Moved aruco dictionary globals into getPredefinedDictionary so constructed on first use, to allow lazy-loading of the opencv DLL

tesseract 4.x headers require C++11 (can't be used by OpenCV 3.4 without C++11)

  * limit the affect by using HAVE_CUDA

Enhance EdgeAwareInterpolation no. matches limitation (#2209)

* sparse_match_interpolators EdgeAwareInterpolation - enhance limitation of the maximal number of matches to be interpolated from SHORT_MAX to INT_MAX by substituting short by int types

* sparse_match_interpolators EdgeAwareInterpolation - enhance limitation of the maximal number of matches to be interpolated from SHORT_MAX to INT_MAX by substituting short by int types

* fix wrong allocation of labels mats CV16S to CV_32S

Signed-off-by: SSE4 <tomskside@gmail.com>

add python sample

* issue-2060: implement ability to return scores for edgeboxes implementation

* issue-2060: address feedback to add some foolproofing

* doc: fix return parameter for void method

* docs: add missing parameter description

* added detection of inverted Markers, tests included

* corrections added, default preserved, identation, var, comment

Move background subtraction python tutorial to proper location (#2120)

* add tutorial from OpenCV-Python Tutorials

* remove trailing whitespace

* remove MOG2 part

* add cross-link to MOG2 tutorial

Incomplete sentence corrected. Resolves Issue #1997 (#2000)

* Incomplete sentence corrected. Issue #1997 

Resolves Issue #1997

``` FaceRecognizer::train(InputArrayOfArrays src, InputArray labels); ```

The type `InputArray` can accept `vector<T>` or `Mat` types as mentioned here: https://docs.opencv.org/3.4/d4/d32/classcv_1_1__InputArray.html

* Trailing whitespaces problem resolved

* fix rectifyPatch bug when patch orientation is not extracted