* Fix for Homogenous precision #14242:
 - moved scale computation to an inline function
 - use std::numeric_limits<float>::epsilon() instead of != 0.0

* Fix for Homogenous precision #14242:
 - fixed warnings for type conversion

* Fix for Homogenous precision #14242:
 - use float epsilon() for truncation of doubles

backporting commit c9f3f4d1

BUG: Qt5 components not correctly detected if only Qt5_DIR is provided, when this is supported by Qt CMake system.

Fixed by updating Qt5 find_package call to use modern CMake syntax

Fixes: https://github.com/opencv/opencv/issues/14169

Grammatical errors for help() in detect_mser.cpp (#14122)

* Grammatical errors for help() statement

Corrected spelling of "synthetic" and added grammatical clarification for keys to press to change view or use mouse.

* Adjustment of superfluous spaces

similar to '--test_require_data' parameter

* fix getting platformIDs in initializeContextFromD3D11Device

fixes build with VTK

backporting of commit: 25247b9f

Fixes a broken link for histograms of Oriented Gradients.

Asynchronous API from Intel's Inference Engine (#13694)

* Add forwardAsync for asynchronous mode from Intel's Inference Engine

* Python test for forwardAsync

* Replace Future_Mat to AsyncMat

* Shadow AsyncMat

* Isolate InferRequest callback

* Manage exceptions in Async API of IE