diff --git a/modules/core/src/kmeans.cpp b/modules/core/src/kmeans.cpp
index bca8b9bd79..544882ef1d 100644
--- a/modules/core/src/kmeans.cpp
+++ b/modules/core/src/kmeans.cpp
@@ -143,6 +143,8 @@ static void generateCentersPP(const Mat& data, Mat& _out_centers,
                 std::swap(tdist, tdist2);
             }
         }
+        if (bestCenter < 0)
+            CV_Error(Error::StsNoConv, "kmeans: can't update cluster center (check input for huge or NaN values)");
         centers[k] = bestCenter;
         sum0 = bestSum;
         std::swap(dist, tdist);
diff --git a/modules/core/test/test_math.cpp b/modules/core/test/test_math.cpp
index 8b13a391cb..066475b19e 100644
--- a/modules/core/test/test_math.cpp
+++ b/modules/core/test/test_math.cpp
@@ -2949,6 +2949,34 @@ TEST(Core_KMeans, compactness)
     }
 }
 
+TEST(Core_KMeans, bad_input)
+{
+    const int N = 100;
+    const int attempts = 4;
+    const TermCriteria crit = TermCriteria(TermCriteria::COUNT, 5, 0); // low number of iterations
+    const int K = 3;
+    Mat data(N, 1, CV_32FC2);
+    cv::randu(data, Scalar(-200, -200), Scalar(200, 200));
+    {
+        SCOPED_TRACE("Huge value");
+        data.at<Vec2f>(10, 0) = Vec2f(1e20f, 0);
+        Mat labels, centers;
+        EXPECT_ANY_THROW(kmeans(data, K, labels, crit, attempts, KMEANS_PP_CENTERS, centers));
+    }
+    {
+        SCOPED_TRACE("Negative value");
+        data.at<Vec2f>(10, 0) = Vec2f(0, -1e20f);
+        Mat labels, centers;
+        EXPECT_ANY_THROW(kmeans(data, K, labels, crit, attempts, KMEANS_PP_CENTERS, centers));
+    }
+    {
+        SCOPED_TRACE("NaN");
+        data.at<Vec2f>(10, 0) = Vec2f(0, std::numeric_limits<float>::quiet_NaN());
+        Mat labels, centers;
+        EXPECT_ANY_THROW(kmeans(data, K, labels, crit, attempts, KMEANS_PP_CENTERS, centers));
+    }
+}
+
 TEST(CovariationMatrixVectorOfMat, accuracy)
 {
     unsigned int col_problem_size = 8, row_problem_size = 8, vector_size = 16;