边缘检测#
实验介绍#
Sobel算子,也被称为Sobel-Feldman算子,或者Sobel滤波,是在图像处理和计算机视觉得到广泛应用的一种图像边缘检测算法。它由斯坦福大学人工智能实验室（SAIL）的Irwin Sobel和Gray Feldman而得名。 Sobel算子使用两个（3x3）矩阵来对原图进行卷积运算以计算出两个方向的灰度差分（偏导）的估计值（一个水平方向、一个竖直方向）。我们假定A是原始图像（彩色图像需先转换为灰度图像）,Gx和Gy分别是在横向及纵向的灰度偏导的近似值（即两个方向上对原图的平面卷积结果）。数学表达如下：
对应的计算过程如下：
Gx = [ f(x+1,y-1)+2*f(x+1,y)+f(x+1,y+1)] - [f(x-1,y-1)+2*f(x-1,y)+f(x-1,y+1) ]
Gy = [ f(x-1,y-1) + 2f(x,y-1) + f(x+1,y-1)] - [f(x-1, y+1) + 2*f(x,y+1)+f(x+1,y+1)
上式中,f(x,y)为图像A中(x,y)处的灰度值。由此便可以计算出每个点的Gx和Gy。 对于图像中的每个点,其梯度的估计值G便可以通过两个方向的梯度Gx和Gy借由下式得出：
模块主要代码#
#include "edge_detector.h"
void ExtractPixel(XF_TNAME(XF_8UC3,XF_NPPC1)&src,ap_uint<8>dst[3])
{
unsigned int i,j=0;
for(i=0;i<24;i+=8)
{
dst[j]=src.range(i+7,i);
j++;
}
}
template<int ROWS,int COLS>
void thresholding(xf::cv::Mat<XF_8UC1,IMG_MAX_ROWS, IMG_MAX_COLS,XF_NPPC1> &src,
xf::cv::Mat<XF_8UC1,IMG_MAX_ROWS, IMG_MAX_COLS,XF_NPPC1> &dst,
ap_uint<8>&threshold)
{
XF_TNAME(XF_8UC1,XF_NPPC1) pixel_src, pixel_dst;
unsigned int i,j=0;
ap_uint<8>value;
if(threshold==0)
{
for(i=0;i<ROWS;i++)
{
for(j=0;j<COLS;j++)
{
#pragma HLS PIPELINE
pixel_src=src.read(i*COLS+j);
dst.write(i*COLS+j,pixel_src);
}
}
}
else
{
for(i=0;i<ROWS;i++)
{
for(j=0;j<COLS;j++)
{
#pragma HLS PIPELINE
pixel_src=src.read(i*COLS+j);
if(pixel_src>threshold)
{
pixel_dst=0;
}
else
{
pixel_dst=255;
}
dst.write(i*COLS+j,pixel_dst);
}
}
}
}
template <int ROWS, int COLS>
void xfrgb2gray(xf::cv::Mat<XF_8UC3, ROWS, COLS, XF_NPPC1>& src,
xf::cv::Mat<XF_8UC1, ROWS, COLS, XF_NPPC1>& dst)
{
unsigned int i,j=0;
XF_TNAME(XF_8UC3,XF_NPPC1)pixel_src;
XF_TNAME(XF_8UC1,XF_NPPC1)pixel_dst;
ap_uint<8>value[3];
ap_uint<8>gray;
for(i=0;i<ROWS;i++)
{
for(j=0;j<COLS;j++)
{
pixel_src=src.read(i*COLS+j);
ExtractPixel(pixel_src,value);
gray=CalculateGRAY(value[0],value[1],value[2]);
pixel_dst.range(7,0)=gray;
dst.write(i*COLS+j,pixel_dst);
}
}
}
template<int ROWS,int COLS>
void AddWeightedKernel(xf::cv::Mat<XF_8UC1,ROWS,COLS,XF_NPPC1>&src1,
float alpha,
xf::cv::Mat<XF_8UC1,ROWS,COLS,XF_NPPC1>&src2,
float beta,
float gamma,
xf::cv::Mat<XF_8UC1,ROWS,COLS,XF_NPPC1>&dst
)
{
ap_fixed<16,8,AP_RND>value_src1=alpha;
ap_fixed<16,8,AP_RND>value_src2=beta;
ap_fixed<16,8,AP_RND>value_src3=gamma;
XF_TNAME(XF_8UC1,XF_NPPC1)pixel1;
XF_TNAME(XF_8UC1,XF_NPPC1)pixel2;
XF_TNAME(XF_8UC1,XF_NPPC1)pixel3;
ap_int<24>firstcmp;
ap_int<24>secondcmp;
ap_int<16>thirdcmp;
ap_uint<8>value;
ap_uint<8>value_cmp1;
ap_uint<8>value_cmp2;
unsigned int i,j=0;
for(i=0;i<ROWS;i++)
{
for(j=0;j<COLS;j++)
{
#pragma HLS pipeline
pixel1=src1.read(i*COLS+j);
pixel2=src2.read(i*COLS+j);
value_cmp1=pixel1.range(7,0);
value_cmp2=pixel2.range(7,0);
firstcmp=(ap_int<24>)value_cmp1*value_src1;
secondcmp=(ap_int<24>)value_cmp2*value_src2;
thirdcmp=(ap_int<16>)firstcmp+secondcmp+value_src3;
if(thirdcmp>255)
{
thirdcmp=255;
}
else if(thirdcmp<0)
{
thirdcmp=0;
}
value=thirdcmp;
pixel3.range(7,0)=value;
dst.write(i*COLS+j,pixel3);
}
}
}
template<int ROWS,int COLS>
void xfgray2rgb(xf::cv::Mat<XF_8UC1,ROWS,COLS,XF_NPPC1>&src,xf::cv::Mat<XF_8UC3,ROWS,COLS,XF_NPPC1> &dst)
{
unsigned int i,j=0;
XF_TNAME(XF_8UC1,XF_NPPC1)temp_src;
XF_TNAME(XF_8UC3,XF_NPPC1)temp_dst;
ap_uint<8>value;
for(i=0;i<ROWS;i++)
{
for(j=0;j<COLS;j++)
{
#pragma HLS PIPELINE
temp_src=src.read(i*COLS+j);
value=temp_src.range(7,0);
temp_dst.range(7,0)=value;
temp_dst.range(15,8)=value;
temp_dst.range(23,16)=value;
dst.write(i*COLS+j,temp_dst);
}
}
}
void edge_detector(ustream_t &src, ustream_t &dst, ap_uint<8> threshold)
{
#pragma HLS DATAFLOW
#pragma HLS INTERFACE axis port=src
#pragma HLS INTERFACE axis port=dst
#pragma HLS INTERFACE mode=s_axilite port=threshold register
#pragma HLS INTERFACE mode=ap_ctrl_none port=return
xf::cv::Mat<XF_8UC3,IMG_MAX_ROWS, IMG_MAX_COLS,XF_NPPC1> srcImg;
#pragma HLS STREAM depth=1920 type=fifo variable=srcImg
xf::cv::Mat<XF_8UC3,IMG_MAX_ROWS, IMG_MAX_COLS,XF_NPPC1> rgbSobel;
#pragma HLS STREAM depth=1920 type=fifo variable=rgbSobel
xf::cv::Mat<XF_8UC1,IMG_MAX_ROWS, IMG_MAX_COLS,XF_NPPC1> grayImg;
#pragma HLS STREAM depth=1920 type=fifo variable=grayImg
xf::cv::Mat<XF_8UC1,IMG_MAX_ROWS, IMG_MAX_COLS,XF_NPPC1> sobelImg_x;
#pragma HLS STREAM depth=1920 variable=sobelImg_x
xf::cv::Mat<XF_8UC1,IMG_MAX_ROWS, IMG_MAX_COLS,XF_NPPC1> sobelImg_y;
#pragma HLS STREAM depth=1920 type=fifo variable=sobelImg_y
xf::cv::Mat<XF_8UC1,IMG_MAX_ROWS, IMG_MAX_COLS,XF_NPPC1> sobelImg;
#pragma HLS STREAM depth=1920 type=fifo variable=sobelImg
xf::cv::Mat<XF_8UC1,IMG_MAX_ROWS, IMG_MAX_COLS,XF_NPPC1> thresholdImg;
#pragma HLS STREAM depth=1920 variable=thresholdImg
xf::cv::AXIvideo2xfMat(src, srcImg);
xfrgb2gray<IMG_MAX_ROWS,IMG_MAX_COLS>(srcImg, grayImg);
xf::cv::xFSobelFilter3x3<XF_8UC1, XF_8UC1,IMG_MAX_ROWS, IMG_MAX_COLS, XF_CHANNELS(XF_8UC1,XF_NPPC1), XF_DEPTH(XF_8UC1,XF_NPPC1), XF_DEPTH(XF_8UC1,XF_NPPC1),
XF_NPPC1,_XFCVDEPTH_DEFAULT,_XFCVDEPTH_DEFAULT,_XFCVDEPTH_DEFAULT,XF_WORDWIDTH(XF_8UC1,XF_NPPC1), XF_WORDWIDTH(XF_8UC1,XF_NPPC1), (IMG_MAX_COLS >> XF_BITSHIFT(XF_NPPC1)),false>(
grayImg,sobelImg_x,sobelImg_y,grayImg.rows,grayImg.cols>>XF_BITSHIFT(XF_NPPC1));
AddWeightedKernel<IMG_MAX_ROWS,IMG_MAX_COLS>(sobelImg_x,0.5f,sobelImg_y,0.5f,0.0f,sobelImg);
thresholding<IMG_MAX_ROWS,IMG_MAX_COLS>(sobelImg, thresholdImg,threshold);
xfgray2rgb<IMG_MAX_ROWS,IMG_MAX_COLS>(thresholdImg, rgbSobel);
xf::cv::xfMat2AXIvideo(rgbSobel, dst);
}
工程路径#
名称 |
路径 |
|---|---|
vivado 工程 |
vivado/ edge_detector |