documentation: avoid links to 'master' branch from 3.4 maintenance branch

modules/core/doc/cuda.markdown

@@ -82,4 +82,4 @@ Block Matching algorithm has been successfully parallelized using the following
 3.  Merge the results into a single disparity map.
 
 With this algorithm, a dual GPU gave a 180% performance increase comparing to the single Fermi GPU.
-For a source code example, see <https://github.com/opencv/opencv/tree/master/samples/gpu/>.
+For a source code example, see <https://github.com/opencv/opencv/tree/3.4/samples/gpu/>.

modules/highgui/include/opencv2/highgui.hpp

@@ -456,7 +456,7 @@ CV_EXPORTS_W Rect getWindowImageRect(const String& winname);
 
 @param winname Name of the window.
 @param onMouse Mouse callback. See OpenCV samples, such as
-<https://github.com/opencv/opencv/tree/master/samples/cpp/ffilldemo.cpp>, on how to specify and
+<https://github.com/opencv/opencv/tree/3.4/samples/cpp/ffilldemo.cpp>, on how to specify and
 use the callback.
 @param userdata The optional parameter passed to the callback.
  */

modules/objdetect/include/opencv2/objdetect.hpp

@@ -91,7 +91,7 @@ compensate for the differences in the size of areas. The sums of pixel values ov
 regions are calculated rapidly using integral images (see below and the integral description).
 
 To see the object detector at work, have a look at the facedetect demo:
-<https://github.com/opencv/opencv/tree/master/samples/cpp/dbt_face_detection.cpp>
+<https://github.com/opencv/opencv/tree/3.4/samples/cpp/dbt_face_detection.cpp>
 
 The following reference is for the detection part only. There is a separate application called
 opencv_traincascade that can train a cascade of boosted classifiers from a set of samples.