diff --git a/modules/core/include/opencv2/core.hpp b/modules/core/include/opencv2/core.hpp
index 48023844a984ddf366db546d8524c3f098f9f82e..70ea4f8c1fa2f8bee806f1bf2ea2f15bed164f05 100644
--- a/modules/core/include/opencv2/core.hpp
+++ b/modules/core/include/opencv2/core.hpp
@@ -819,12 +819,45 @@ mixChannels , or split .
 @param minLoc pointer to the returned minimum location (in 2D case); NULL is used if not required.
 @param maxLoc pointer to the returned maximum location (in 2D case); NULL is used if not required.
 @param mask optional mask used to select a sub-array.
-@sa max, min, compare, inRange, extractImageCOI, mixChannels, split, Mat::reshape
+@sa max, min, reduceArgMin, reduceArgMax, compare, inRange, extractImageCOI, mixChannels, split, Mat::reshape
 */
 CV_EXPORTS_W void minMaxLoc(InputArray src, CV_OUT double* minVal,
                             CV_OUT double* maxVal = 0, CV_OUT Point* minLoc = 0,
                             CV_OUT Point* maxLoc = 0, InputArray mask = noArray());
 
+/**
+ * @brief Finds indices of min elements along provided axis
+ *
+ * @note
+ *      - If input or output array is not continuous, this function will create an internal copy.
+ *      - NaN handling is left unspecified, see patchNaNs().
+ *      - The returned index is always in bounds of input matrix.
+ *
+ * @param src input single-channel array.
+ * @param dst output array of type CV_32SC1 with the same dimensionality as src,
+ * except for axis being reduced - it should be set to 1.
+ * @param lastIndex whether to get the index of first or last occurrence of min.
+ * @param axis axis to reduce along.
+ * @sa reduceArgMax, minMaxLoc, min, max, compare, reduce
+ */
+CV_EXPORTS_W void reduceArgMin(InputArray src, OutputArray dst, int axis, bool lastIndex = false);
+
+/**
+ * @brief Finds indices of max elements along provided axis
+ *
+ * @note
+ *      - If input or output array is not continuous, this function will create an internal copy.
+ *      - NaN handling is left unspecified, see patchNaNs().
+ *      - The returned index is always in bounds of input matrix.
+ *
+ * @param src input single-channel array.
+ * @param dst output array of type CV_32SC1 with the same dimensionality as src,
+ * except for axis being reduced - it should be set to 1.
+ * @param lastIndex whether to get the index of first or last occurrence of max.
+ * @param axis axis to reduce along.
+ * @sa reduceArgMin, minMaxLoc, min, max, compare, reduce
+ */
+CV_EXPORTS_W void reduceArgMax(InputArray src, OutputArray dst, int axis, bool lastIndex = false);
 
 /** @brief Finds the global minimum and maximum in an array
 
@@ -886,7 +919,7 @@ a single row. 1 means that the matrix is reduced to a single column.
 @param rtype reduction operation that could be one of #ReduceTypes
 @param dtype when negative, the output vector will have the same type as the input matrix,
 otherwise, its type will be CV_MAKE_TYPE(CV_MAT_DEPTH(dtype), src.channels()).
-@sa repeat
+@sa repeat, reduceArgMin, reduceArgMax
 */
 CV_EXPORTS_W void reduce(InputArray src, OutputArray dst, int dim, int rtype, int dtype = -1);
 
diff --git a/modules/core/include/opencv2/core/detail/dispatch_helper.impl.hpp b/modules/core/include/opencv2/core/detail/dispatch_helper.impl.hpp
new file mode 100644
index 0000000000000000000000000000000000000000..d6ec6769220031ccc2664d7bc96f5747023b97a6
--- /dev/null
+++ b/modules/core/include/opencv2/core/detail/dispatch_helper.impl.hpp
@@ -0,0 +1,49 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_DETAIL_DISPATCH_HELPER_IMPL_HPP
+#define OPENCV_CORE_DETAIL_DISPATCH_HELPER_IMPL_HPP
+
+//! @cond IGNORED
+
+namespace cv {
+namespace detail {
+
+template<template<typename> class Functor, typename... Args>
+static inline void depthDispatch(const int depth, Args&&... args)
+{
+    switch (depth)
+    {
+        case CV_8U:
+            Functor<uint8_t>{}(std::forward<Args>(args)...);
+            break;
+        case CV_8S:
+            Functor<int8_t>{}(std::forward<Args>(args)...);
+            break;
+        case CV_16U:
+            Functor<uint16_t>{}(std::forward<Args>(args)...);
+            break;
+        case CV_16S:
+            Functor<int16_t>{}(std::forward<Args>(args)...);
+            break;
+        case CV_32S:
+            Functor<int32_t>{}(std::forward<Args>(args)...);
+            break;
+        case CV_32F:
+            Functor<float>{}(std::forward<Args>(args)...);
+            break;
+        case CV_64F:
+            Functor<double>{}(std::forward<Args>(args)...);
+            break;
+        case CV_16F:
+        default:
+            CV_Error(cv::Error::BadDepth, "Unsupported matrix type.");
+    };
+}
+
+}}
+
+//! @endcond
+
+#endif //OPENCV_CORE_DETAIL_DISPATCH_HELPER_IMPL_HPP
diff --git a/modules/core/perf/perf_reduce.cpp b/modules/core/perf/perf_reduce.cpp
index 4fc5aa55f995a9767b77407e3c989271bf0f33a1..8f9c2e8349f9fbd6f262f2643e6631656ef2b07f 100644
--- a/modules/core/perf/perf_reduce.cpp
+++ b/modules/core/perf/perf_reduce.cpp
@@ -65,4 +65,33 @@ PERF_TEST_P(Size_MatType_ROp, reduceC,
     SANITY_CHECK(vec, 1);
 }
 
+typedef tuple<Size, MatType, int> Size_MatType_RMode_t;
+typedef perf::TestBaseWithParam<Size_MatType_RMode_t> Size_MatType_RMode;
+
+PERF_TEST_P(Size_MatType_RMode, DISABLED_reduceArgMinMax, testing::Combine(
+        testing::Values(TYPICAL_MAT_SIZES),
+        testing::Values(CV_8U, CV_32F),
+        testing::Values(0, 1)
+)
+)
+{
+    Size srcSize = get<0>(GetParam());
+    int matType = get<1>(GetParam());
+    int axis = get<2>(GetParam());
+
+    Mat src(srcSize, matType);
+
+    std::vector<int> dstSize(src.dims);
+    std::copy(src.size.p, src.size.p + src.dims, dstSize.begin());
+    dstSize[axis] = 1;
+
+    Mat dst(dstSize, CV_32S, 0.);
+
+    declare.in(src, WARMUP_RNG).out(dst);
+
+    TEST_CYCLE() cv::reduceArgMin(src, dst, axis, true);
+
+    SANITY_CHECK_NOTHING();
+}
+
 } // namespace
diff --git a/modules/core/src/minmax.cpp b/modules/core/src/minmax.cpp
index 61bddc3d356725ccfc8b35ea99ccdc2656f22d46..96a7d027ada9898350949eb8fc62fb786b4841d2 100644
--- a/modules/core/src/minmax.cpp
+++ b/modules/core/src/minmax.cpp
@@ -7,6 +7,9 @@
 #include "opencl_kernels_core.hpp"
 #include "opencv2/core/openvx/ovx_defs.hpp"
 #include "stat.hpp"
+#include "opencv2/core/detail/dispatch_helper.impl.hpp"
+
+#include <algorithm>
 
 #undef HAVE_IPP
 #undef CV_IPP_RUN_FAST
@@ -1570,3 +1573,118 @@ void cv::minMaxLoc( InputArray _img, double* minVal, double* maxVal,
     if( maxLoc )
         std::swap(maxLoc->x, maxLoc->y);
 }
+
+enum class ReduceMode
+{
+    FIRST_MIN = 0, //!< get index of first min occurrence
+    LAST_MIN  = 1, //!< get index of last min occurrence
+    FIRST_MAX = 2, //!< get index of first max occurrence
+    LAST_MAX  = 3, //!< get index of last max occurrence
+};
+
+template <typename T>
+struct reduceMinMaxImpl
+{
+    void operator()(const cv::Mat& src, cv::Mat& dst, ReduceMode mode, const int axis) const
+    {
+        switch(mode)
+        {
+        case ReduceMode::FIRST_MIN:
+            reduceMinMaxApply<std::less>(src, dst, axis);
+            break;
+        case ReduceMode::LAST_MIN:
+            reduceMinMaxApply<std::less_equal>(src, dst, axis);
+            break;
+        case ReduceMode::FIRST_MAX:
+            reduceMinMaxApply<std::greater>(src, dst, axis);
+            break;
+        case ReduceMode::LAST_MAX:
+            reduceMinMaxApply<std::greater_equal>(src, dst, axis);
+            break;
+        }
+    }
+
+    template <template<class> class Cmp>
+    static void reduceMinMaxApply(const cv::Mat& src, cv::Mat& dst, const int axis)
+    {
+        Cmp<T> cmp;
+
+        const auto *src_ptr = src.ptr<T>();
+        auto *dst_ptr = dst.ptr<int32_t>();
+
+        const size_t outer_size = src.total(0, axis);
+        const auto mid_size = static_cast<size_t>(src.size[axis]);
+
+        const size_t outer_step = src.total(axis);
+        const size_t dst_step = dst.total(axis);
+
+        const size_t mid_step = src.total(axis + 1);
+
+        for (size_t outer = 0; outer < outer_size; ++outer)
+        {
+            const size_t outer_offset = outer * outer_step;
+            const size_t dst_offset = outer * dst_step;
+            for (size_t mid = 0; mid != mid_size; ++mid)
+            {
+                const size_t src_offset = outer_offset + mid * mid_step;
+                for (size_t inner = 0; inner < mid_step; inner++)
+                {
+                    int32_t& index = dst_ptr[dst_offset + inner];
+
+                    const size_t prev = outer_offset + index * mid_step + inner;
+                    const size_t curr = src_offset + inner;
+
+                    if (cmp(src_ptr[curr], src_ptr[prev]))
+                    {
+                        index = static_cast<int32_t>(mid);
+                    }
+                }
+            }
+        }
+    }
+};
+
+static void reduceMinMax(cv::InputArray src, cv::OutputArray dst, ReduceMode mode, int axis)
+{
+    CV_INSTRUMENT_REGION();
+
+    cv::Mat srcMat = src.getMat();
+    axis = (axis + srcMat.dims) % srcMat.dims;
+    CV_Assert(srcMat.channels() == 1 && axis >= 0 && axis < srcMat.dims);
+
+    std::vector<int> sizes(srcMat.dims);
+    std::copy(srcMat.size.p, srcMat.size.p + srcMat.dims, sizes.begin());
+    sizes[axis] = 1;
+
+    dst.create(srcMat.dims, sizes.data(), CV_32SC1); // indices
+    cv::Mat dstMat = dst.getMat();
+    dstMat.setTo(cv::Scalar::all(0));
+
+    if (!srcMat.isContinuous())
+    {
+        srcMat = srcMat.clone();
+    }
+
+    bool needs_copy = !dstMat.isContinuous();
+    if (needs_copy)
+    {
+        dstMat = dstMat.clone();
+    }
+
+    cv::detail::depthDispatch<reduceMinMaxImpl>(srcMat.depth(), srcMat, dstMat, mode, axis);
+
+    if (needs_copy)
+    {
+        dstMat.copyTo(dst);
+    }
+}
+
+void cv::reduceArgMin(InputArray src, OutputArray dst, int axis, bool lastIndex)
+{
+    reduceMinMax(src, dst, lastIndex ? ReduceMode::LAST_MIN : ReduceMode::FIRST_MIN, axis);
+}
+
+void cv::reduceArgMax(InputArray src, OutputArray dst, int axis, bool lastIndex)
+{
+    reduceMinMax(src, dst, lastIndex ? ReduceMode::LAST_MAX : ReduceMode::FIRST_MAX, axis);
+}
diff --git a/modules/core/test/ref_reduce_arg.impl.hpp b/modules/core/test/ref_reduce_arg.impl.hpp
new file mode 100644
index 0000000000000000000000000000000000000000..22b7c2b7a650b5cbeffa62d742505e9d0ff90e29
--- /dev/null
+++ b/modules/core/test/ref_reduce_arg.impl.hpp
@@ -0,0 +1,78 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_TEST_REF_REDUCE_ARG_HPP
+#define OPENCV_TEST_REF_REDUCE_ARG_HPP
+
+#include "opencv2/core/detail/dispatch_helper.impl.hpp"
+
+#include <algorithm>
+#include <numeric>
+
+namespace cvtest {
+
+template <class Cmp, typename T>
+struct reduceMinMaxImpl
+{
+    void operator()(const cv::Mat& src, cv::Mat& dst, const int axis) const
+    {
+        Cmp cmp;
+        std::vector<int> sizes(src.dims);
+        std::copy(src.size.p, src.size.p + src.dims, sizes.begin());
+
+        std::vector<cv::Range> idx(sizes.size(), cv::Range(0, 1));
+        idx[axis] = cv::Range::all();
+        const int n = std::accumulate(begin(sizes), end(sizes), 1, std::multiplies<int>());
+        const std::vector<int> newShape{1, src.size[axis]};
+        for (int i = 0; i < n ; ++i)
+        {
+            cv::Mat sub = src(idx);
+
+            auto begin = sub.begin<T>();
+            auto it = std::min_element(begin, sub.end<T>(), cmp);
+            *dst(idx).ptr<int32_t>() = static_cast<int32_t>(std::distance(begin, it));
+
+            for (int j = static_cast<int>(idx.size()) - 1; j >= 0; --j)
+            {
+                if (j == axis)
+                {
+                    continue;
+                }
+                const int old_s = idx[j].start;
+                const int new_s = (old_s + 1) % sizes[j];
+                if (new_s > old_s)
+                {
+                    idx[j] = cv::Range(new_s, new_s + 1);
+                    break;
+                }
+                idx[j] = cv::Range(0, 1);
+            }
+        }
+    }
+};
+
+template<template<class> class Cmp>
+struct MinMaxReducer{
+    template <typename T>
+    using Impl = reduceMinMaxImpl<Cmp<T>, T>;
+
+    static void reduce(const Mat& src, Mat& dst, int axis)
+    {
+        axis = (axis + src.dims) % src.dims;
+        CV_Assert(src.channels() == 1 && axis >= 0 && axis < src.dims);
+
+        std::vector<int> sizes(src.dims);
+        std::copy(src.size.p, src.size.p + src.dims, sizes.begin());
+        sizes[axis] = 1;
+
+        dst.create(sizes, CV_32SC1); // indices
+        dst.setTo(cv::Scalar::all(0));
+
+        cv::detail::depthDispatch<Impl>(src.depth(), src, dst, axis);
+    }
+};
+
+}
+
+#endif //OPENCV_TEST_REF_REDUCE_ARG_HPP
diff --git a/modules/core/test/test_arithm.cpp b/modules/core/test/test_arithm.cpp
index 9e8e242d604a37c3378a65cfe723ab52cc6d4b4b..4a9522d3d1539bf95ad0772ce6544e4bcd0cdffe 100644
--- a/modules/core/test/test_arithm.cpp
+++ b/modules/core/test/test_arithm.cpp
@@ -2,6 +2,7 @@
 // It is subject to the license terms in the LICENSE file found in the top-level directory
 // of this distribution and at http://opencv.org/license.html.
 #include "test_precomp.hpp"
+#include "ref_reduce_arg.impl.hpp"
 
 namespace opencv_test { namespace {
 
@@ -1387,6 +1388,73 @@ struct MinMaxLocOp : public BaseElemWiseOp
     }
 };
 
+struct reduceArgMinMaxOp : public BaseElemWiseOp
+{
+    reduceArgMinMaxOp() : BaseElemWiseOp(1, FIX_ALPHA+FIX_BETA+FIX_GAMMA, 1, 1, Scalar::all(0))
+    {
+        context = ARITHM_MAX_NDIMS*2 + 2;
+    };
+    int getRandomType(RNG& rng) override
+    {
+        return cvtest::randomType(rng, _OutputArray::DEPTH_MASK_ALL_BUT_8S, 1, 1);
+    }
+    void getRandomSize(RNG& rng, vector<int>& size) override
+    {
+        cvtest::randomSize(rng, 2, ARITHM_MAX_NDIMS, 6, size);
+    }
+    void generateScalars(int depth, RNG& rng) override
+    {
+        BaseElemWiseOp::generateScalars(depth, rng);
+        isLast = (randInt(rng) % 2 == 0);
+        isMax = (randInt(rng) % 2 == 0);
+        axis = randInt(rng);
+    }
+    int getAxis(const Mat& src) const
+    {
+        int dims = src.dims;
+        return static_cast<int>(axis % (2 * dims)) - dims; // [-dims; dims - 1]
+    }
+    void op(const vector<Mat>& src, Mat& dst, const Mat&) override
+    {
+        const Mat& inp = src[0];
+        const int axis_ = getAxis(inp);
+        if (isMax)
+        {
+            cv::reduceArgMax(inp, dst, axis_, isLast);
+        }
+        else
+        {
+            cv::reduceArgMin(inp, dst, axis_, isLast);
+        }
+    }
+    void refop(const vector<Mat>& src, Mat& dst, const Mat&) override
+    {
+        const Mat& inp = src[0];
+        const int axis_ = getAxis(inp);
+
+        if (!isLast && !isMax)
+        {
+            cvtest::MinMaxReducer<std::less>::reduce(inp, dst, axis_);
+        }
+        else if (!isLast && isMax)
+        {
+            cvtest::MinMaxReducer<std::greater>::reduce(inp, dst, axis_);
+        }
+        else if (isLast && !isMax)
+        {
+            cvtest::MinMaxReducer<std::less_equal>::reduce(inp, dst, axis_);
+        }
+        else
+        {
+            cvtest::MinMaxReducer<std::greater_equal>::reduce(inp, dst, axis_);
+        }
+    }
+
+    bool isLast;
+    bool isMax;
+    uint32_t axis;
+};
+
 
 typedef Ptr<BaseElemWiseOp> ElemWiseOpPtr;
 class ElemWiseTest : public ::testing::TestWithParam<ElemWiseOpPtr> {};
@@ -1492,6 +1560,7 @@ INSTANTIATE_TEST_CASE_P(Core_MeanStdDev, ElemWiseTest, ::testing::Values(ElemWis
 INSTANTIATE_TEST_CASE_P(Core_Sum, ElemWiseTest, ::testing::Values(ElemWiseOpPtr(new SumOp)));
 INSTANTIATE_TEST_CASE_P(Core_Norm, ElemWiseTest, ::testing::Values(ElemWiseOpPtr(new NormOp)));
 INSTANTIATE_TEST_CASE_P(Core_MinMaxLoc, ElemWiseTest, ::testing::Values(ElemWiseOpPtr(new MinMaxLocOp)));
+INSTANTIATE_TEST_CASE_P(Core_reduceArgMinMax, ElemWiseTest, ::testing::Values(ElemWiseOpPtr(new reduceArgMinMaxOp)));
 INSTANTIATE_TEST_CASE_P(Core_CartToPolarToCart, ElemWiseTest, ::testing::Values(ElemWiseOpPtr(new CartToPolarToCartOp)));