resolve undeclared identifier close error

UIImages with alpha were ending up with garbage pixels in background (random memory values).  Need to initialize matrix pixels before drawing UIImage with alpha on it.

Note: didn’t fix Grayscale image with alpha stripping alpha in UIImage -> Mat conversion.

Adding of destructor and placement new constructors for classes wrapped with CV_EXPORTS_W_SIMPLE macro

The TiffDecoder keeps an open file handle.
As a consequence the file cannot be removed
before the TiffDecoder closes the file.

Replaced quiet return from MSER:detectRegion function with exeption throwing to notify user in case small image is processed

      * accumulate, accumulateSquare, accumulateProduct and accumulateWeighted

1. if a component's variation is a global minimum than it should be a local minimum
2. for the small image with invert and blur, the MSERs number should be 20

Exp and Log IPP implementations are changed to hal interface;

consequences of changes

accuracy test

Solve issue 7063

  * The difference becomes too large when multiply is done in int16
  * To reproduce the test failure, IPP has to be switched off

"declare.in(img, WARMUP_RNG)" overrides input image with random values

Also added support for different input sizes (via cv::resize)

There is an issue with processing of abs(short) function for
negative argument.

Affected OpenCL devices:
- iGPU: Intel(R) HD Graphics 520 (OpenCL 2.0 )
- CPU: Intel(R) Core(TM) i5-6300U CPU @ 2.40GHz (OpenCL 2.0 (Build 10094))

GraphCutSeamFinderGpu requires cudalegacy.

  * avoid using vld1_f16 and vst1_f16 on gcc 4 series (Ubuntu 14.04)
  * guard correctly with #if
  * use static inline

the merge() calls growHistory() too many times such that:
1. some CompHistory nodes not used have been created
2. some CompHistory node's val equal their parents

* Common Canny parallelization added. TBB and single thread code removed. Final pass vectorized with SSE2 intrinsics.

* wrong #ifdef replaced with #if

* Merged to actual Canny version

* Merged common parallelized Canny with actual Canny implementation

* Remove 'Mutex *mutex' and pass 'Mutex mutex' from outside to parallelCanny

* Replaced extern Mutex with intern mutable Mutex.

  * stop using 'CUDA' prefix
  * add explanation

Problem is observed with test_loop_times=100
in corner cases like 93x1 or similar