use _mm_sad_epu8 instead of _mm_movemask_epi8 in countNonZero. Therefore, we do not need any popcount table and this solves http://code.opencv.org/issues/4333

http://code.opencv.org/issues/4124

avoid possible problems with zero values in tone mapping algorithms (http://code.opencv.org/issues/4020)

added test to prove that remap does not leak memory (http://code.opencv.org/issues/2502). disabled the test for now to save execution time.

significantly improved parallel non-local means by using granularity parameter in parallel_for_ loop. Because the algorithm deals with sliding sums, it's essential that each thread has enough work to do, otherwise the algorithm gets higher theoretical complexity and thus there is no speedup comparing to 1-thread code (at best).

increased singularity epsilon in LU decomposition. This solved singular case from http://code.opencv.org/issues/3305. Added the respective test.

added support for n-channel (n>4) images in warpAffine/warpPerspective/remap: http://code.opencv.org/issues/4190

when true, use signed gradient for the angular histogram computation.
default to false for backward compatibility.

A define was forgotten for the codecs and there was a typo for
another.