diff --git a/modules/gpu/src/filtering_npp.cpp b/modules/gpu/src/filtering_npp.cpp
index 8ceb44e8bd..ae97244f2a 100644
--- a/modules/gpu/src/filtering_npp.cpp
+++ b/modules/gpu/src/filtering_npp.cpp
@@ -64,8 +64,19 @@ namespace
         CV_Assert(src.type() == CV_8U || src.type() == CV_8UC4);        
         CV_Assert(kernel.type() == CV_8U && (kernel.cols & 1) != 0 && (kernel.rows & 1) != 0);
 
+        if (anchor.x == -1)
+            anchor.x = 0;
+        if (anchor.y == -1)
+            anchor.y = 0;
+
         // in NPP for Cuda 3.1 only such anchor is supported.
-        CV_Assert(anchor.x == kernel.cols/2 && anchor.y == kernel.rows/2);
+        CV_Assert(anchor.x == 0 && anchor.y == 0);
+
+        if (iterations == 0)
+        {
+            src.copyTo(dst);
+            return;
+        }
 
         const Mat& cont_krnl = (kernel.isContinuous() ? kernel : kernel.clone()).reshape(1, 1);
         GpuMat gpu_krnl(cont_krnl);
@@ -84,8 +95,9 @@ namespace
         
         dst.create(src.size(), src.type());
 
-        for(int i = 0; i < iterations; ++i)
-            nppSafeCall( func(src.ptr<Npp8u>(), src.step, dst.ptr<Npp8u>(), dst.step, sz, gpu_krnl.ptr<Npp8u>(), mask_sz, anc) );
+        nppSafeCall( func(src.ptr<Npp8u>(), src.step, dst.ptr<Npp8u>(), dst.step, sz, gpu_krnl.ptr<Npp8u>(), mask_sz, anc) );
+        for(int i = 1; i < iterations; ++i)
+            nppSafeCall( func(dst.ptr<Npp8u>(), dst.step, dst.ptr<Npp8u>(), dst.step, sz, gpu_krnl.ptr<Npp8u>(), mask_sz, anc) );
     }
 }
 
diff --git a/tests/gpu/src/gputest_main.cpp b/tests/gpu/src/gputest_main.cpp
index 463b2b7509..d4b9b3f5eb 100644
--- a/tests/gpu/src/gputest_main.cpp
+++ b/tests/gpu/src/gputest_main.cpp
@@ -47,9 +47,9 @@ const char* blacklist[] =
 {    
     "GPU-NppImageSum",              // crash
     "GPU-MatOperatorAsyncCall",     // crash
-    "GPU-NppErode",                 // npp func returns error code (CUDA_KERNEL_LAUNCH_ERROR or TEXTURE_BIND_ERROR)
-    "GPU-NppDilate",                // npp func returns error code (CUDA_KERNEL_LAUNCH_ERROR or TEXTURE_BIND_ERROR)
-    "GPU-NppMorphologyEx",          // npp func returns error code (CUDA_KERNEL_LAUNCH_ERROR or TEXTURE_BIND_ERROR)
+    //"GPU-NppErode",                 // npp func returns error code (CUDA_KERNEL_LAUNCH_ERROR or TEXTURE_BIND_ERROR)
+    //"GPU-NppDilate",                // npp func returns error code (CUDA_KERNEL_LAUNCH_ERROR or TEXTURE_BIND_ERROR)
+    //"GPU-NppMorphologyEx",          // npp func returns error code (CUDA_KERNEL_LAUNCH_ERROR or TEXTURE_BIND_ERROR)
     //"GPU-NppImageDivide",           // different round mode
     //"GPU-NppImageMeanStdDev",       // different precision
     //"GPU-NppImageMinNax",           // npp bug
diff --git a/tests/gpu/src/morf_filters.cpp b/tests/gpu/src/morf_filters.cpp
index 1f5c5a7144..f90a33fa08 100644
--- a/tests/gpu/src/morf_filters.cpp
+++ b/tests/gpu/src/morf_filters.cpp
@@ -134,7 +134,7 @@ protected:
 	virtual int test(const Mat& img)
     {
         GpuMat kernel(Mat::ones(3, 3, CV_8U));
-        Point anchor(1,1);
+        Point anchor(-1, -1);
         int iters = 3;
 
 	    cv::Mat cpuRes;
@@ -158,7 +158,7 @@ protected:
 	virtual int test(const Mat& img)
     {
         GpuMat kernel(Mat::ones(3, 3, CV_8U));
-        Point anchor(1,1);
+        Point anchor(-1, -1);
         int iters = 3;
 
 	    cv::Mat cpuRes;
@@ -186,7 +186,7 @@ protected:
         int num = sizeof(ops)/sizeof(ops[0]);
 
         GpuMat kernel(Mat::ones(3, 3, CV_8U));
-        Point anchor(1,1);
+        Point anchor(-1, -1);
         int iters = 3;
 
         for(int i = 0; i < num; ++i)