WarpAffine

Affine Transformation is used to express rotation, scaling and translation operations.

xFWarpAffine takes inverse transformation as input and transforms the input image by matrix multiplication with 2x3 matrix.Calculate the input position corresponding to every output pixel in the following way.

Note: Prior scaling of greater than or equal to 0.25 and less than or equal to 8 are supported and throws an assertion if the transformation matrix passed is having unsupported scaling factors.

API Syntax

template<int INTERPOLATION_TYPE, int SRC_T, int ROWS, int COLS, int NPC=1>
void xFwarpAffine(xF::Mat<SRC_T, ROWS, COLS, XF_NPPC8> & _src, xF::Mat<SRC_T, ROWS, COLS, XF_NPPC8> & _dst, float* transformation_matrix)

Parameter Descriptions

The following table describes the template and the function parameters.

Table 1. xFwarpAffine Function Parameter Descriptions
Parameter	Description
INTERPOLATION_TYPE	Interpolation technique to be used XF_INTERPOLATION_NN - Nearest Neighbor XF_INTERPOLATION_BILINEAR - Bilinear
SRC_T	Input pixel type.
ROWS	Maximum height of input and output image (must be a multiple of 8)
COLS	Maximum width of input and output image (must be a multiple of 8)
NPC	Number of pixels to be processed per cycle; possible options are XF_NPPC1 and XF_NPPC8 for 1 pixel and 8 pixel operations respectively.
_src	Input Image pointer
_dst	Output Image pointer
transformation_matrix	Inverse Affine Transformation matrix

Resource Utilization

The following table shows the resource utilization of xFWarpAffine in different configurations as generated in Vivado HLS 2017.1 version tool for the part Xilinx Xczu9eg-ffvb1156-1-i-es1.

Table 2. xFwarpAffine Function Resource Utilization Summary
Name	Utilization (at 250 MHz)
	Normal Operation (1 pixel) Mode		Resource Optimized (8 pixel) Mode
	Nearest Neighbour	Bilinear Inerpolation	Nearest Neighbour	Bilinear Interpolation
BRAM	200	194	200	200
DSP48E	121	125	139	155
FF	9523	10004	8070	9570
LUT	9928	10592	13296	13894
CLB	2390	2435	2932	2829

Note: NO is single pixel processing and RO is 4-pixel processing.

Performance Estimate

The following table shows the performance of Affine Transformation in different configurations, as generated in Vivado HLS 2017.1 version tool for the part Xilinx Xczu9eg-ffvb1156-1-i-es1.

Table 3. xFwarpAffine Function Performance Estimate Summary
Latency Estimate
1pixel(300 MHz)		8pixel(150 MHz)
Nearest Neighbour	Bilinear Interpolation	Nearest Neighbour	Bilinear Interpolation
Max(in ms)	Max(in ms)	Max(in ms)	Max(in ms)
10.4	19.3	10.7	10.4