diff --git a/modules/gpuimgproc/include/opencv2/gpuimgproc.hpp b/modules/gpuimgproc/include/opencv2/gpuimgproc.hpp
index 2fd17d9b35..24b8b64f0a 100644
--- a/modules/gpuimgproc/include/opencv2/gpuimgproc.hpp
+++ b/modules/gpuimgproc/include/opencv2/gpuimgproc.hpp
@@ -409,17 +409,17 @@ CV_EXPORTS Ptr<CornersDetector> createGoodFeaturesToTrackDetector(int srcType, i
 ///////////////////////////// Mean Shift //////////////////////////////
 
 //! Does mean shift filtering on GPU.
-CV_EXPORTS void meanShiftFiltering(const GpuMat& src, GpuMat& dst, int sp, int sr,
+CV_EXPORTS void meanShiftFiltering(InputArray src, OutputArray dst, int sp, int sr,
                                    TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1),
                                    Stream& stream = Stream::Null());
 
 //! Does mean shift procedure on GPU.
-CV_EXPORTS void meanShiftProc(const GpuMat& src, GpuMat& dstr, GpuMat& dstsp, int sp, int sr,
+CV_EXPORTS void meanShiftProc(InputArray src, OutputArray dstr, OutputArray dstsp, int sp, int sr,
                               TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1),
                               Stream& stream = Stream::Null());
 
 //! Does mean shift segmentation with elimination of small regions.
-CV_EXPORTS void meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr, int minsize,
+CV_EXPORTS void meanShiftSegmentation(InputArray src, OutputArray dst, int sp, int sr, int minsize,
                                       TermCriteria criteria = TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 5, 1));
 
 /////////////////////////// Match Template ////////////////////////////
diff --git a/modules/gpuimgproc/src/mean_shift.cpp b/modules/gpuimgproc/src/mean_shift.cpp
index e30f95bf98..26368ca5ac 100644
--- a/modules/gpuimgproc/src/mean_shift.cpp
+++ b/modules/gpuimgproc/src/mean_shift.cpp
@@ -47,13 +47,13 @@ using namespace cv::gpu;
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-void cv::gpu::meanShiftFiltering(const GpuMat&, GpuMat&, int, int, TermCriteria, Stream&) { throw_no_cuda(); }
-void cv::gpu::meanShiftProc(const GpuMat&, GpuMat&, GpuMat&, int, int, TermCriteria, Stream&) { throw_no_cuda(); }
+void cv::gpu::meanShiftFiltering(InputArray, OutputArray, int, int, TermCriteria, Stream&) { throw_no_cuda(); }
+void cv::gpu::meanShiftProc(InputArray, OutputArray, OutputArray, int, int, TermCriteria, Stream&) { throw_no_cuda(); }
 
 #else /* !defined (HAVE_CUDA) */
 
 ////////////////////////////////////////////////////////////////////////
-// meanShiftFiltering_GPU
+// meanShiftFiltering
 
 namespace cv { namespace gpu { namespace cudev
 {
@@ -63,27 +63,26 @@ namespace cv { namespace gpu { namespace cudev
     }
 }}}
 
-void cv::gpu::meanShiftFiltering(const GpuMat& src, GpuMat& dst, int sp, int sr, TermCriteria criteria, Stream& stream)
+void cv::gpu::meanShiftFiltering(InputArray _src, OutputArray _dst, int sp, int sr, TermCriteria criteria, Stream& stream)
 {
     using namespace ::cv::gpu::cudev::imgproc;
 
-    if( src.empty() )
-        CV_Error( cv::Error::StsBadArg, "The input image is empty" );
+    GpuMat src = _src.getGpuMat();
 
-    if( src.depth() != CV_8U || src.channels() != 4 )
-        CV_Error( cv::Error::StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
+    CV_Assert( src.type() == CV_8UC4 );
 
-    dst.create( src.size(), CV_8UC4 );
+    _dst.create(src.size(), CV_8UC4);
+    GpuMat dst = _dst.getGpuMat();
 
-    if( !(criteria.type & TermCriteria::MAX_ITER) )
+    if (!(criteria.type & TermCriteria::MAX_ITER))
         criteria.maxCount = 5;
 
     int maxIter = std::min(std::max(criteria.maxCount, 1), 100);
 
-    float eps;
-    if( !(criteria.type & TermCriteria::EPS) )
-        eps = 1.f;
-    eps = (float)std::max(criteria.epsilon, 0.0);
+    if (!(criteria.type & TermCriteria::EPS))
+        criteria.epsilon = 1.f;
+
+    float eps = (float) std::max(criteria.epsilon, 0.0);
 
     meanShiftFiltering_gpu(src, dst, sp, sr, maxIter, eps, StreamAccessor::getStream(stream));
 }
@@ -99,28 +98,29 @@ namespace cv { namespace gpu { namespace cudev
     }
 }}}
 
-void cv::gpu::meanShiftProc(const GpuMat& src, GpuMat& dstr, GpuMat& dstsp, int sp, int sr, TermCriteria criteria, Stream& stream)
+void cv::gpu::meanShiftProc(InputArray _src, OutputArray _dstr, OutputArray _dstsp, int sp, int sr, TermCriteria criteria, Stream& stream)
 {
     using namespace ::cv::gpu::cudev::imgproc;
 
-    if( src.empty() )
-        CV_Error( cv::Error::StsBadArg, "The input image is empty" );
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC4 );
 
-    if( src.depth() != CV_8U || src.channels() != 4 )
-        CV_Error( cv::Error::StsUnsupportedFormat, "Only 8-bit, 4-channel images are supported" );
+    _dstr.create(src.size(), CV_8UC4);
+    _dstsp.create(src.size(), CV_16SC2);
 
-    dstr.create( src.size(), CV_8UC4 );
-    dstsp.create( src.size(), CV_16SC2 );
+    GpuMat dstr = _dstr.getGpuMat();
+    GpuMat dstsp = _dstsp.getGpuMat();
 
-    if( !(criteria.type & TermCriteria::MAX_ITER) )
+    if (!(criteria.type & TermCriteria::MAX_ITER))
         criteria.maxCount = 5;
 
     int maxIter = std::min(std::max(criteria.maxCount, 1), 100);
 
-    float eps;
-    if( !(criteria.type & TermCriteria::EPS) )
-        eps = 1.f;
-    eps = (float)std::max(criteria.epsilon, 0.0);
+    if (!(criteria.type & TermCriteria::EPS))
+        criteria.epsilon = 1.f;
+
+    float eps = (float) std::max(criteria.epsilon, 0.0);
 
     meanShiftProc_gpu(src, dstr, dstsp, sp, sr, maxIter, eps, StreamAccessor::getStream(stream));
 }
diff --git a/modules/gpuimgproc/src/mssegmentation.cpp b/modules/gpuimgproc/src/mssegmentation.cpp
index 7f02168e1a..ec1c5feb4b 100644
--- a/modules/gpuimgproc/src/mssegmentation.cpp
+++ b/modules/gpuimgproc/src/mssegmentation.cpp
@@ -43,7 +43,7 @@
 
 #if !defined HAVE_CUDA || defined(CUDA_DISABLER)
 
-void cv::gpu::meanShiftSegmentation(const GpuMat&, Mat&, int, int, int, TermCriteria) { throw_no_cuda(); }
+void cv::gpu::meanShiftSegmentation(InputArray, OutputArray, int, int, int, TermCriteria) { throw_no_cuda(); }
 
 #else
 
@@ -222,9 +222,12 @@ inline int dist2(const cv::Vec2s& lhs, const cv::Vec2s& rhs)
 } // anonymous namespace
 
 
-void cv::gpu::meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr, int minsize, TermCriteria criteria)
+void cv::gpu::meanShiftSegmentation(InputArray _src, OutputArray _dst, int sp, int sr, int minsize, TermCriteria criteria)
 {
-    CV_Assert(src.type() == CV_8UC4);
+    GpuMat src = _src.getGpuMat();
+
+    CV_Assert( src.type() == CV_8UC4 );
+
     const int nrows = src.rows;
     const int ncols = src.cols;
     const int hr = sr;
@@ -232,7 +235,7 @@ void cv::gpu::meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr,
 
     // Perform mean shift procedure and obtain region and spatial maps
     GpuMat d_rmap, d_spmap;
-    meanShiftProc(src, d_rmap, d_spmap, sp, sr, criteria);
+    gpu::meanShiftProc(src, d_rmap, d_spmap, sp, sr, criteria);
     Mat rmap(d_rmap);
     Mat spmap(d_spmap);
 
@@ -337,7 +340,7 @@ void cv::gpu::meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr,
     }
 
     // Sort all graph's edges connecting differnet components (in asceding order)
-    sort(edges.begin(), edges.end());
+    std::sort(edges.begin(), edges.end());
 
     // Exclude small components (starting from the nearest couple)
     for (size_t i = 0; i < edges.size(); ++i)
@@ -366,7 +369,8 @@ void cv::gpu::meanShiftSegmentation(const GpuMat& src, Mat& dst, int sp, int sr,
     }
 
     // Create final image, color of each segment is the average color of its pixels
-    dst.create(src.size(), src.type());
+    _dst.create(src.size(), src.type());
+    Mat dst = _dst.getMat();
 
     for (int y = 0; y < nrows; ++y)
     {