diff --git a/modules/imgproc/perf/perf_matchTemplate.cpp b/modules/imgproc/perf/perf_matchTemplate.cpp
new file mode 100644
index 0000000000..0c727d5a48
--- /dev/null
+++ b/modules/imgproc/perf/perf_matchTemplate.cpp
@@ -0,0 +1,81 @@
+#include "perf_precomp.hpp"
+
+using namespace std;
+using namespace cv;
+using namespace perf;
+using std::tr1::make_tuple;
+using std::tr1::get;
+
+CV_ENUM(MethodType, CV_TM_SQDIFF, CV_TM_SQDIFF_NORMED, CV_TM_CCORR, CV_TM_CCORR_NORMED, CV_TM_CCOEFF, CV_TM_CCOEFF_NORMED)
+
+typedef std::tr1::tuple<Size, Size, MethodType> ImgSize_TmplSize_Method_t;
+typedef perf::TestBaseWithParam<ImgSize_TmplSize_Method_t> ImgSize_TmplSize_Method;
+
+PERF_TEST_P(ImgSize_TmplSize_Method, matchTemplateSmall,
+            testing::Combine(
+                testing::Values(szSmall128, cv::Size(320, 240),
+                                cv::Size(640, 480), cv::Size(800, 600),
+                                cv::Size(1024, 768), cv::Size(1280, 1024)),
+                testing::Values(cv::Size(12, 12), cv::Size(28, 9),
+                                cv::Size(8, 30), cv::Size(16, 16)),
+                testing::ValuesIn(MethodType::all())
+                )
+            )
+{
+    Size imgSz = get<0>(GetParam());
+    Size tmplSz = get<1>(GetParam());
+    int method = get<2>(GetParam());
+
+    Mat img(imgSz, CV_8UC1);
+    Mat tmpl(tmplSz, CV_8UC1);
+    Mat result(imgSz - tmplSz + Size(1,1), CV_32F);
+
+    declare
+        .in(img, WARMUP_RNG)
+        .in(tmpl, WARMUP_RNG)
+        .out(result);
+
+    TEST_CYCLE() matchTemplate(img, tmpl, result, method);
+
+    bool isNormed =
+        method == CV_TM_CCORR_NORMED ||
+        method == CV_TM_SQDIFF_NORMED ||
+        method == CV_TM_CCOEFF_NORMED;
+    double eps = isNormed ? 1e-6
+        : 255 * 255 * tmpl.total() * 1e-6;
+
+    SANITY_CHECK(result, eps);
+}
+
+PERF_TEST_P(ImgSize_TmplSize_Method, matchTemplateBig,
+            testing::Combine(
+                testing::Values(cv::Size(1280, 1024)),
+                testing::Values(cv::Size(1260, 1000), cv::Size(1261, 1013)),
+                testing::ValuesIn(MethodType::all())
+                )
+    )
+{
+    Size imgSz = get<0>(GetParam());
+    Size tmplSz = get<1>(GetParam());
+    int method = get<2>(GetParam());
+
+    Mat img(imgSz, CV_8UC1);
+    Mat tmpl(tmplSz, CV_8UC1);
+    Mat result(imgSz - tmplSz + Size(1,1), CV_32F);
+
+    declare
+        .in(img, WARMUP_RNG)
+        .in(tmpl, WARMUP_RNG)
+        .out(result);
+
+    TEST_CYCLE() matchTemplate(img, tmpl, result, method);
+
+    bool isNormed =
+        method == CV_TM_CCORR_NORMED ||
+        method == CV_TM_SQDIFF_NORMED ||
+        method == CV_TM_CCOEFF_NORMED;
+    double eps = isNormed ? 1e-6
+        : 255 * 255 * tmpl.total() * 1e-6;
+
+    SANITY_CHECK(result, eps);
+}
diff --git a/modules/imgproc/src/templmatch.cpp b/modules/imgproc/src/templmatch.cpp
index 0a3b95d7c7..ec7a92a223 100644
--- a/modules/imgproc/src/templmatch.cpp
+++ b/modules/imgproc/src/templmatch.cpp
@@ -252,6 +252,11 @@ void cv::matchTemplate( InputArray _img, InputArray _templ, OutputArray _result,
     _result.create(corrSize, CV_32F);
     Mat result = _result.getMat();
 
+#ifdef HAVE_TEGRA_OPTIMIZATION
+    if (tegra::matchTemplate(img, templ, result, method))
+        return;
+#endif
+
     int cn = img.channels();
     crossCorr( img, templ, result, result.size(), result.type(), Point(0,0), 0, 0);