/// <summary>
 /// Image threshold.<para/>
 /// If for a comparison operations sourcePixel is less than nThresholdLT is true, the pixel is set
 /// to nValueLT, else if sourcePixel is greater than nThresholdGT the pixel is set to nValueGT, otherwise it is set to sourcePixel.
 /// </summary>
 /// <param name="dest">Destination image</param>
 /// <param name="nThresholdLT">The thresholdLT value.</param>
 /// <param name="nValueLT">The thresholdLT replacement value.</param>
 /// <param name="nThresholdGT">The thresholdGT value.</param>
 /// <param name="nValueGT">The thresholdGT replacement value.</param>
 public void ThresholdLTGT(NPPImage_16uC1 dest, ushort nThresholdLT, ushort nValueLT, ushort nThresholdGT, ushort nValueGT)
 {
     status = NPPNativeMethods.NPPi.Threshold.nppiThreshold_LTValGTVal_16u_C1R(_devPtrRoi, _pitch, dest.DevicePointerRoi, dest.Pitch, _sizeRoi, nThresholdLT, nValueLT, nThresholdGT, nValueGT);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiThreshold_LTValGTVal_16u_C1R", status));
     NPPException.CheckNppStatus(status, this);
 }

		/// <summary>
		/// resizes planar images.
		/// </summary>
		/// <param name="src0">Source image (Channel 0)</param>
		/// <param name="src1">Source image (Channel 1)</param>
		/// <param name="src2">Source image (Channel 2)</param>
		/// <param name="dest0">Destination image (Channel 0)</param>
		/// <param name="dest1">Destination image (Channel 1)</param>
		/// <param name="dest2">Destination image (Channel 2)</param>
		/// <param name="xFactor">X scaling factor</param>
		/// <param name="yFactor">Y scaling factor</param>
		/// <param name="eInterpolation">Interpolation mode</param>
		public static void Resize(NPPImage_16uC1 src0, NPPImage_16uC1 src1, NPPImage_16uC1 src2, NPPImage_16uC1 dest0, NPPImage_16uC1 dest1, NPPImage_16uC1 dest2, double xFactor, double yFactor, InterpolationMode eInterpolation)
		{
			CUdeviceptr[] src = new CUdeviceptr[] { src0.DevicePointer, src1.DevicePointer, src2.DevicePointer };
			CUdeviceptr[] dst = new CUdeviceptr[] { dest0.DevicePointerRoi, dest1.DevicePointerRoi, dest2.DevicePointerRoi };
			NppStatus status = NPPNativeMethods.NPPi.GeometricTransforms.nppiResize_16u_P3R(src, src0.Size, src0.Pitch, new NppiRect(src0.PointRoi, src0.SizeRoi), dst, dest0.Pitch, dest0.SizeRoi, xFactor, yFactor, eInterpolation);
			Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiResize_16u_P3R", status));
			NPPException.CheckNppStatus(status, null);
		}

		/// <summary>
		/// planar image remap.
		/// </summary>
		/// <param name="src0">Source image (Channel 0)</param>
		/// <param name="src1">Source image (Channel 1)</param>
		/// <param name="src2">Source image (Channel 2)</param>
		/// <param name="dest0">Destination image (Channel 0)</param>
		/// <param name="dest1">Destination image (Channel 1)</param>
		/// <param name="dest2">Destination image (Channel 2)</param>
		/// <param name="pXMap">Device memory pointer to 2D image array of X coordinate values to be used when sampling source image. </param>
		/// <param name="pYMap">Device memory pointer to 2D image array of Y coordinate values to be used when sampling source image. </param>
		/// <param name="eInterpolation">The type of eInterpolation to perform resampling.</param>
		public static void Remap(NPPImage_16uC1 src0, NPPImage_16uC1 src1, NPPImage_16uC1 src2, NPPImage_16uC1 dest0, NPPImage_16uC1 dest1, NPPImage_16uC1 dest2, NPPImage_32fC1 pXMap, NPPImage_32fC1 pYMap, InterpolationMode eInterpolation)
		{
			CUdeviceptr[] src = new CUdeviceptr[] { src0.DevicePointer, src1.DevicePointer, src2.DevicePointer };
			CUdeviceptr[] dst = new CUdeviceptr[] { dest0.DevicePointerRoi, dest1.DevicePointerRoi, dest2.DevicePointerRoi };
			NppiRect srcRect = new NppiRect(src0.PointRoi, src0.SizeRoi);
			NppStatus status = NPPNativeMethods.NPPi.Remap.nppiRemap_16u_P3R(src, src0.SizeRoi, src0.Pitch, srcRect, pXMap.DevicePointerRoi, pXMap.Pitch, pYMap.DevicePointerRoi, pYMap.Pitch, dst, dest0.Pitch, dest0.SizeRoi, eInterpolation);
			Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiRemap_16u_P3R", status));
			NPPException.CheckNppStatus(status, null);
		}

		/// <summary>
		/// Image copy.
		/// </summary>
		/// <param name="dst">Destination image</param>
		/// <param name="channel">Channel number. This number is added to the dst pointer</param>
		public void Copy(NPPImage_16uC1 dst, int channel)
		{
			if (channel < 0 | channel >= _channels) throw new ArgumentOutOfRangeException("channel", "channel must be in range [0..2].");
			status = NPPNativeMethods.NPPi.MemCopy.nppiCopy_16u_C3C1R(_devPtrRoi + channel * _typeSize, _pitch, dst.DevicePointerRoi, dst.Pitch, _sizeRoi);
			Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiCopy_16u_C3C1R", status));
			NPPException.CheckNppStatus(status, this);
		}

 /// <summary>
 /// In place image division, scale by 2^(-nScaleFactor), then clamp to saturated value.
 /// </summary>
 /// <param name="src2">2nd source image</param>
 /// <param name="nScaleFactor">scaling factor</param>
 public void Div(NPPImage_16uC1 src2, int nScaleFactor)
 {
     status = NPPNativeMethods.NPPi.Div.nppiDiv_16u_C1IRSfs(src2.DevicePointerRoi, src2.Pitch, _devPtrRoi, _pitch, _sizeRoi, nScaleFactor);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiDiv_16u_C1IRSfs", status));
     NPPException.CheckNppStatus(status, this);
 }

		/// <summary>
		/// 32-bit unsigned to 16-bit unsigned conversion.
		/// </summary>
		/// <param name="dst">Destination image</param>
		/// <param name="roundMode">Round mode</param>
		/// <param name="scaleFactor">scaling factor</param>
		public void Convert(NPPImage_16uC1 dst, NppRoundMode roundMode, int scaleFactor)
		{
			status = NPPNativeMethods.NPPi.BitDepthConversion.nppiConvert_32u16u_C1RSfs(_devPtrRoi, _pitch, dst.DevicePointerRoi, dst.Pitch, _sizeRoi, roundMode, scaleFactor);
			Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiConvert_32u16u_C1RSfs", status));
			NPPException.CheckNppStatus(status, this);
		}

 /// <summary>
 /// Compare pSrc1's pixels with corresponding pixels in pSrc2.
 /// </summary>
 /// <param name="src2">2nd source image</param>
 /// <param name="dest">Destination image</param>
 /// <param name="eComparisonOperation">Specifies the comparison operation to be used in the pixel comparison.</param>
 public void Compare(NPPImage_16uC1 src2, NPPImage_8uC1 dest, NppCmpOp eComparisonOperation)
 {
     status = NPPNativeMethods.NPPi.Compare.nppiCompare_16u_C1R(_devPtrRoi, _pitch, src2.DevicePointerRoi, src2.Pitch, dest.DevicePointerRoi, dest.Pitch, _sizeRoi, eComparisonOperation);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiCompare_16u_C1R", status));
     NPPException.CheckNppStatus(status, this);
 }

 /// <summary>
 /// Masked Operation 8-bit unsigned image copy.
 /// </summary>
 /// <param name="dst">Destination image</param>
 /// <param name="mask">Mask image</param>
 public void Copy(NPPImage_16uC1 dst, NPPImage_8uC1 mask)
 {
     status = NPPNativeMethods.NPPi.MemCopy.nppiCopy_16u_C1MR(_devPtrRoi, _pitch, dst.DevicePointerRoi, dst.Pitch, _sizeRoi, mask.DevicePointerRoi, mask.Pitch);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiCopy_16u_C1MR", status));
     NPPException.CheckNppStatus(status, this);
 }

 /// <summary>
 /// In place image logical Xor.
 /// </summary>
 /// <param name="src2">2nd source image</param>
 public void Xor(NPPImage_16uC1 src2)
 {
     status = NPPNativeMethods.NPPi.Xor.nppiXor_16u_C1IR(src2.DevicePointerRoi, src2.Pitch, _devPtrRoi, _pitch, _sizeRoi);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiXor_16u_C1IR", status));
     NPPException.CheckNppStatus(status, this);
 }

 /// <summary>
 /// An input color twist matrix with floating-point pixel values is applied
 /// within ROI.
 /// </summary>
 /// <param name="dest">Destination image</param>
 /// <param name="twistMatrix">The color twist matrix with floating-point pixel values [3,4].</param>
 public void ColorTwist(NPPImage_16uC1 dest, float[,] twistMatrix)
 {
     status = NPPNativeMethods.NPPi.ColorProcessing.nppiColorTwist32f_16u_C1R(_devPtr, _pitch, dest.DevicePointer, dest.Pitch, _sizeRoi, twistMatrix);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiColorTwist32f_16u_C1R", status));
     NPPException.CheckNppStatus(status, this);
 }

 /// <summary>
 /// Perspective transform of an image.<para/>
 /// This function performs perspective warping of a the specified
 /// quadrangle in the source image to the specified quadrangle in the
 /// destination image. The function nppiWarpPerspectiveQuad uses the same
 /// formulas for pixel mapping as in nppiWarpPerspective function. The
 /// transform coefficients are computed internally.
 /// The transformed part of the source image is resampled using the specified
 /// interpolation method and written to the destination ROI.<para/>
 /// NPPI specific recommendation: <para/>
 /// The function operates using 2 types of kernels: fast and accurate. The fast
 /// method is about 4 times faster than its accurate variant,
 /// but doesn't perform memory access checks and requires the destination ROI
 /// to be 64 bytes aligned. Hence any destination ROI is 
 /// chunked into 3 vertical stripes: the first and the third are processed by
 /// accurate kernels and the central one is processed by the fast one.
 /// In order to get the maximum available speed of execution, the projection of
 /// destination ROI onto image addresses must be 64 bytes aligned. This is
 /// always true if the values <para/>
 /// <code>(int)((void *)(pDst + dstRoi.x))</code> and <para/>
 /// <code>(int)((void *)(pDst + dstRoi.x + dstRoi.width))</code> <para/>
 /// are multiples of 64. Another rule of thumb is to specify destination ROI in
 /// such way that left and right sides of the projected image are separated from
 /// the ROI by at least 63 bytes from each side. However, this requires the
 /// whole ROI to be part of allocated memory. In case when the conditions above
 /// are not satisfied, the function may decrease in speed slightly and will
 /// return NPP_MISALIGNED_DST_ROI_WARNING warning.
 /// </summary>
 /// <param name="srcQuad">Source quadrangle [4,2]</param>
 /// <param name="dest">Destination image</param>
 /// <param name="destQuad">Destination quadrangle [4,2]</param>
 /// <param name="eInterpolation">Interpolation mode: can be <see cref="InterpolationMode.NearestNeighbor"/>, <see cref="InterpolationMode.Linear"/> or <see cref="InterpolationMode.Cubic"/></param>
 public void WarpPerspectiveQuad(double[,] srcQuad, NPPImage_16uC1 dest, double[,] destQuad, InterpolationMode eInterpolation)
 {
     NppiRect rectIn = new NppiRect(_pointRoi, _sizeRoi);
     NppiRect rectOut = new NppiRect(dest.PointRoi, dest.SizeRoi);
     status = NPPNativeMethods.NPPi.PerspectiveTransforms.nppiWarpPerspectiveQuad_16u_C1R(_devPtr, _sizeOriginal, _pitch, rectIn, srcQuad, dest.DevicePointer, dest.Pitch, rectOut, destQuad, eInterpolation);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiWarpPerspectiveQuad_16u_C1R", status));
     NPPException.CheckNppStatus(status, this);
 }

        /// <summary>
        /// image average relative error.
        /// </summary>
        /// <param name="src2">2nd source image</param>
        /// <param name="pError">Pointer to the computed error.</param>
        /// <param name="buffer">Pointer to the user-allocated scratch buffer required for the AverageRelativeError operation.</param>
        public void AverageRelativeError(NPPImage_16uC1 src2, CudaDeviceVariable<double> pError, CudaDeviceVariable<byte> buffer)
        {
            int bufferSize = AverageRelativeErrorGetBufferHostSize();
            if (bufferSize > buffer.Size) throw new NPPException("Provided buffer is too small.");

            status = NPPNativeMethods.NPPi.AverageRelativeError.nppiAverageRelativeError_16u_C1R(_devPtrRoi, _pitch, src2.DevicePointerRoi, src2.Pitch, _sizeRoi, pError.DevicePointer, buffer.DevicePointer);
            Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiAverageRelativeError_16u_C1R", status));
            NPPException.CheckNppStatus(status, this);
        }

 /// <summary>
 /// image transpose
 /// </summary>
 /// <param name="dest">Destination image</param>
 public void Transpose(NPPImage_16uC1 dest)
 {
     status = NPPNativeMethods.NPPi.Transpose.nppiTranspose_16u_C1R(_devPtrRoi, _pitch, dest.DevicePointerRoi, dest.Pitch, _sizeRoi);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiTranspose_16u_C1R", status));
     NPPException.CheckNppStatus(status, this);
 }

 /// <summary>
 /// 3x3 dilation with border control.
 /// </summary>
 /// <param name="dest">Destination image</param>
 /// <param name="eBorderType">The border type operation to be applied at source image border boundaries.</param>
 public void Dilate3x3Border(NPPImage_16uC1 dest, NppiBorderType eBorderType)
 {
     status = NPPNativeMethods.NPPi.Dilate3x3Border.nppiDilate3x3Border_16u_C1R(_devPtr, _pitch, _sizeOriginal, _pointRoi, dest.DevicePointerRoi, dest.Pitch, _sizeRoi, eBorderType);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiDilate3x3Border_16u_C1R", status));
     NPPException.CheckNppStatus(status, this);
 }

 /// <summary>
 /// linearly interpolated source image subpixel coordinate color copy.
 /// </summary>
 /// <param name="dst">Destination-Image</param>
 /// <param name="nDx">Fractional part of source image X coordinate.</param>
 /// <param name="nDy">Fractional part of source image Y coordinate.</param>
 public void CopySubpix(NPPImage_16uC1 dst, float nDx, float nDy)
 {
     status = NPPNativeMethods.NPPi.CopySubpix.nppiCopySubpix_16u_C1R(_devPtrRoi, _pitch, dst.DevicePointerRoi, dst.Pitch, _sizeRoi, nDx, nDy);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiCopySubpix_16u_C1R", status));
     NPPException.CheckNppStatus(status, this);
 }

 /// <summary>
 /// Dilation computes the output pixel as the maximum pixel value of the pixels under the mask. Pixels who’s
 /// corresponding mask values are zero to not participate in the maximum search. With border control.
 /// </summary>
 /// <param name="dest">Destination image</param>
 /// <param name="Mask">Pointer to the start address of the mask array.</param>
 /// <param name="aMaskSize">Width and Height mask array.</param>
 /// <param name="oAnchor">X and Y offsets of the mask origin frame of reference w.r.t the source pixel.</param>
 /// <param name="eBorderType">The border type operation to be applied at source image border boundaries.</param>
 public void DilateBorder(NPPImage_16uC1 dest, CudaDeviceVariable<byte> Mask, NppiSize aMaskSize, NppiPoint oAnchor, NppiBorderType eBorderType)
 {
     status = NPPNativeMethods.NPPi.DilationWithBorderControl.nppiDilateBorder_16u_C1R(_devPtr, _pitch, _sizeOriginal, _pointRoi, dest.DevicePointerRoi, dest.Pitch, _sizeRoi, Mask.DevicePointer, aMaskSize, oAnchor, eBorderType);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiDilateBorder_16u_C1R", status));
     NPPException.CheckNppStatus(status, this);
 }

 /// <summary>
 /// image copy with the borders wrapped by replication of source image pixel colors.
 /// </summary>
 /// <param name="dst">Destination-Image</param>
 /// <param name="nTopBorderHeight">Height (in pixels) of the top border. The height of the border at the bottom of
 /// the destination ROI is implicitly defined by the size of the source ROI: nBottomBorderHeight =
 /// oDstSizeROI.height - nTopBorderHeight - oSrcSizeROI.height.</param>
 /// <param name="nLeftBorderWidth">Width (in pixels) of the left border. The width of the border at the right side of
 /// the destination ROI is implicitly defined by the size of the source ROI: nRightBorderWidth =
 /// oDstSizeROI.width - nLeftBorderWidth - oSrcSizeROI.width.</param>
 public void CopyWrapBorder(NPPImage_16uC1 dst, int nTopBorderHeight, int nLeftBorderWidth)
 {
     status = NPPNativeMethods.NPPi.CopyWrapBorder.nppiCopyWrapBorder_16u_C1R(_devPtrRoi, _pitch, _sizeRoi, dst.DevicePointerRoi, dst.Pitch, dst.SizeRoi, nTopBorderHeight, nLeftBorderWidth);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiCopyWrapBorder_16u_C1R", status));
     NPPException.CheckNppStatus(status, this);
 }

        /// <summary>
        /// CrossCorrSame_NormLevel.
        /// </summary>
        /// <param name="tpl">template image.</param>
        /// <param name="dst">Destination image</param>
        /// <param name="buffer">Allocated device memory with size of at <see cref="SameNormLevelGetBufferHostSize()"/></param>
        public void CrossCorrSame_NormLevel(NPPImage_16uC1 tpl, NPPImage_32fC1 dst, CudaDeviceVariable<byte> buffer)
        {
            int bufferSize = SameNormLevelGetBufferHostSize();
            if (bufferSize > buffer.Size) throw new NPPException("Provided buffer is too small.");

            status = NPPNativeMethods.NPPi.ImageProximity.nppiCrossCorrSame_NormLevel_16u32f_C1R(_devPtrRoi, _pitch, _sizeRoi, tpl.DevicePointerRoi, tpl.Pitch, tpl.SizeRoi, dst.DevicePointer, dst.Pitch, buffer.DevicePointer);
            Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiCrossCorrSame_NormLevel_16u32f_C1R", status));
            NPPException.CheckNppStatus(status, this);
        }

		/// <summary>
		/// Perspective transform of an image.<para/>
		/// This function performs perspective warping of a the specified
		/// quadrangle in the source image to the specified quadrangle in the
		/// destination image. The function nppiWarpPerspectiveQuad uses the same
		/// formulas for pixel mapping as in nppiWarpPerspective function. The
		/// transform coefficients are computed internally.
		/// The transformed part of the source image is resampled using the specified
		/// interpolation method and written to the destination ROI.<para/>
		/// NPPI specific recommendation: <para/>
		/// The function operates using 2 types of kernels: fast and accurate. The fast
		/// method is about 4 times faster than its accurate variant,
		/// but doesn't perform memory access checks and requires the destination ROI
		/// to be 64 bytes aligned. Hence any destination ROI is 
		/// chunked into 3 vertical stripes: the first and the third are processed by
		/// accurate kernels and the central one is processed by the fast one.
		/// In order to get the maximum available speed of execution, the projection of
		/// destination ROI onto image addresses must be 64 bytes aligned. This is
		/// always true if the values <para/>
		/// <code>(int)((void *)(pDst + dstRoi.x))</code> and <para/>
		/// <code>(int)((void *)(pDst + dstRoi.x + dstRoi.width))</code> <para/>
		/// are multiples of 64. Another rule of thumb is to specify destination ROI in
		/// such way that left and right sides of the projected image are separated from
		/// the ROI by at least 63 bytes from each side. However, this requires the
		/// whole ROI to be part of allocated memory. In case when the conditions above
		/// are not satisfied, the function may decrease in speed slightly and will
		/// return NPP_MISALIGNED_DST_ROI_WARNING warning.
		/// </summary>
		/// <param name="src0">Source image (Channel 0)</param>
		/// <param name="src1">Source image (Channel 1)</param>
		/// <param name="src2">Source image (Channel 2)</param>
		/// <param name="srcQuad">Source quadrangle [4,2]</param>
		/// <param name="dest0">Destination image (Channel 0)</param>
		/// <param name="dest1">Destination image (Channel 1)</param>
		/// <param name="dest2">Destination image (Channel 2)</param>
		/// <param name="destQuad">Destination quadrangle [4,2]</param>
		/// <param name="eInterpolation">Interpolation mode: can be <see cref="InterpolationMode.NearestNeighbor"/>, <see cref="InterpolationMode.Linear"/> or <see cref="InterpolationMode.Cubic"/></param>
		public static void WarpPerspectiveQuad(NPPImage_16uC1 src0, NPPImage_16uC1 src1, NPPImage_16uC1 src2, double[,] srcQuad, NPPImage_16uC1 dest0, NPPImage_16uC1 dest1, NPPImage_16uC1 dest2, double[,] destQuad, InterpolationMode eInterpolation)
		{
			NppiRect rectIn = new NppiRect(src0.PointRoi, src0.SizeRoi);
			NppiRect rectOut = new NppiRect(dest0.PointRoi, dest0.SizeRoi);

			CUdeviceptr[] src = new CUdeviceptr[] { src0.DevicePointer, src1.DevicePointer, src2.DevicePointer };
			CUdeviceptr[] dst = new CUdeviceptr[] { dest0.DevicePointer, dest1.DevicePointer, dest2.DevicePointer };

			NppStatus status = NPPNativeMethods.NPPi.PerspectiveTransforms.nppiWarpPerspectiveQuad_16u_P3R(src, src0.Size, src0.Pitch, rectIn, srcQuad, dst, dest0.Pitch, rectOut, destQuad, eInterpolation);
			Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiWarpPerspectiveQuad_16u_P3R", status));
			NPPException.CheckNppStatus(status, null);
		}

 /// <summary>
 /// image CrossCorrValid_Norm.
 /// </summary>
 /// <param name="tpl">template image.</param>
 /// <param name="dst">Destination-Image</param>
 public void CrossCorrValid_Norm(NPPImage_16uC1 tpl, NPPImage_32fC1 dst)
 {
     status = NPPNativeMethods.NPPi.ImageProximity.nppiCrossCorrValid_Norm_16u32f_C1R(_devPtrRoi, _pitch, _sizeRoi, tpl.DevicePointerRoi, tpl.Pitch, tpl.SizeRoi, dst.DevicePointerRoi, dst.Pitch);
     Debug.WriteLine(String.Format("{0:G}, {1}: {2}", DateTime.Now, "nppiCrossCorrValid_Norm_16u32f_C1R", status));
     NPPException.CheckNppStatus(status, this);
 }