Usage:

import numpy as np
import cv2
img = cv2.imread('messi5.jpg',0)
sift = cv2.SIFT()
kp = sift.detect(img)

pts = cv2.KeyPoint_convert(kp)

kps = cv2.KeyPoint_convert(pts.reshape(-1,1,2))

overlap = cv2.KeyPoint_overlap(kp[0],kp[1])

Conflicts:
	modules/gpuwarping/src/cuda/resize.cu
	modules/gpuwarping/src/resize.cpp
	modules/gpuwarping/test/test_resize.cpp
	modules/ocl/perf/main.cpp
	modules/ocl/perf/perf_calib3d.cpp
	modules/ocl/perf/perf_canny.cpp
	modules/ocl/perf/perf_color.cpp
	modules/ocl/perf/perf_haar.cpp
	modules/ocl/perf/perf_match_template.cpp
	modules/ocl/perf/perf_precomp.cpp
	modules/ocl/perf/perf_precomp.hpp

Conflicts:
	cmake/OpenCVConfig.cmake
	cmake/OpenCVLegacyOptions.cmake
	modules/contrib/src/retina.cpp
	modules/gpu/doc/camera_calibration_and_3d_reconstruction.rst
	modules/gpu/doc/video.rst
	modules/gpu/src/speckle_filtering.cpp
	modules/python/src2/cv2.cv.hpp
	modules/python/test/test2.py
	samples/python/watershed.py

Conflicts:
	modules/contrib/src/inputoutput.cpp
	modules/gpu/perf/perf_imgproc.cpp
	modules/gpuarithm/perf/perf_element_operations.cpp
	modules/gpuarithm/src/element_operations.cpp
	modules/ts/src/precomp.hpp

correct range computations, it is difficult to do the more efficient single pass 3N/2 implementations correctly in this particular setting without costing more than they save

Who knows, maybe the user's Sphinx installation uses a Python we couldn't find.

Most importantly, added support for multiple include directores,
but also did some general cleanup.

Rationale: we already depend on it (e.g. some scripts use print_function).

None of this should be relevant as of CMake 2.8.7.