum test, it doesn't fail on Quadro anymore (when seed is 000001af5a11badd) after BFM test, but something definitely wrong with NPP_Staging's transpose