diff --git a/CMakeLists.txt b/CMakeLists.txt
index 7831430e69..6621a2f99f 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -110,14 +110,15 @@ endif()
 
 # Optional 3rd party components
 # ===================================================
-OCV_OPTION(WITH_1394           "Include IEEE1394 support"                    ON   IF (UNIX AND NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_1394           "Include IEEE1394 support"                    ON   IF (UNIX AND NOT ANDROID AND NOT IOS AND NOT CARMA) )
 OCV_OPTION(WITH_AVFOUNDATION   "Use AVFoundation for Video I/O"              ON   IF IOS)
 OCV_OPTION(WITH_CARBON         "Use Carbon for UI instead of Cocoa"          OFF  IF APPLE )
-OCV_OPTION(WITH_CUBLAS         "Include NVidia Cuda Basic Linear Algebra Subprograms (BLAS) library support" OFF IF (CMAKE_VERSION VERSION_GREATER "2.8" AND NOT ANDROID AND NOT IOS) )
 OCV_OPTION(WITH_CUDA           "Include NVidia Cuda Runtime support"         ON   IF (CMAKE_VERSION VERSION_GREATER "2.8" AND NOT ANDROID AND NOT IOS) )
 OCV_OPTION(WITH_CUFFT          "Include NVidia Cuda Fast Fourier Transform (FFT) library support"            ON  IF (CMAKE_VERSION VERSION_GREATER "2.8" AND NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_CUBLAS         "Include NVidia Cuda Basic Linear Algebra Subprograms (BLAS) library support" OFF IF (CMAKE_VERSION VERSION_GREATER "2.8" AND NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_NVCUVID        "Include NVidia Video Decoding library support"                               OFF IF (CMAKE_VERSION VERSION_GREATER "2.8" AND NOT ANDROID AND NOT IOS AND NOT APPLE) )
 OCV_OPTION(WITH_EIGEN          "Include Eigen2/Eigen3 support"               ON)
-OCV_OPTION(WITH_FFMPEG         "Include FFMPEG support"                      ON   IF (NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_FFMPEG         "Include FFMPEG support"                      ON   IF (NOT ANDROID AND NOT IOS))
 OCV_OPTION(WITH_GSTREAMER      "Include Gstreamer support"                   ON   IF (UNIX AND NOT APPLE AND NOT ANDROID) )
 OCV_OPTION(WITH_GTK            "Include GTK support"                         ON   IF (UNIX AND NOT APPLE AND NOT ANDROID) )
 OCV_OPTION(WITH_IMAGEIO        "ImageIO support for OS X"                    OFF  IF APPLE)
@@ -140,9 +141,9 @@ OCV_OPTION(WITH_V4L            "Include Video 4 Linux support"               ON
 OCV_OPTION(WITH_VIDEOINPUT     "Build HighGUI with DirectShow support"       ON   IF WIN32 )
 OCV_OPTION(WITH_XIMEA          "Include XIMEA cameras support"               OFF  IF (NOT ANDROID AND NOT APPLE) )
 OCV_OPTION(WITH_XINE           "Include Xine support (GPL)"                  OFF  IF (UNIX AND NOT APPLE AND NOT ANDROID) )
-OCV_OPTION(WITH_OPENCL         "Include OpenCL Runtime support"              OFF  IF (NOT ANDROID AND NOT IOS) )
-OCV_OPTION(WITH_OPENCLAMDFFT   "Include AMD OpenCL FFT library support"      OFF  IF (NOT ANDROID AND NOT IOS) )
-OCV_OPTION(WITH_OPENCLAMDBLAS  "Include AMD OpenCL BLAS library support"     OFF  IF (NOT ANDROID AND NOT IOS) )
+OCV_OPTION(WITH_OPENCL         "Include OpenCL Runtime support"              OFF  IF (NOT ANDROID AND NOT IOS AND NOT CARMA) )
+OCV_OPTION(WITH_OPENCLAMDFFT   "Include AMD OpenCL FFT library support"      OFF  IF (NOT ANDROID AND NOT IOS AND NOT CARMA) )
+OCV_OPTION(WITH_OPENCLAMDBLAS  "Include AMD OpenCL BLAS library support"     OFF  IF (NOT ANDROID AND NOT IOS AND NOT CARMA) )
 
 
 # OpenCV build components
@@ -161,12 +162,12 @@ OCV_OPTION(BUILD_ANDROID_SERVICE    "Build OpenCV Manager for Google Play" OFF I
 OCV_OPTION(BUILD_ANDROID_PACKAGE    "Build platform-specific package for Google Play" OFF IF ANDROID )
 
 # 3rd party libs
-OCV_OPTION(BUILD_ZLIB               "Build zlib from source"         WIN32 OR APPLE )
-OCV_OPTION(BUILD_TIFF               "Build libtiff from source"      WIN32 OR ANDROID OR APPLE )
-OCV_OPTION(BUILD_JASPER             "Build libjasper from source"    WIN32 OR ANDROID OR APPLE )
-OCV_OPTION(BUILD_JPEG               "Build libjpeg from source"      WIN32 OR ANDROID OR APPLE )
-OCV_OPTION(BUILD_PNG                "Build libpng from source"       WIN32 OR ANDROID OR APPLE )
-OCV_OPTION(BUILD_OPENEXR            "Build openexr from source"      WIN32 OR ANDROID OR APPLE )
+OCV_OPTION(BUILD_ZLIB               "Build zlib from source"         WIN32 OR APPLE OR CARMA )
+OCV_OPTION(BUILD_TIFF               "Build libtiff from source"      WIN32 OR ANDROID OR APPLE OR CARMA )
+OCV_OPTION(BUILD_JASPER             "Build libjasper from source"    WIN32 OR ANDROID OR APPLE OR CARMA )
+OCV_OPTION(BUILD_JPEG               "Build libjpeg from source"      WIN32 OR ANDROID OR APPLE OR CARMA )
+OCV_OPTION(BUILD_PNG                "Build libpng from source"       WIN32 OR ANDROID OR APPLE OR CARMA )
+OCV_OPTION(BUILD_OPENEXR            "Build openexr from source"      WIN32 OR ANDROID OR APPLE OR CARMA )
 
 
 # OpenCV installation options
@@ -776,8 +777,9 @@ if(HAVE_CUDA)
   status("")
   status("  NVIDIA CUDA")
 
-  status("    Use CUFFT:"            HAVE_CUFFT  THEN YES ELSE NO)
-  status("    Use CUBLAS:"           HAVE_CUBLAS THEN YES ELSE NO)
+  status("    Use CUFFT:"            HAVE_CUFFT   THEN YES ELSE NO)
+  status("    Use CUBLAS:"           HAVE_CUBLAS  THEN YES ELSE NO)
+  status("    USE NVCUVID:"          HAVE_NVCUVID THEN YES ELSE NO)
   status("    NVIDIA GPU arch:"      ${OPENCV_CUDA_ARCH_BIN})
   status("    NVIDIA PTX archs:"     ${OPENCV_CUDA_ARCH_PTX})
   status("    Use fast math:"        CUDA_FAST_MATH THEN YES ELSE NO)
diff --git a/cmake/OpenCVDetectCUDA.cmake b/cmake/OpenCVDetectCUDA.cmake
index c1cd83866b..25cfdd225c 100644
--- a/cmake/OpenCVDetectCUDA.cmake
+++ b/cmake/OpenCVDetectCUDA.cmake
@@ -3,17 +3,17 @@ if(${CMAKE_VERSION} VERSION_LESS "2.8.3")
   return()
 endif()
 
-if (WIN32 AND NOT MSVC)
-  message(STATUS "CUDA compilation is disabled (due to only Visual Studio compiler suppoted on your platform).")
+if(WIN32 AND NOT MSVC)
+  message(STATUS "CUDA compilation is disabled (due to only Visual Studio compiler supported on your platform).")
   return()
 endif()
 
-if (CMAKE_COMPILER_IS_GNUCXX AND NOT APPLE AND CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
-  message(STATUS "CUDA compilation is disabled (due to Clang unsuppoted on your platform).")
+if(CMAKE_COMPILER_IS_GNUCXX AND NOT APPLE AND CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
+  message(STATUS "CUDA compilation is disabled (due to Clang unsupported on your platform).")
   return()
 endif()
 
-find_package(CUDA 4.1)
+find_package(CUDA 4.2 QUIET)
 
 if(CUDA_FOUND)
   set(HAVE_CUDA 1)
@@ -26,16 +26,21 @@ if(CUDA_FOUND)
     set(HAVE_CUBLAS 1)
   endif()
 
+  if(WITH_NVCUVID)
+    find_cuda_helper_libs(nvcuvid)
+    set(HAVE_NVCUVID 1)
+  endif()
+
   message(STATUS "CUDA detected: " ${CUDA_VERSION})
 
-  if(${CUDA_VERSION_STRING} VERSION_GREATER "4.1")
-    set(CUDA_ARCH_BIN "1.1 1.2 1.3 2.0 2.1(2.0) 3.0" CACHE STRING "Specify 'real' GPU architectures to build binaries for, BIN(PTX) format is supported")
+  if (CARMA)
+    set(CUDA_ARCH_BIN "2.1(2.0) 3.0" CACHE STRING "Specify 'real' GPU architectures to build binaries for, BIN(PTX) format is supported")
+    set(CUDA_ARCH_PTX "3.0" CACHE STRING "Specify 'virtual' PTX architectures to build PTX intermediate code for")
   else()
-    set(CUDA_ARCH_BIN "1.1 1.2 1.3 2.0 2.1(2.0)" CACHE STRING "Specify 'real' GPU architectures to build binaries for, BIN(PTX) format is supported")
+    set(CUDA_ARCH_BIN "1.1 1.2 1.3 2.0 2.1(2.0) 3.0" CACHE STRING "Specify 'real' GPU architectures to build binaries for, BIN(PTX) format is supported")
+    set(CUDA_ARCH_PTX "2.0 3.0" CACHE STRING "Specify 'virtual' PTX architectures to build PTX intermediate code for")
   endif()
 
-  set(CUDA_ARCH_PTX "2.0" CACHE STRING "Specify 'virtual' PTX architectures to build PTX intermediate code for")
-
   string(REGEX REPLACE "\\." "" ARCH_BIN_NO_POINTS "${CUDA_ARCH_BIN}")
   string(REGEX REPLACE "\\." "" ARCH_PTX_NO_POINTS "${CUDA_ARCH_PTX}")
 
@@ -72,11 +77,20 @@ if(CUDA_FOUND)
 
   # Tell NVCC to add PTX intermediate code for the specified architectures
   string(REGEX MATCHALL "[0-9]+" ARCH_LIST "${ARCH_PTX_NO_POINTS}")
-    foreach(ARCH IN LISTS ARCH_LIST)
-      set(NVCC_FLAGS_EXTRA ${NVCC_FLAGS_EXTRA} -gencode arch=compute_${ARCH},code=compute_${ARCH})
-      set(OPENCV_CUDA_ARCH_PTX "${OPENCV_CUDA_ARCH_PTX} ${ARCH}")
-      set(OPENCV_CUDA_ARCH_FEATURES "${OPENCV_CUDA_ARCH_FEATURES} ${ARCH}")
-    endforeach()
+  foreach(ARCH IN LISTS ARCH_LIST)
+    set(NVCC_FLAGS_EXTRA ${NVCC_FLAGS_EXTRA} -gencode arch=compute_${ARCH},code=compute_${ARCH})
+    set(OPENCV_CUDA_ARCH_PTX "${OPENCV_CUDA_ARCH_PTX} ${ARCH}")
+    set(OPENCV_CUDA_ARCH_FEATURES "${OPENCV_CUDA_ARCH_FEATURES} ${ARCH}")
+  endforeach()
+
+  if(CARMA)
+    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} --target-cpu-architecture=ARM" )
+
+    if (CMAKE_VERSION VERSION_LESS 2.8.10)
+      set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -ccbin=${CMAKE_CXX_COMPILER}" )
+    endif()
+
+  endif()
 
   # These vars will be processed in other scripts
   set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} ${NVCC_FLAGS_EXTRA})
@@ -84,7 +98,7 @@ if(CUDA_FOUND)
 
   message(STATUS "CUDA NVCC target flags: ${CUDA_NVCC_FLAGS}")
 
-  OCV_OPTION(CUDA_FAST_MATH  "Enable --use_fast_math for CUDA compiler " OFF)
+  OCV_OPTION(CUDA_FAST_MATH "Enable --use_fast_math for CUDA compiler " OFF)
 
   if(CUDA_FAST_MATH)
     set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} --use_fast_math)
@@ -92,7 +106,6 @@ if(CUDA_FOUND)
 
   mark_as_advanced(CUDA_BUILD_CUBIN CUDA_BUILD_EMULATION CUDA_VERBOSE_BUILD CUDA_SDK_ROOT_DIR)
 
-  unset(CUDA_npp_LIBRARY CACHE)
   find_cuda_helper_libs(npp)
 
   macro(ocv_cuda_compile VAR)
@@ -106,15 +119,15 @@ if(CUDA_FOUND)
       string(REPLACE "-ggdb3" "" ${var} "${${var}}")
     endforeach()
 
-    if (BUILD_SHARED_LIBS)
+    if(BUILD_SHARED_LIBS)
       set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -Xcompiler -DCVAPI_EXPORTS)
     endif()
 
     if(UNIX OR APPLE)
-      set (CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -Xcompiler -fPIC)
+      set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -Xcompiler -fPIC)
     endif()
     if(APPLE)
-      set (CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -Xcompiler -fno-finite-math-only)
+      set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} -Xcompiler -fno-finite-math-only)
     endif()
 
     # disabled because of multiple warnings during building nvcc auto generated files
diff --git a/cmake/templates/OpenCVConfig.cmake.in b/cmake/templates/OpenCVConfig.cmake.in
index 4cf87536d8..de0a377a7c 100644
--- a/cmake/templates/OpenCVConfig.cmake.in
+++ b/cmake/templates/OpenCVConfig.cmake.in
@@ -42,8 +42,9 @@
 set(OpenCV_COMPUTE_CAPABILITIES @OpenCV_CUDA_CC_CONFIGCMAKE@)
 
 set(OpenCV_CUDA_VERSION @OpenCV_CUDA_VERSION@)
-set(OpenCV_USE_CUBLAS @HAVE_CUBLAS@)
-set(OpenCV_USE_CUFFT  @HAVE_CUFFT@)
+set(OpenCV_USE_CUBLAS   @HAVE_CUBLAS@)
+set(OpenCV_USE_CUFFT    @HAVE_CUFFT@)
+set(OpenCV_USE_NVCUVID  @HAVE_NVCUVID@)
 
 # Android API level from which OpenCV has been compiled is remembered
 set(OpenCV_ANDROID_NATIVE_API_LEVEL @OpenCV_ANDROID_NATIVE_API_LEVEL_CONFIGCMAKE@)
@@ -218,17 +219,22 @@ foreach(__opttype OPT DBG)
   else()
     #TODO: duplicates are annoying but they should not be the problem
   endif()
-  # fix hard coded paths for CUDA libraries under Windows
-  if(WIN32 AND OpenCV_CUDA_VERSION AND NOT OpenCV_SHARED)
+  
+  # CUDA
+  if(OpenCV_CUDA_VERSION AND (CARMA OR (WIN32 AND NOT OpenCV_SHARED)))
     if(NOT CUDA_FOUND)
       find_package(CUDA ${OpenCV_CUDA_VERSION} EXACT REQUIRED)
     else()
       if(NOT CUDA_VERSION_STRING VERSION_EQUAL OpenCV_CUDA_VERSION)
-        message(FATAL_ERROR "OpenCV static library compiled with CUDA ${OpenCV_CUDA_VERSION} support. Please, use the same version or rebuild OpenCV with CUDA ${CUDA_VERSION_STRING}")
+        if(WIN32)
+          message(FATAL_ERROR "OpenCV static library was compiled with CUDA ${OpenCV_CUDA_VERSION} support. Please, use the same version or rebuild OpenCV with CUDA ${CUDA_VERSION_STRING}")
+        else()
+          message(FATAL_ERROR "OpenCV library for CARMA was compiled with CUDA ${OpenCV_CUDA_VERSION} support. Please, use the same version or rebuild OpenCV with CUDA ${CUDA_VERSION_STRING}")
+        endif()
       endif()
     endif()
 
-    list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY} ${CUDA_nvcuvid_LIBRARY} ${CUDA_nvcuvenc_LIBRARY})
+    list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY})
 
     if(OpenCV_USE_CUBLAS)
       list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_CUBLAS_LIBRARIES})
@@ -238,6 +244,13 @@ foreach(__opttype OPT DBG)
       list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_CUFFT_LIBRARIES})
     endif()
 
+    if(OpenCV_USE_NVCUVID)
+      list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_nvcuvid_LIBRARIES})
+    endif()
+
+    if(WIN32)
+      list(APPEND OpenCV_EXTRA_LIBS_${__opttype} ${CUDA_nvcuvenc_LIBRARIES})
+    endif()
   endif()
 endforeach()
 
diff --git a/cmake/templates/cvconfig.h.cmake b/cmake/templates/cvconfig.h.cmake
index 368905fd9c..aa372260e4 100644
--- a/cmake/templates/cvconfig.h.cmake
+++ b/cmake/templates/cvconfig.h.cmake
@@ -175,21 +175,15 @@
 /* NVidia Cuda Runtime API*/
 #cmakedefine HAVE_CUDA
 
-/* OpenCL Support */
-#cmakedefine HAVE_OPENCL
-
-/* AMD's OpenCL Fast Fourier Transform Library*/
-#cmakedefine HAVE_CLAMDFFT
-
-/* AMD's Basic Linear Algebra Subprograms Library*/
-#cmakedefine HAVE_CLAMDBLAS
-
 /* NVidia Cuda Fast Fourier Transform (FFT) API*/
 #cmakedefine HAVE_CUFFT
 
 /* NVidia Cuda Basic Linear Algebra Subprograms (BLAS) API*/
 #cmakedefine HAVE_CUBLAS
 
+/* NVidia Video Decoding API*/
+#cmakedefine HAVE_NVCUVID
+
 /* Compile for 'real' NVIDIA GPU architectures */
 #define CUDA_ARCH_BIN "${OPENCV_CUDA_ARCH_BIN}"
 
@@ -202,6 +196,15 @@
 /* Create PTX or BIN for 1.0 compute capability */
 #cmakedefine CUDA_ARCH_BIN_OR_PTX_10
 
+/* OpenCL Support */
+#cmakedefine HAVE_OPENCL
+
+/* AMD's OpenCL Fast Fourier Transform Library*/
+#cmakedefine HAVE_CLAMDFFT
+
+/* AMD's Basic Linear Algebra Subprograms Library*/
+#cmakedefine HAVE_CLAMDBLAS
+
 /* VideoInput library */
 #cmakedefine HAVE_VIDEOINPUT
 
diff --git a/modules/core/CMakeLists.txt b/modules/core/CMakeLists.txt
index cfa14cdcdb..4c5112e3f9 100644
--- a/modules/core/CMakeLists.txt
+++ b/modules/core/CMakeLists.txt
@@ -10,7 +10,6 @@ if(HAVE_CUDA)
   file(GLOB lib_cuda "src/cuda/*.cu")
   ocv_cuda_compile(cuda_objs ${lib_cuda})
 
-
   set(cuda_link_libs ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY})
 else()
   set(lib_cuda "")
diff --git a/modules/gpu/CMakeLists.txt b/modules/gpu/CMakeLists.txt
index cf69ba1259..0c9f709b15 100644
--- a/modules/gpu/CMakeLists.txt
+++ b/modules/gpu/CMakeLists.txt
@@ -5,7 +5,7 @@ endif()
 set(the_description "GPU-accelerated Computer Vision")
 ocv_add_module(gpu opencv_imgproc opencv_calib3d opencv_objdetect opencv_video opencv_nonfree opencv_photo opencv_legacy)
 
-ocv_module_include_directories("${CMAKE_CURRENT_SOURCE_DIR}/src/cuda" "${CMAKE_CURRENT_SOURCE_DIR}/../highgui/src")
+ocv_module_include_directories("${CMAKE_CURRENT_SOURCE_DIR}/src/cuda")
 
 file(GLOB lib_hdrs               "include/opencv2/${name}/*.hpp"               "include/opencv2/${name}/*.h")
 file(GLOB lib_device_hdrs        "include/opencv2/${name}/device/*.hpp"        "include/opencv2/${name}/device/*.h")
@@ -15,24 +15,21 @@ file(GLOB lib_cuda_hdrs          "src/cuda/*.hpp" "src/cuda/*.h")
 file(GLOB lib_srcs               "src/*.cpp")
 file(GLOB lib_cuda               "src/cuda/*.cu*")
 
-source_group("Include" FILES ${lib_hdrs})
-source_group("Src\\Host" FILES ${lib_srcs} ${lib_int_hdrs})
-source_group("Src\\Cuda" FILES ${lib_cuda} ${lib_cuda_hdrs})
-source_group("Device" FILES ${lib_device_hdrs})
+source_group("Include"        FILES ${lib_hdrs})
+source_group("Src\\Host"      FILES ${lib_srcs} ${lib_int_hdrs})
+source_group("Src\\Cuda"      FILES ${lib_cuda} ${lib_cuda_hdrs})
+source_group("Device"         FILES ${lib_device_hdrs})
 source_group("Device\\Detail" FILES ${lib_device_hdrs_detail})
 
 if (HAVE_CUDA)
-  file(GLOB_RECURSE ncv_srcs "src/nvidia/*.cpp")
+  file(GLOB_RECURSE ncv_srcs "src/nvidia/*.cpp" "src/nvidia/*.h*")
   file(GLOB_RECURSE ncv_cuda "src/nvidia/*.cu")
-  file(GLOB_RECURSE ncv_hdrs "src/nvidia/*.hpp" "src/nvidia/*.h")
-  set(ncv_files ${ncv_srcs} ${ncv_hdrs} ${ncv_cuda})
+  set(ncv_files ${ncv_srcs} ${ncv_cuda})
 
   source_group("Src\\NVidia" FILES ${ncv_files})
   ocv_include_directories("src/nvidia" "src/nvidia/core" "src/nvidia/NPP_staging" ${CUDA_INCLUDE_DIRS})
   ocv_warnings_disable(CMAKE_CXX_FLAGS -Wundef -Wmissing-declarations /wd4211 /wd4201 /wd4100 /wd4505 /wd4408)
   string(REPLACE "-Wsign-promo" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
-
-  #set (CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-keep")
   #set (CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xcompiler;/EHsc-;")
 
   if(MSVC)
@@ -47,12 +44,9 @@ if (HAVE_CUDA)
 
   ocv_cuda_compile(cuda_objs ${lib_cuda} ${ncv_cuda})
 
-  #CUDA_BUILD_CLEAN_TARGET()
-
   set(cuda_link_libs ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY})
 
-  if(NOT APPLE)
-    find_cuda_helper_libs(nvcuvid)
+  if(WITH_NVCUVID)
     set(cuda_link_libs ${cuda_link_libs} ${CUDA_nvcuvid_LIBRARY})
   endif()
 
@@ -61,7 +55,7 @@ if (HAVE_CUDA)
     set(cuda_link_libs ${cuda_link_libs} ${CUDA_nvcuvenc_LIBRARY})
   endif()
 
-  if(NOT APPLE AND WITH_FFMPEG)
+  if(WITH_FFMPEG)
     set(cuda_link_libs ${cuda_link_libs} ${HIGHGUI_LIBRARIES})
   endif()
 else()
diff --git a/modules/gpu/misc/carma.toolchain.cmake b/modules/gpu/misc/carma.toolchain.cmake
new file mode 100644
index 0000000000..18f0e0f934
--- /dev/null
+++ b/modules/gpu/misc/carma.toolchain.cmake
@@ -0,0 +1,26 @@
+set(CMAKE_SYSTEM_NAME Linux)
+set(CMAKE_SYSTEM_VERSION 1)
+set(CMAKE_SYSTEM_PROCESSOR arm)
+
+set(CMAKE_C_COMPILER    arm-linux-gnueabi-gcc-4.5)
+set(CMAKE_CXX_COMPILER  arm-linux-gnueabi-g++-4.5)
+
+#suppress compiller varning
+set( CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-psabi" )
+set( CMAKE_C_FLAGS   "${CMAKE_C_FLAGS}   -Wno-psabi" )
+
+# can be any other plases
+set(__arm_linux_eabi_root /usr/arm-linux-gnueabi)
+
+set(CMAKE_FIND_ROOT_PATH ${CMAKE_FIND_ROOT_PATH} ${__arm_linux_eabi_root})
+
+if(EXISTS ${CUDA_TOOLKIT_ROOT_DIR})
+    set(CMAKE_FIND_ROOT_PATH ${CMAKE_FIND_ROOT_PATH} ${CUDA_TOOLKIT_ROOT_DIR})
+endif()
+
+set(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)
+set(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)
+set(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM ONLY)
+
+set(CARMA 1)
+add_definitions(-DCARMA)
diff --git a/modules/gpu/src/cuvid_video_source.cpp b/modules/gpu/src/cuvid_video_source.cpp
index 5c46913539..7d45b8fee9 100644
--- a/modules/gpu/src/cuvid_video_source.cpp
+++ b/modules/gpu/src/cuvid_video_source.cpp
@@ -1,7 +1,7 @@
 #include "cuvid_video_source.h"
 #include "cu_safe_call.h"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
 cv::gpu::detail::CuvidVideoSource::CuvidVideoSource(const std::string& fname)
 {
diff --git a/modules/gpu/src/cuvid_video_source.h b/modules/gpu/src/cuvid_video_source.h
index a06353fedd..1c4c0e5e00 100644
--- a/modules/gpu/src/cuvid_video_source.h
+++ b/modules/gpu/src/cuvid_video_source.h
@@ -45,7 +45,7 @@
 
 #include "precomp.hpp"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
 namespace cv { namespace gpu
 {
diff --git a/modules/gpu/src/ffmpeg_video_source.cpp b/modules/gpu/src/ffmpeg_video_source.cpp
index 71553d9915..bd3d70058a 100644
--- a/modules/gpu/src/ffmpeg_video_source.cpp
+++ b/modules/gpu/src/ffmpeg_video_source.cpp
@@ -42,12 +42,12 @@
 
 #include "ffmpeg_video_source.h"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
-#ifdef HAVE_FFMPEG
-    #include "cap_ffmpeg_impl.hpp"
+#if defined(HAVE_FFMPEG) && defined(BUILD_SHARED_LIBS)
+    #include "../src/cap_ffmpeg_impl.hpp"
 #else
-    #include "cap_ffmpeg_api.hpp"
+    #include "../src/cap_ffmpeg_api.hpp"
 #endif
 
 namespace
diff --git a/modules/gpu/src/ffmpeg_video_source.h b/modules/gpu/src/ffmpeg_video_source.h
index 66bb7c71ca..41bf0cfd0e 100644
--- a/modules/gpu/src/ffmpeg_video_source.h
+++ b/modules/gpu/src/ffmpeg_video_source.h
@@ -46,7 +46,7 @@
 #include "precomp.hpp"
 #include "thread_wrappers.h"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
 struct InputMediaStream_FFMPEG;
 
diff --git a/modules/gpu/src/frame_queue.cpp b/modules/gpu/src/frame_queue.cpp
index 4ed6d0f147..0d93918732 100644
--- a/modules/gpu/src/frame_queue.cpp
+++ b/modules/gpu/src/frame_queue.cpp
@@ -42,7 +42,7 @@
 
 #include "frame_queue.h"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
 cv::gpu::detail::FrameQueue::FrameQueue() :
     endOfDecode_(0),
diff --git a/modules/gpu/src/frame_queue.h b/modules/gpu/src/frame_queue.h
index 5c049a28e5..00c44a4015 100644
--- a/modules/gpu/src/frame_queue.h
+++ b/modules/gpu/src/frame_queue.h
@@ -46,7 +46,7 @@
 #include "precomp.hpp"
 #include "thread_wrappers.h"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
 namespace cv { namespace gpu
 {
diff --git a/modules/gpu/src/precomp.hpp b/modules/gpu/src/precomp.hpp
index c37332ae45..7c53185107 100644
--- a/modules/gpu/src/precomp.hpp
+++ b/modules/gpu/src/precomp.hpp
@@ -97,12 +97,12 @@
         #include <cublas.h>
     #endif
 
-    #ifndef __APPLE__
+    #ifdef HAVE_NVCUVID
         #include <nvcuvid.h>
-    #endif
 
-    #ifdef WIN32
-        #include <NVEncoderAPI.h>
+        #ifdef WIN32
+            #include <NVEncoderAPI.h>
+        #endif
     #endif
 
     #include "internal_shared.hpp"
diff --git a/modules/gpu/src/thread_wrappers.cpp b/modules/gpu/src/thread_wrappers.cpp
index fcfa59827e..f8099aca91 100644
--- a/modules/gpu/src/thread_wrappers.cpp
+++ b/modules/gpu/src/thread_wrappers.cpp
@@ -42,7 +42,7 @@
 
 #include "thread_wrappers.h"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
 #ifdef WIN32
     #define NOMINMAX
diff --git a/modules/gpu/src/thread_wrappers.h b/modules/gpu/src/thread_wrappers.h
index 9198d3e906..78e675216c 100644
--- a/modules/gpu/src/thread_wrappers.h
+++ b/modules/gpu/src/thread_wrappers.h
@@ -45,7 +45,7 @@
 
 #include "precomp.hpp"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
 namespace cv { namespace gpu
 {
diff --git a/modules/gpu/src/video_decoder.cpp b/modules/gpu/src/video_decoder.cpp
index 41e1d26078..c1e8f86a4c 100644
--- a/modules/gpu/src/video_decoder.cpp
+++ b/modules/gpu/src/video_decoder.cpp
@@ -43,7 +43,7 @@
 #include "video_decoder.h"
 #include "frame_queue.h"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
 void cv::gpu::detail::VideoDecoder::create(const VideoReader_GPU::FormatInfo& videoFormat)
 {
diff --git a/modules/gpu/src/video_decoder.h b/modules/gpu/src/video_decoder.h
index 9c4dfa9294..634643b889 100644
--- a/modules/gpu/src/video_decoder.h
+++ b/modules/gpu/src/video_decoder.h
@@ -46,7 +46,7 @@
 #include "precomp.hpp"
 #include "cu_safe_call.h"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
 namespace cv { namespace gpu
 {
diff --git a/modules/gpu/src/video_parser.cpp b/modules/gpu/src/video_parser.cpp
index 2ff14bd75a..70187fcb2f 100644
--- a/modules/gpu/src/video_parser.cpp
+++ b/modules/gpu/src/video_parser.cpp
@@ -43,7 +43,7 @@
 #include "video_parser.h"
 #include "cu_safe_call.h"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
 cv::gpu::detail::VideoParser::VideoParser(VideoDecoder* videoDecoder, FrameQueue* frameQueue) :
     videoDecoder_(videoDecoder), frameQueue_(frameQueue), unparsedPackets_(0), hasError_(false)
diff --git a/modules/gpu/src/video_parser.h b/modules/gpu/src/video_parser.h
index 3935ffcefe..44a53710bf 100644
--- a/modules/gpu/src/video_parser.h
+++ b/modules/gpu/src/video_parser.h
@@ -48,7 +48,7 @@
 #include "frame_queue.h"
 #include "video_decoder.h"
 
-#if defined(HAVE_CUDA) && !defined(__APPLE__)
+#if defined(HAVE_CUDA) && defined(HAVE_NVCUVID)
 
 namespace cv { namespace gpu
 {
diff --git a/modules/gpu/src/video_reader.cpp b/modules/gpu/src/video_reader.cpp
index df3c2ff2f0..3224902c6d 100644
--- a/modules/gpu/src/video_reader.cpp
+++ b/modules/gpu/src/video_reader.cpp
@@ -42,7 +42,7 @@
 
 #include "precomp.hpp"
 
-#if !defined HAVE_CUDA || defined(CUDA_DISABLER) || defined(__APPLE__)
+#if !defined(HAVE_CUDA) || defined(CUDA_DISABLER) || !defined(HAVE_NVCUVID)
 
 class cv::gpu::VideoReader_GPU::Impl
 {
diff --git a/modules/gpu/src/video_writer.cpp b/modules/gpu/src/video_writer.cpp
index 036512d9af..fe44a16f7c 100644
--- a/modules/gpu/src/video_writer.cpp
+++ b/modules/gpu/src/video_writer.cpp
@@ -42,7 +42,7 @@
 
 #include "precomp.hpp"
 
-#if !defined HAVE_CUDA || defined(CUDA_DISABLER) || !defined(WIN32)
+#if !defined(HAVE_CUDA) || defined(CUDA_DISABLER) || !defined(HAVE_NVCUVID) || !defined(WIN32)
 
 class cv::gpu::VideoWriter_GPU::Impl
 {
@@ -71,9 +71,9 @@ void cv::gpu::VideoWriter_GPU::EncoderParams::save(const std::string&) const { t
 #else // !defined HAVE_CUDA || !defined WIN32
 
 #ifdef HAVE_FFMPEG
-    #include "cap_ffmpeg_impl.hpp"
+    #include "../src/cap_ffmpeg_impl.hpp"
 #else
-    #include "cap_ffmpeg_api.hpp"
+    #include "../src/cap_ffmpeg_api.hpp"
 #endif
 
 
@@ -1001,4 +1001,4 @@ void cv::gpu::VideoWriter_GPU::EncoderParams::save(const std::string& configFile
     cv::write(fs, "DisableSPSPPS"  , DisableSPSPPS);
 }
 
-#endif // !defined HAVE_CUDA || !defined WIN32
\ No newline at end of file
+#endif // !defined HAVE_CUDA || !defined WIN32