diff --git a/modules/gpu/doc/operations_on_matrices.rst b/modules/gpu/doc/operations_on_matrices.rst
index 7f586a1b02..d1762f442a 100644
--- a/modules/gpu/doc/operations_on_matrices.rst
+++ b/modules/gpu/doc/operations_on_matrices.rst
@@ -242,3 +242,33 @@ Converts polar coordinates into Cartesian.
     :param stream: Stream for the asynchronous version.
 
 .. seealso:: :ocv:func:`polarToCart`
+
+
+
+gpu::normalize
+--------------
+Normalizes the norm or value range of an array.
+
+.. ocv:function:: void gpu::normalize(const GpuMat& src, GpuMat& dst, double alpha = 1, double beta = 0, int norm_type = NORM_L2, int dtype = -1, const GpuMat& mask = GpuMat())
+
+.. ocv:function:: void gpu::normalize(const GpuMat& src, GpuMat& dst, double a, double b, int norm_type, int dtype, const GpuMat& mask, GpuMat& norm_buf, GpuMat& cvt_buf)
+
+    :param src: input array.
+
+    :param dst: output array of the same size as  ``src`` .
+
+    :param alpha: norm value to normalize to or the lower range boundary in case of the range normalization.
+
+    :param beta: upper range boundary in case of the range normalization; it is not used for the norm normalization.
+
+    :param normType: normalization type (see the details below).
+
+    :param dtype: when negative, the output array has the same type as ``src``; otherwise, it has the same number of channels as  ``src`` and the depth ``=CV_MAT_DEPTH(dtype)``.
+
+    :param mask: optional operation mask.
+
+    :param norm_buf: Optional buffer to avoid extra memory allocations. It is resized automatically.
+
+    :param cvt_buf: Optional buffer to avoid extra memory allocations. It is resized automatically.
+
+.. seealso:: :ocv:func:`normalize`
diff --git a/modules/gpu/include/opencv2/gpu/gpu.hpp b/modules/gpu/include/opencv2/gpu/gpu.hpp
index 046339bd04..38eafdbdcf 100644
--- a/modules/gpu/include/opencv2/gpu/gpu.hpp
+++ b/modules/gpu/include/opencv2/gpu/gpu.hpp
@@ -465,6 +465,12 @@ CV_EXPORTS void cartToPolar(const GpuMat& x, const GpuMat& y, GpuMat& magnitude,
 //! supports only floating-point source
 CV_EXPORTS void polarToCart(const GpuMat& magnitude, const GpuMat& angle, GpuMat& x, GpuMat& y, bool angleInDegrees = false, Stream& stream = Stream::Null());
 
+//! scales and shifts array elements so that either the specified norm (alpha) or the minimum (alpha) and maximum (beta) array values get the specified values
+CV_EXPORTS void normalize(const GpuMat& src, GpuMat& dst, double alpha = 1, double beta = 0,
+                          int norm_type = NORM_L2, int dtype = -1, const GpuMat& mask = GpuMat());
+CV_EXPORTS void normalize(const GpuMat& src, GpuMat& dst, double a, double b,
+                          int norm_type, int dtype, const GpuMat& mask, GpuMat& norm_buf, GpuMat& cvt_buf);
+
 
 //////////////////////////// Per-element operations ////////////////////////////////////
 
diff --git a/modules/gpu/perf/perf_core.cpp b/modules/gpu/perf/perf_core.cpp
index 6b407f49b5..b97c4999cd 100644
--- a/modules/gpu/perf/perf_core.cpp
+++ b/modules/gpu/perf/perf_core.cpp
@@ -1926,4 +1926,48 @@ PERF_TEST_P(Sz_Depth_Cn_Code_Dim, Core_Reduce, Combine(
     }
 }
 
+//////////////////////////////////////////////////////////////////////
+// Normalize
+
+DEF_PARAM_TEST(Sz_Depth_NormType, cv::Size, MatDepth, NormType);
+
+PERF_TEST_P(Sz_Depth_NormType, Core_Normalize, Combine(
+    GPU_TYPICAL_MAT_SIZES,
+    Values(CV_8U, CV_16U, CV_32F, CV_64F),
+    Values(NormType(cv::NORM_INF),
+           NormType(cv::NORM_L1),
+           NormType(cv::NORM_L2),
+           NormType(cv::NORM_MINMAX))
+    ))
+{
+    cv::Size size = GET_PARAM(0);
+    int type = GET_PARAM(1);
+    int norm_type = GET_PARAM(2);
+
+    double alpha = 1;
+    double beta = 0;
+
+    cv::Mat src(size, type);
+    fillRandom(src);
+
+    if (PERF_RUN_GPU())
+    {
+        cv::gpu::GpuMat d_src(src);
+        cv::gpu::GpuMat d_dst;
+        cv::gpu::GpuMat d_norm_buf, d_cvt_buf;
+
+        TEST_CYCLE() cv::gpu::normalize(d_src, d_dst, alpha, beta, norm_type, type, cv::gpu::GpuMat(), d_norm_buf, d_cvt_buf);
+
+        GPU_SANITY_CHECK(d_dst, 1);
+    }
+    else
+    {
+        cv::Mat dst;
+
+        TEST_CYCLE() cv::normalize(src, dst, alpha, beta, norm_type, type);
+
+        CPU_SANITY_CHECK(dst, 1);
+    }
+}
+
 } // namespace
diff --git a/modules/gpu/perf/utility.hpp b/modules/gpu/perf/utility.hpp
index 09b84f53aa..6782b93768 100644
--- a/modules/gpu/perf/utility.hpp
+++ b/modules/gpu/perf/utility.hpp
@@ -17,7 +17,7 @@ CV_ENUM(BorderMode, cv::BORDER_REFLECT101, cv::BORDER_REPLICATE, cv::BORDER_CONS
 
 CV_ENUM(Interpolation, cv::INTER_NEAREST, cv::INTER_LINEAR, cv::INTER_CUBIC, cv::INTER_AREA)
 #define ALL_INTERPOLATIONS testing::ValuesIn(Interpolation::all())
-CV_ENUM(NormType, cv::NORM_INF, cv::NORM_L1, cv::NORM_L2, cv::NORM_HAMMING)
+CV_ENUM(NormType, cv::NORM_INF, cv::NORM_L1, cv::NORM_L2, cv::NORM_HAMMING, cv::NORM_MINMAX)
 
 const int Gray = 1, TwoChannel = 2, BGR = 3, BGRA = 4;
 CV_ENUM(MatCn, Gray, TwoChannel, BGR, BGRA)
diff --git a/modules/gpu/src/arithm.cpp b/modules/gpu/src/arithm.cpp
index 242febded9..7e0aaab680 100644
--- a/modules/gpu/src/arithm.cpp
+++ b/modules/gpu/src/arithm.cpp
@@ -59,6 +59,8 @@ void cv::gpu::magnitudeSqr(const GpuMat&, const GpuMat&, GpuMat&, Stream&) { thr
 void cv::gpu::phase(const GpuMat&, const GpuMat&, GpuMat&, bool, Stream&) { throw_nogpu(); }
 void cv::gpu::cartToPolar(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, bool, Stream&) { throw_nogpu(); }
 void cv::gpu::polarToCart(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, bool, Stream&) { throw_nogpu(); }
+void cv::gpu::normalize(const GpuMat&, GpuMat&, double, double, int, int, const GpuMat&) { throw_nogpu(); }
+void cv::gpu::normalize(const GpuMat&, GpuMat&, double, double, int, int, const GpuMat&, GpuMat&, GpuMat&) { throw_nogpu(); }
 
 #else /* !defined (HAVE_CUDA) */
 
@@ -529,4 +531,47 @@ void cv::gpu::polarToCart(const GpuMat& magnitude, const GpuMat& angle, GpuMat&
     polarToCart_caller(magnitude, angle, x, y, angleInDegrees, StreamAccessor::getStream(stream));
 }
 
+////////////////////////////////////////////////////////////////////////
+// normalize
+
+void cv::gpu::normalize(const GpuMat& src, GpuMat& dst, double a, double b, int norm_type, int dtype, const GpuMat& mask)
+{
+    GpuMat norm_buf;
+    GpuMat cvt_buf;
+    normalize(src, dst, a, b, norm_type, dtype, mask, norm_buf, cvt_buf);
+}
+
+void cv::gpu::normalize(const GpuMat& src, GpuMat& dst, double a, double b, int norm_type, int dtype, const GpuMat& mask, GpuMat& norm_buf, GpuMat& cvt_buf)
+{
+    double scale = 1, shift = 0;
+    if (norm_type == NORM_MINMAX)
+    {
+        double smin = 0, smax = 0;
+        double dmin = std::min(a, b), dmax = std::max(a, b);
+        minMax(src, &smin, &smax, mask, norm_buf);
+        scale = (dmax - dmin) * (smax - smin > numeric_limits<double>::epsilon() ? 1.0 / (smax - smin) : 0.0);
+        shift = dmin - smin * scale;
+    }
+    else if (norm_type == NORM_L2 || norm_type == NORM_L1 || norm_type == NORM_INF)
+    {
+        scale = norm(src, norm_type, mask, norm_buf);
+        scale = scale > numeric_limits<double>::epsilon() ? a / scale : 0.0;
+        shift = 0;
+    }
+    else
+    {
+        CV_Error(CV_StsBadArg, "Unknown/unsupported norm type");
+    }
+
+    if (mask.empty())
+    {
+        src.convertTo(dst, dtype, scale, shift);
+    }
+    else
+    {
+        src.convertTo(cvt_buf, dtype, scale, shift);
+        cvt_buf.copyTo(dst, mask);
+    }
+}
+
 #endif /* !defined (HAVE_CUDA) */
diff --git a/modules/gpu/test/test_core.cpp b/modules/gpu/test/test_core.cpp
index ad0cf53ffd..736256cd51 100644
--- a/modules/gpu/test/test_core.cpp
+++ b/modules/gpu/test/test_core.cpp
@@ -3540,4 +3540,70 @@ INSTANTIATE_TEST_CASE_P(GPU_Core, Reduce, testing::Combine(
     ALL_REDUCE_CODES,
     WHOLE_SUBMAT));
 
+//////////////////////////////////////////////////////////////////////////////
+// Normalize
+
+PARAM_TEST_CASE(Normalize, cv::gpu::DeviceInfo, cv::Size, MatDepth, NormCode, UseRoi)
+{
+    cv::gpu::DeviceInfo devInfo;
+    cv::Size size;
+    int type;
+    int norm_type;
+    bool useRoi;
+
+    double alpha;
+    double beta;
+
+    virtual void SetUp()
+    {
+        devInfo = GET_PARAM(0);
+        size = GET_PARAM(1);
+        type = GET_PARAM(2);
+        norm_type = GET_PARAM(3);
+        useRoi = GET_PARAM(4);
+
+        cv::gpu::setDevice(devInfo.deviceID());
+
+        alpha = 1;
+        beta = 0;
+    }
+
+};
+
+GPU_TEST_P(Normalize, WithOutMask)
+{
+    cv::Mat src = randomMat(size, type);
+
+    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
+    cv::gpu::normalize(loadMat(src, useRoi), dst, alpha, beta, norm_type, type);
+
+    cv::Mat dst_gold;
+    cv::normalize(src, dst_gold, alpha, beta, norm_type, type);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 1e-6);
+}
+
+GPU_TEST_P(Normalize, WithMask)
+{
+    cv::Mat src = randomMat(size, type);
+    cv::Mat mask = randomMat(size, CV_8UC1, 0, 2);
+
+    cv::gpu::GpuMat dst = createMat(size, type, useRoi);
+    dst.setTo(cv::Scalar::all(0));
+    cv::gpu::normalize(loadMat(src, useRoi), dst, alpha, beta, norm_type, type, loadMat(mask, useRoi));
+
+    cv::Mat dst_gold(size, type);
+    dst_gold.setTo(cv::Scalar::all(0));
+    cv::normalize(src, dst_gold, alpha, beta, norm_type, type, mask);
+
+    EXPECT_MAT_NEAR(dst_gold, dst, 1e-6);
+}
+
+INSTANTIATE_TEST_CASE_P(GPU_Core, Normalize, testing::Combine(
+    ALL_DEVICES,
+    DIFFERENT_SIZES,
+    ALL_DEPTH,
+    testing::Values(NormCode(cv::NORM_L1), NormCode(cv::NORM_L2), NormCode(cv::NORM_INF), NormCode(cv::NORM_MINMAX)),
+    WHOLE_SUBMAT));
+
 #endif // HAVE_CUDA