diff --git a/modules/video/include/opencv2/video/tracking.hpp b/modules/video/include/opencv2/video/tracking.hpp
index 8dbcfbf216..df34a9f97c 100644
--- a/modules/video/include/opencv2/video/tracking.hpp
+++ b/modules/video/include/opencv2/video/tracking.hpp
@@ -564,6 +564,12 @@ public:
     /** @copybrief getGamma @see getGamma */
     CV_WRAP virtual void setGamma(float val) = 0;
 
+    /** @brief Norm value shift for robust penalizer
+    @see setEpsilon */
+    CV_WRAP virtual float getEpsilon() const = 0;
+    /** @copybrief getEpsilon @see getEpsilon */
+    CV_WRAP virtual void setEpsilon(float val) = 0;
+
     /** @brief Creates an instance of VariationalRefinement
     */
     CV_WRAP static Ptr<VariationalRefinement> create();
@@ -645,6 +651,12 @@ public:
     /** @copybrief getVariationalRefinementGamma @see getVariationalRefinementGamma */
     CV_WRAP virtual void setVariationalRefinementGamma(float val) = 0;
 
+    /** @brief Norm value shift for robust penalizer
+    @see setVariationalRefinementEpsilon */
+    CV_WRAP virtual float getVariationalRefinementEpsilon() const = 0;
+    /** @copybrief getVariationalRefinementEpsilon @see getVariationalRefinementEpsilon */
+    CV_WRAP virtual void setVariationalRefinementEpsilon(float val) = 0;
+
 
     /** @brief Whether to use mean-normalization of patches when computing patch distance. It is turned on
         by default as it typically provides a noticeable quality boost because of increased robustness to
diff --git a/modules/video/src/dis_flow.cpp b/modules/video/src/dis_flow.cpp
index 40ac4517a4..75090d093d 100644
--- a/modules/video/src/dis_flow.cpp
+++ b/modules/video/src/dis_flow.cpp
@@ -67,6 +67,7 @@ class DISOpticalFlowImpl CV_FINAL : public DISOpticalFlow
     float variational_refinement_alpha;
     float variational_refinement_gamma;
     float variational_refinement_delta;
+    float variational_refinement_epsilon;
     bool use_mean_normalization;
     bool use_spatial_propagation;
 
@@ -92,6 +93,8 @@ class DISOpticalFlowImpl CV_FINAL : public DISOpticalFlow
     void setVariationalRefinementDelta(float val) CV_OVERRIDE { variational_refinement_delta = val; }
     float getVariationalRefinementGamma() const CV_OVERRIDE { return variational_refinement_gamma; }
     void setVariationalRefinementGamma(float val) CV_OVERRIDE { variational_refinement_gamma = val; }
+    float getVariationalRefinementEpsilon() const CV_OVERRIDE { return variational_refinement_epsilon; }
+    void setVariationalRefinementEpsilon(float val) CV_OVERRIDE { variational_refinement_epsilon = val; }
 
     bool getUseMeanNormalization() const CV_OVERRIDE { return use_mean_normalization; }
     void setUseMeanNormalization(bool val) CV_OVERRIDE { use_mean_normalization = val; }
@@ -219,6 +222,7 @@ DISOpticalFlowImpl::DISOpticalFlowImpl()
     variational_refinement_alpha = 20.f;
     variational_refinement_gamma = 10.f;
     variational_refinement_delta = 5.f;
+    variational_refinement_epsilon = 0.01f;
 
     border_size = 16;
     use_mean_normalization = true;
@@ -306,6 +310,7 @@ void DISOpticalFlowImpl::prepareBuffers(Mat &I0, Mat &I1, Mat &flow, bool use_fl
             variational_refinement_processors[i]->setAlpha(variational_refinement_alpha);
             variational_refinement_processors[i]->setDelta(variational_refinement_delta);
             variational_refinement_processors[i]->setGamma(variational_refinement_gamma);
+            variational_refinement_processors[i]->setEpsilon(variational_refinement_epsilon);
             variational_refinement_processors[i]->setSorIterations(5);
             variational_refinement_processors[i]->setFixedPointIterations(variational_refinement_iter);
 
@@ -1274,6 +1279,7 @@ void DISOpticalFlowImpl::ocl_prepareBuffers(UMat &I0, UMat &I1, InputArray flow,
             variational_refinement_processors[i]->setAlpha(variational_refinement_alpha);
             variational_refinement_processors[i]->setDelta(variational_refinement_delta);
             variational_refinement_processors[i]->setGamma(variational_refinement_gamma);
+            variational_refinement_processors[i]->setEpsilon(variational_refinement_epsilon);
             variational_refinement_processors[i]->setSorIterations(5);
             variational_refinement_processors[i]->setFixedPointIterations(variational_refinement_iter);
 
diff --git a/modules/video/src/variational_refinement.cpp b/modules/video/src/variational_refinement.cpp
index 968bce6717..b1891c60df 100644
--- a/modules/video/src/variational_refinement.cpp
+++ b/modules/video/src/variational_refinement.cpp
@@ -76,6 +76,8 @@ class VariationalRefinementImpl CV_FINAL : public VariationalRefinement
     void setDelta(float val) CV_OVERRIDE { delta = val; }
     float getGamma() const CV_OVERRIDE { return gamma; }
     void setGamma(float val) CV_OVERRIDE { gamma = val; }
+    float getEpsilon() const CV_OVERRIDE { return epsilon; }
+    void setEpsilon(float val) CV_OVERRIDE { epsilon = val; }
 
   protected: //!< internal buffers
     /* This struct defines a special data layout for Mat_<float>. Original buffer is split into two: one for "red"