diff --git a/modules/photo/src/fast_nlmeans_denoising_invoker.hpp b/modules/photo/src/fast_nlmeans_denoising_invoker.hpp
index 01588b03db..2ebf76af40 100644
--- a/modules/photo/src/fast_nlmeans_denoising_invoker.hpp
+++ b/modules/photo/src/fast_nlmeans_denoising_invoker.hpp
@@ -80,15 +80,15 @@ private:
     std::vector<IT> almost_dist2weight_;
 
     void calcDistSumsForFirstElementInRow(
-        int i, Array2d<IT>& dist_sums,
-        Array3d<IT>& col_dist_sums,
-        Array3d<IT>& up_col_dist_sums) const;
+        int i, Array2d<int>& dist_sums,
+        Array3d<int>& col_dist_sums,
+        Array3d<int>& up_col_dist_sums) const;
 
     void calcDistSumsForElementInFirstRow(
         int i, int j, int first_col_num,
-        Array2d<IT>& dist_sums,
-        Array3d<IT>& col_dist_sums,
-        Array3d<IT>& up_col_dist_sums) const;
+        Array2d<int>& dist_sums,
+        Array3d<int>& col_dist_sums,
+        Array3d<int>& up_col_dist_sums) const;
 };
 
 inline int getNearestPowerOf2(int value)
@@ -128,8 +128,8 @@ FastNlMeansDenoisingInvoker<T, IT, UIT, D>::FastNlMeansDenoisingInvoker(
     almost_template_window_size_sq_bin_shift_ = getNearestPowerOf2(template_window_size_sq);
     double almost_dist2actual_dist_multiplier = ((double)(1 << almost_template_window_size_sq_bin_shift_)) / template_window_size_sq;
 
-    IT max_dist = D::template maxDist<T, IT>();
-    size_t almost_max_dist = (size_t)(max_dist / almost_dist2actual_dist_multiplier + 1);
+    int max_dist = D::template maxDist<T>();
+    int almost_max_dist = (int)(max_dist / almost_dist2actual_dist_multiplier + 1);
     almost_dist2weight_.resize(almost_max_dist);
 
     const double WEIGHT_THRESHOLD = 0.001;
@@ -156,14 +156,14 @@ void FastNlMeansDenoisingInvoker<T, IT, UIT, D>::operator() (const Range& range)
     int row_to = range.end - 1;
 
     // sums of cols anf rows for current pixel p
-    Array2d<IT> dist_sums(search_window_size_, search_window_size_);
+    Array2d<int> dist_sums(search_window_size_, search_window_size_);
 
     // for lazy calc optimization (sum of cols for current pixel)
-    Array3d<IT> col_dist_sums(template_window_size_, search_window_size_, search_window_size_);
+    Array3d<int> col_dist_sums(template_window_size_, search_window_size_, search_window_size_);
 
     int first_col_num = -1;
     // last elements of column sum (for each element in row)
-    Array3d<IT> up_col_dist_sums(src_.cols, search_window_size_, search_window_size_);
+    Array3d<int> up_col_dist_sums(src_.cols, search_window_size_, search_window_size_);
 
     for (int i = row_from; i <= row_to; i++)
     {
@@ -202,9 +202,9 @@ void FastNlMeansDenoisingInvoker<T, IT, UIT, D>::operator() (const Range& range)
 
                     for (int y = 0; y < search_window_size; y++)
                     {
-                        IT * dist_sums_row = dist_sums.row_ptr(y);
-                        IT * col_dist_sums_row = col_dist_sums.row_ptr(first_col_num, y);
-                        IT * up_col_dist_sums_row = up_col_dist_sums.row_ptr(j, y);
+                        int * dist_sums_row = dist_sums.row_ptr(y);
+                        int * col_dist_sums_row = col_dist_sums.row_ptr(first_col_num, y);
+                        int * up_col_dist_sums_row = up_col_dist_sums.row_ptr(j, y);
 
                         const T * b_up_ptr = extended_src_.ptr<T>(start_by - template_window_half_size_ - 1 + y);
                         const T * b_down_ptr = extended_src_.ptr<T>(start_by + template_window_half_size_ + y);
@@ -215,7 +215,7 @@ void FastNlMeansDenoisingInvoker<T, IT, UIT, D>::operator() (const Range& range)
                             dist_sums_row[x] -= col_dist_sums_row[x];
 
                             int bx = start_bx + x;
-                            col_dist_sums_row[x] = up_col_dist_sums_row[x] + D::template calcUpDownDist<T, IT>(a_up, a_down, b_up_ptr[bx], b_down_ptr[bx]);
+                            col_dist_sums_row[x] = up_col_dist_sums_row[x] + D::template calcUpDownDist<T>(a_up, a_down, b_up_ptr[bx], b_down_ptr[bx]);
 
                             dist_sums_row[x] += col_dist_sums_row[x];
                             up_col_dist_sums_row[x] = col_dist_sums_row[x];
@@ -234,10 +234,10 @@ void FastNlMeansDenoisingInvoker<T, IT, UIT, D>::operator() (const Range& range)
             for (int y = 0; y < search_window_size_; y++)
             {
                 const T* cur_row_ptr = extended_src_.ptr<T>(border_size_ + search_window_y + y);
-                IT* dist_sums_row = dist_sums.row_ptr(y);
+                int* dist_sums_row = dist_sums.row_ptr(y);
                 for (int x = 0; x < search_window_size_; x++)
                 {
-                    size_t almostAvgDist = (size_t)(dist_sums_row[x] >> almost_template_window_size_sq_bin_shift_);
+                    int almostAvgDist = dist_sums_row[x] >> almost_template_window_size_sq_bin_shift_;
                     IT weight = almost_dist2weight_[almostAvgDist];
                     weights_sum += weight;
 
@@ -257,9 +257,9 @@ void FastNlMeansDenoisingInvoker<T, IT, UIT, D>::operator() (const Range& range)
 template <typename T, typename IT, typename UIT, typename D>
 inline void FastNlMeansDenoisingInvoker<T, IT, UIT, D>::calcDistSumsForFirstElementInRow(
     int i,
-    Array2d<IT>& dist_sums,
-    Array3d<IT>& col_dist_sums,
-    Array3d<IT>& up_col_dist_sums) const
+    Array2d<int>& dist_sums,
+    Array3d<int>& col_dist_sums,
+    Array3d<int>& up_col_dist_sums) const
 {
     int j = 0;
 
@@ -276,7 +276,7 @@ inline void FastNlMeansDenoisingInvoker<T, IT, UIT, D>::calcDistSumsForFirstElem
             for (int ty = -template_window_half_size_; ty <= template_window_half_size_; ty++)
                 for (int tx = -template_window_half_size_; tx <= template_window_half_size_; tx++)
                 {
-                    int dist = D::template calcDist<T, IT>(extended_src_,
+                    int dist = D::template calcDist<T>(extended_src_,
                         border_size_ + i + ty, border_size_ + j + tx,
                         border_size_ + start_y + ty, border_size_ + start_x + tx);
 
@@ -291,9 +291,9 @@ inline void FastNlMeansDenoisingInvoker<T, IT, UIT, D>::calcDistSumsForFirstElem
 template <typename T, typename IT, typename UIT, typename D>
 inline void FastNlMeansDenoisingInvoker<T, IT, UIT, D>::calcDistSumsForElementInFirstRow(
     int i, int j, int first_col_num,
-    Array2d<IT>& dist_sums,
-    Array3d<IT>& col_dist_sums,
-    Array3d<IT>& up_col_dist_sums) const
+    Array2d<int>& dist_sums,
+    Array3d<int>& col_dist_sums,
+    Array3d<int>& up_col_dist_sums) const
 {
     int ay = border_size_ + i;
     int ax = border_size_ + j + template_window_half_size_;
@@ -312,7 +312,7 @@ inline void FastNlMeansDenoisingInvoker<T, IT, UIT, D>::calcDistSumsForElementIn
             int by = start_by + y;
             int bx = start_bx + x;
             for (int ty = -template_window_half_size_; ty <= template_window_half_size_; ty++)
-                col_dist_sums[new_last_col_num][y][x] += D::template calcDist<T,IT>(extended_src_, ay + ty, ax, by + ty, bx);
+                col_dist_sums[new_last_col_num][y][x] += D::template calcDist<T>(extended_src_, ay + ty, ax, by + ty, bx);
 
             dist_sums[y][x] += col_dist_sums[new_last_col_num][y][x];
             up_col_dist_sums[j][y][x] = col_dist_sums[new_last_col_num][y][x];
diff --git a/modules/photo/src/fast_nlmeans_denoising_invoker_commons.hpp b/modules/photo/src/fast_nlmeans_denoising_invoker_commons.hpp
index d77ca3e1f7..dbb4c5eb33 100644
--- a/modules/photo/src/fast_nlmeans_denoising_invoker_commons.hpp
+++ b/modules/photo/src/fast_nlmeans_denoising_invoker_commons.hpp
@@ -83,63 +83,63 @@ template <typename T> struct pixelInfo: public pixelInfo_<T>
 
 class DistAbs
 {
-    template <typename T, typename IT> struct calcDist_
+    template <typename T> struct calcDist_
     {
-        static inline IT f(const T a, const T b)
+        static inline int f(const T a, const T b)
         {
-            return std::abs((IT)(a-b));
+            return std::abs((int)(a-b));
         }
     };
 
-    template <typename ET, typename IT> struct calcDist_<Vec<ET, 2>, IT>
+    template <typename ET> struct calcDist_<Vec<ET, 2> >
     {
-        static inline IT f(const Vec<ET, 2> a, const Vec<ET, 2> b)
+        static inline int f(const Vec<ET, 2> a, const Vec<ET, 2> b)
         {
-            return std::abs((IT)(a[0]-b[0])) + std::abs((IT)(a[1]-b[1]));
+            return std::abs((int)(a[0]-b[0])) + std::abs((int)(a[1]-b[1]));
         }
     };
 
-    template <typename ET, typename IT> struct calcDist_<Vec<ET, 3>, IT>
+    template <typename ET> struct calcDist_<Vec<ET, 3> >
     {
-        static inline IT f(const Vec<ET, 3> a, const Vec<ET, 3> b)
+        static inline int f(const Vec<ET, 3> a, const Vec<ET, 3> b)
         {
             return
-                std::abs((IT)(a[0]-b[0])) +
-                std::abs((IT)(a[1]-b[1])) +
-                std::abs((IT)(a[2]-b[2]));
+                std::abs((int)(a[0]-b[0])) +
+                std::abs((int)(a[1]-b[1])) +
+                std::abs((int)(a[2]-b[2]));
         }
     };
 
-    template <typename ET, typename IT> struct calcDist_<Vec<ET, 4>, IT>
+    template <typename ET> struct calcDist_<Vec<ET, 4> >
     {
-        static inline IT f(const Vec<ET, 4> a, const Vec<ET, 4> b)
+        static inline int f(const Vec<ET, 4> a, const Vec<ET, 4> b)
         {
             return
-                std::abs((IT)(a[0]-b[0])) +
-                std::abs((IT)(a[1]-b[1])) +
-                std::abs((IT)(a[2]-b[2])) +
-                std::abs((IT)(a[3]-b[3]));
+                std::abs((int)(a[0]-b[0])) +
+                std::abs((int)(a[1]-b[1])) +
+                std::abs((int)(a[2]-b[2])) +
+                std::abs((int)(a[3]-b[3]));
         }
     };
 
 public:
-    template <typename T, typename IT> static inline IT calcDist(const T a, const T b)
+    template <typename T> static inline int calcDist(const T a, const T b)
     {
-        return calcDist_<T, IT>::f(a, b);
+        return calcDist_<T>::f(a, b);
     }
 
-    template <typename T, typename IT>
-    static inline IT calcDist(const Mat& m, int i1, int j1, int i2, int j2)
+    template <typename T>
+    static inline int calcDist(const Mat& m, int i1, int j1, int i2, int j2)
     {
         const T a = m.at<T>(i1, j1);
         const T b = m.at<T>(i2, j2);
-        return calcDist<T, IT>(a,b);
+        return calcDist<T>(a,b);
     }
 
-    template <typename T, typename IT>
-    static inline IT calcUpDownDist(T a_up, T a_down, T b_up, T b_down)
+    template <typename T>
+    static inline int calcUpDownDist(T a_up, T a_down, T b_up, T b_down)
     {
-        return calcDist<T, IT>(a_down, b_down) - calcDist<T, IT>(a_up, b_up);
+        return calcDist<T>(a_down, b_down) - calcDist<T>(a_up, b_up);
     };
 
     template <typename T>
@@ -148,93 +148,93 @@ public:
         return std::exp(-dist*dist / (h * h * pixelInfo<T>::channels));
     }
 
-    template <typename T, typename IT>
+    template <typename T>
     static double maxDist()
     {
-        return (IT)pixelInfo<T>::sampleMax() * (IT)pixelInfo<T>::channels;
+        return (int)pixelInfo<T>::sampleMax() * pixelInfo<T>::channels;
     }
 };
 
 class DistSquared
 {
-    template <typename T, typename IT> struct calcDist_
+    template <typename T> struct calcDist_
     {
-        static inline IT f(const T a, const T b)
+        static inline int f(const T a, const T b)
         {
-            return (IT)(a-b) * (IT)(a-b);
+            return (int)(a-b) * (int)(a-b);
         }
     };
 
-    template <typename ET, typename IT> struct calcDist_<Vec<ET, 2>, IT>
+    template <typename ET> struct calcDist_<Vec<ET, 2> >
     {
-        static inline IT f(const Vec<ET, 2> a, const Vec<ET, 2> b)
+        static inline int f(const Vec<ET, 2> a, const Vec<ET, 2> b)
         {
-            return (IT)(a[0]-b[0])*(IT)(a[0]-b[0]) + (IT)(a[1]-b[1])*(IT)(a[1]-b[1]);
+            return (int)(a[0]-b[0])*(int)(a[0]-b[0]) + (int)(a[1]-b[1])*(int)(a[1]-b[1]);
         }
     };
 
-    template <typename ET, typename IT> struct calcDist_<Vec<ET, 3>, IT>
+    template <typename ET> struct calcDist_<Vec<ET, 3> >
     {
-        static inline IT f(const Vec<ET, 3> a, const Vec<ET, 3> b)
+        static inline int f(const Vec<ET, 3> a, const Vec<ET, 3> b)
         {
             return
-                (IT)(a[0]-b[0])*(IT)(a[0]-b[0]) +
-                (IT)(a[1]-b[1])*(IT)(a[1]-b[1]) +
-                (IT)(a[2]-b[2])*(IT)(a[2]-b[2]);
+                (int)(a[0]-b[0])*(int)(a[0]-b[0]) +
+                (int)(a[1]-b[1])*(int)(a[1]-b[1]) +
+                (int)(a[2]-b[2])*(int)(a[2]-b[2]);
         }
     };
 
-    template <typename ET, typename IT> struct calcDist_<Vec<ET, 4>, IT>
+    template <typename ET> struct calcDist_<Vec<ET, 4> >
     {
-        static inline IT f(const Vec<ET, 4> a, const Vec<ET, 4> b)
+        static inline int f(const Vec<ET, 4> a, const Vec<ET, 4> b)
         {
             return
-                (IT)(a[0]-b[0])*(IT)(a[0]-b[0]) +
-                (IT)(a[1]-b[1])*(IT)(a[1]-b[1]) +
-                (IT)(a[2]-b[2])*(IT)(a[2]-b[2]) +
-                (IT)(a[3]-b[3])*(IT)(a[3]-b[3]);
+                (int)(a[0]-b[0])*(int)(a[0]-b[0]) +
+                (int)(a[1]-b[1])*(int)(a[1]-b[1]) +
+                (int)(a[2]-b[2])*(int)(a[2]-b[2]) +
+                (int)(a[3]-b[3])*(int)(a[3]-b[3]);
         }
     };
 
-    template <typename T, typename IT> struct calcUpDownDist_
+    template <typename T> struct calcUpDownDist_
     {
-        static inline IT f(T a_up, T a_down, T b_up, T b_down)
+        static inline int f(T a_up, T a_down, T b_up, T b_down)
         {
-            IT A = a_down - b_down;
-            IT B = a_up - b_up;
+            int A = a_down - b_down;
+            int B = a_up - b_up;
             return (A-B)*(A+B);
         }
     };
 
-    template <typename ET, int n, typename IT> struct calcUpDownDist_<Vec<ET, n>, IT>
+    template <typename ET, int n> struct calcUpDownDist_<Vec<ET, n> >
     {
     private:
         typedef Vec<ET, n> T;
     public:
-        static inline IT f(T a_up, T a_down, T b_up, T b_down)
+        static inline int f(T a_up, T a_down, T b_up, T b_down)
         {
-            return calcDist<T, IT>(a_down, b_down) - calcDist<T, IT>(a_up, b_up);
+            return calcDist<T>(a_down, b_down) - calcDist<T>(a_up, b_up);
         }
     };
 
 public:
-    template <typename T, typename IT> static inline IT calcDist(const T a, const T b)
+    template <typename T> static inline int calcDist(const T a, const T b)
     {
-        return calcDist_<T, IT>::f(a, b);
+        return calcDist_<T>::f(a, b);
     }
 
-    template <typename T, typename IT>
-    static inline IT calcDist(const Mat& m, int i1, int j1, int i2, int j2)
+    template <typename T>
+    static inline int calcDist(const Mat& m, int i1, int j1, int i2, int j2)
     {
         const T a = m.at<T>(i1, j1);
         const T b = m.at<T>(i2, j2);
-        return calcDist<T, IT>(a,b);
+        return calcDist<T>(a,b);
     }
 
-    template <typename T, typename IT>
-    static inline IT calcUpDownDist(T a_up, T a_down, T b_up, T b_down)
+    template <typename T>
+    static inline int calcUpDownDist(T a_up, T a_down, T b_up, T b_down)
     {
-        return calcUpDownDist_<T, IT>::f(a_up, a_down, b_up, b_down);
+        return calcUpDownDist_<T>::f(a_up, a_down, b_up, b_down);
     };
 
     template <typename T>
@@ -243,11 +243,11 @@ public:
         return std::exp(-dist / (h * h * pixelInfo<T>::channels));
     }
 
-    template <typename T, typename IT>
+    template <typename T>
     static double maxDist()
     {
-        return (IT)pixelInfo<T>::sampleMax() * (IT)pixelInfo<T>::sampleMax() *
-            (IT)pixelInfo<T>::channels;
+        return (int)pixelInfo<T>::sampleMax() * (int)pixelInfo<T>::sampleMax() *
+            pixelInfo<T>::channels;
     }
 };
 
diff --git a/modules/photo/src/fast_nlmeans_multi_denoising_invoker.hpp b/modules/photo/src/fast_nlmeans_multi_denoising_invoker.hpp
index eb20786435..f1a334040c 100644
--- a/modules/photo/src/fast_nlmeans_multi_denoising_invoker.hpp
+++ b/modules/photo/src/fast_nlmeans_multi_denoising_invoker.hpp
@@ -85,13 +85,13 @@ private:
     int almost_template_window_size_sq_bin_shift;
     std::vector<IT> almost_dist2weight;
 
-    void calcDistSumsForFirstElementInRow(int i, Array3d<IT>& dist_sums,
-                                          Array4d<IT>& col_dist_sums,
-                                          Array4d<IT>& up_col_dist_sums) const;
+    void calcDistSumsForFirstElementInRow(int i, Array3d<int>& dist_sums,
+                                          Array4d<int>& col_dist_sums,
+                                          Array4d<int>& up_col_dist_sums) const;
 
     void calcDistSumsForElementInFirstRow(int i, int j, int first_col_num,
-                                          Array3d<IT>& dist_sums, Array4d<IT>& col_dist_sums,
-                                          Array4d<IT>& up_col_dist_sums) const;
+                                          Array3d<int>& dist_sums, Array4d<int>& col_dist_sums,
+                                          Array4d<int>& up_col_dist_sums) const;
 };
 
 template <typename T, typename IT, typename UIT, typename D>
@@ -139,8 +139,8 @@ FastNlMeansMultiDenoisingInvoker<T, IT, UIT, D>::FastNlMeansMultiDenoisingInvoke
     int almost_template_window_size_sq = 1 << almost_template_window_size_sq_bin_shift;
     double almost_dist2actual_dist_multiplier = (double) almost_template_window_size_sq / template_window_size_sq;
 
-    IT max_dist = D::template maxDist<T,IT>();
-    int almost_max_dist = (int) (max_dist / almost_dist2actual_dist_multiplier + 1);
+    int max_dist = D::template maxDist<T>();
+    int almost_max_dist = (int)(max_dist / almost_dist2actual_dist_multiplier + 1);
     almost_dist2weight.resize(almost_max_dist);
 
     const double WEIGHT_THRESHOLD = 0.001;
@@ -166,13 +166,13 @@ void FastNlMeansMultiDenoisingInvoker<T, IT, UIT, D>::operator() (const Range& r
     int row_from = range.start;
     int row_to = range.end - 1;
 
-    Array3d<IT> dist_sums(temporal_window_size_, search_window_size_, search_window_size_);
+    Array3d<int> dist_sums(temporal_window_size_, search_window_size_, search_window_size_);
 
     // for lazy calc optimization
-    Array4d<IT> col_dist_sums(template_window_size_, temporal_window_size_, search_window_size_, search_window_size_);
+    Array4d<int> col_dist_sums(template_window_size_, temporal_window_size_, search_window_size_, search_window_size_);
 
     int first_col_num = -1;
-    Array4d<IT> up_col_dist_sums(cols_, temporal_window_size_, search_window_size_, search_window_size_);
+    Array4d<int> up_col_dist_sums(cols_, temporal_window_size_, search_window_size_, search_window_size_);
 
     for (int i = row_from; i <= row_to; i++)
     {
@@ -216,15 +216,15 @@ void FastNlMeansMultiDenoisingInvoker<T, IT, UIT, D>::operator() (const Range& r
                     for (int d = 0; d < temporal_window_size_; d++)
                     {
                         Mat cur_extended_src = extended_srcs_[d];
-                        Array2d<IT> cur_dist_sums = dist_sums[d];
-                        Array2d<IT> cur_col_dist_sums = col_dist_sums[first_col_num][d];
-                        Array2d<IT> cur_up_col_dist_sums = up_col_dist_sums[j][d];
+                        Array2d<int> cur_dist_sums = dist_sums[d];
+                        Array2d<int> cur_col_dist_sums = col_dist_sums[first_col_num][d];
+                        Array2d<int> cur_up_col_dist_sums = up_col_dist_sums[j][d];
                         for (int y = 0; y < search_window_size; y++)
                         {
-                            IT* dist_sums_row = cur_dist_sums.row_ptr(y);
+                            int* dist_sums_row = cur_dist_sums.row_ptr(y);
 
-                            IT* col_dist_sums_row = cur_col_dist_sums.row_ptr(y);
-                            IT* up_col_dist_sums_row = cur_up_col_dist_sums.row_ptr(y);
+                            int* col_dist_sums_row = cur_col_dist_sums.row_ptr(y);
+                            int* up_col_dist_sums_row = cur_up_col_dist_sums.row_ptr(y);
 
                             const T* b_up_ptr = cur_extended_src.ptr<T>(start_by - template_window_half_size_ - 1 + y);
                             const T* b_down_ptr = cur_extended_src.ptr<T>(start_by + template_window_half_size_ + y);
@@ -234,7 +234,7 @@ void FastNlMeansMultiDenoisingInvoker<T, IT, UIT, D>::operator() (const Range& r
                                 dist_sums_row[x] -= col_dist_sums_row[x];
 
                                 col_dist_sums_row[x] = up_col_dist_sums_row[x] +
-                                    D::template calcUpDownDist<T, IT>(a_up, a_down, b_up_ptr[start_bx + x], b_down_ptr[start_bx + x]);
+                                    D::template calcUpDownDist<T>(a_up, a_down, b_up_ptr[start_bx + x], b_down_ptr[start_bx + x]);
 
                                 dist_sums_row[x] += col_dist_sums_row[x];
                                 up_col_dist_sums_row[x] = col_dist_sums_row[x];
@@ -260,11 +260,11 @@ void FastNlMeansMultiDenoisingInvoker<T, IT, UIT, D>::operator() (const Range& r
                 {
                     const T* cur_row_ptr = esrc_d.ptr<T>(border_size_ + search_window_y + y);
 
-                    IT* dist_sums_row = dist_sums.row_ptr(d, y);
+                    int* dist_sums_row = dist_sums.row_ptr(d, y);
 
                     for (int x = 0; x < search_window_size_; x++)
                     {
-                        size_t almostAvgDist = (size_t)(dist_sums_row[x] >> almost_template_window_size_sq_bin_shift);
+                        int almostAvgDist = dist_sums_row[x] >> almost_template_window_size_sq_bin_shift;
 
                         IT weight =  almost_dist2weight[almostAvgDist];
                         weights_sum += weight;
@@ -286,7 +286,7 @@ void FastNlMeansMultiDenoisingInvoker<T, IT, UIT, D>::operator() (const Range& r
 
 template <typename T, typename IT, typename UIT, typename D>
 inline void FastNlMeansMultiDenoisingInvoker<T, IT, UIT, D>::calcDistSumsForFirstElementInRow(
-        int i, Array3d<IT>& dist_sums, Array4d<IT>& col_dist_sums, Array4d<IT>& up_col_dist_sums) const
+        int i, Array3d<int>& dist_sums, Array4d<int>& col_dist_sums, Array4d<int>& up_col_dist_sums) const
 {
     int j = 0;
 
@@ -303,14 +303,14 @@ inline void FastNlMeansMultiDenoisingInvoker<T, IT, UIT, D>::calcDistSumsForFirs
                 int start_y = i + y - search_window_half_size_;
                 int start_x = j + x - search_window_half_size_;
 
-                IT* dist_sums_ptr = &dist_sums[d][y][x];
-                IT* col_dist_sums_ptr = &col_dist_sums[0][d][y][x];
+                int* dist_sums_ptr = &dist_sums[d][y][x];
+                int* col_dist_sums_ptr = &col_dist_sums[0][d][y][x];
                 int col_dist_sums_step = col_dist_sums.step_size(0);
                 for (int tx = -template_window_half_size_; tx <= template_window_half_size_; tx++)
                 {
                     for (int ty = -template_window_half_size_; ty <= template_window_half_size_; ty++)
                     {
-                        IT dist = D::template calcDist<T, IT>(
+                        int dist = D::template calcDist<T>(
                                     main_extended_src_.at<T>(border_size_ + i + ty, border_size_ + j + tx),
                                     cur_extended_src.at<T>(border_size_ + start_y + ty, border_size_ + start_x + tx));
 
@@ -327,8 +327,8 @@ inline void FastNlMeansMultiDenoisingInvoker<T, IT, UIT, D>::calcDistSumsForFirs
 
 template <typename T, typename IT, typename UIT, typename D>
 inline void FastNlMeansMultiDenoisingInvoker<T, IT, UIT, D>::calcDistSumsForElementInFirstRow(
-    int i, int j, int first_col_num, Array3d<IT>& dist_sums,
-    Array4d<IT>& col_dist_sums, Array4d<IT>& up_col_dist_sums) const
+    int i, int j, int first_col_num, Array3d<int>& dist_sums,
+    Array4d<int>& col_dist_sums, Array4d<int>& up_col_dist_sums) const
 {
     int ay = border_size_ + i;
     int ax = border_size_ + j + template_window_half_size_;
@@ -350,10 +350,10 @@ inline void FastNlMeansMultiDenoisingInvoker<T, IT, UIT, D>::calcDistSumsForElem
                 int by = start_by + y;
                 int bx = start_bx + x;
 
-                IT* col_dist_sums_ptr = &col_dist_sums[new_last_col_num][d][y][x];
+                int* col_dist_sums_ptr = &col_dist_sums[new_last_col_num][d][y][x];
                 for (int ty = -template_window_half_size_; ty <= template_window_half_size_; ty++)
                 {
-                    *col_dist_sums_ptr += D::template calcDist<T, IT>(
+                    *col_dist_sums_ptr += D::template calcDist<T>(
                                 main_extended_src_.at<T>(ay + ty, ax),
                                 cur_extended_src.at<T>(by + ty, bx));
                 }