diff --git a/modules/core/include/opencv2/core/cuda.hpp b/modules/core/include/opencv2/core/cuda.hpp
index c538392cbd..4c5c522e24 100644
--- a/modules/core/include/opencv2/core/cuda.hpp
+++ b/modules/core/include/opencv2/core/cuda.hpp
@@ -327,6 +327,34 @@ The function does not reallocate memory if the matrix has proper attributes alre
  */
 CV_EXPORTS void ensureSizeIsEnough(int rows, int cols, int type, OutputArray arr);
 
+/** @brief BufferPool for use with CUDA streams
+
+ * BufferPool utilizes cuda::Stream's allocator to create new buffers. It is
+ * particularly useful when BufferPoolUsage is set to true, or a custom
+ * allocator is specified for the cuda::Stream, and you want to implement your
+ * own stream based functions utilizing the same underlying GPU memory
+ * management.
+ */
+class CV_EXPORTS BufferPool
+{
+public:
+
+    //! Gets the BufferPool for the given stream.
+    explicit BufferPool(Stream& stream);
+
+    //! Allocates a new GpuMat of given size and type.
+    GpuMat getBuffer(int rows, int cols, int type);
+
+    //! Allocates a new GpuMat of given size and type.
+    GpuMat getBuffer(Size size, int type) { return getBuffer(size.height, size.width, type); }
+
+    //! Returns the allocator associated with the stream.
+    Ptr<GpuMat::Allocator> getAllocator() const { return allocator_; }
+
+private:
+    Ptr<GpuMat::Allocator> allocator_;
+};
+
 //! BufferPool management (must be called before Stream creation)
 CV_EXPORTS void setBufferPoolUsage(bool on);
 CV_EXPORTS void setBufferPoolConfig(int deviceId, size_t stackSize, int stackCount);
diff --git a/modules/core/include/opencv2/core/private.cuda.hpp b/modules/core/include/opencv2/core/private.cuda.hpp
index 01a4ab3bf9..1214d70304 100644
--- a/modules/core/include/opencv2/core/private.cuda.hpp
+++ b/modules/core/include/opencv2/core/private.cuda.hpp
@@ -102,20 +102,6 @@ static inline void throw_no_cuda() { CV_Error(cv::Error::StsNotImplemented, "The
 
 namespace cv { namespace cuda
 {
-    class CV_EXPORTS BufferPool
-    {
-    public:
-        explicit BufferPool(Stream& stream);
-
-        GpuMat getBuffer(int rows, int cols, int type);
-        GpuMat getBuffer(Size size, int type) { return getBuffer(size.height, size.width, type); }
-
-        GpuMat::Allocator* getAllocator() const { return allocator_; }
-
-    private:
-        GpuMat::Allocator* allocator_;
-    };
-
     static inline void checkNppError(int code, const char* file, const int line, const char* func)
     {
         if (code < 0)
diff --git a/modules/core/src/cuda_stream.cpp b/modules/core/src/cuda_stream.cpp
index 1ea8df37b9..24ecbd4ab0 100644
--- a/modules/core/src/cuda_stream.cpp
+++ b/modules/core/src/cuda_stream.cpp
@@ -668,20 +668,33 @@ void cv::cuda::setBufferPoolConfig(int deviceId, size_t stackSize, int stackCoun
 #endif
 }
 
-#ifdef HAVE_CUDA
-
-cv::cuda::BufferPool::BufferPool(Stream& stream) : allocator_(stream.impl_->stackAllocator.get())
+#ifndef HAVE_CUDA
+cv::cuda::BufferPool::BufferPool(Stream& stream)
+{
+    (void) stream;
+    throw_no_cuda();
+}
+#else
+cv::cuda::BufferPool::BufferPool(Stream& stream) : allocator_(stream.impl_->stackAllocator)
 {
 }
+#endif
 
 GpuMat cv::cuda::BufferPool::getBuffer(int rows, int cols, int type)
 {
+#ifndef HAVE_CUDA
+    (void) rows;
+    (void) cols;
+    (void) type;
+    throw_no_cuda();
+    return GpuMat();
+#else
     GpuMat buf(allocator_);
     buf.create(rows, cols, type);
     return buf;
+#endif
 }
 
-#endif
 
 ////////////////////////////////////////////////////////////////
 // Event