图像特征检测(Image Feature Detection)
作者：王先荣
前言
图像特征提取是计算机视觉和图像处理中的一个概念。它指的是使用计算机提取图像信息，决定每个图像的点是否属于一个图像特征。本文主要探讨如何提取图像中的“角点”这一特征，及其相关的内容。而诸如直方图、边缘、区域等内容在前文中有所提及，请查看相关文章。OpenCv（EmguCv）中实现了多种角点特征的提取方法，包括：Harris角点、ShiTomasi角点、亚像素级角点、SURF角点、Star关键点、FAST关键点、Lepetit关键点等等，本文将逐一介绍如何检测这些角点。在此之前将会先介绍跟角点检测密切相关的一些变换，包括Sobel算子、拉普拉斯算子、Canny算子、霍夫变换。另外，还会介绍一种广泛使用而OpenCv中并未实现的SIFT角点检测，以及最近在OpenCv中实现的MSER区域检测。所要讲述的内容会很多，我这里尽量写一些需要注意的地方及实现代码，而参考手册及书本中有的内容将一笔带过或者不会提及。
Sobel算子
Sobel算子用多项式计算来拟合导数计算，可以用OpenCv中的cvSobel函数或者EmguCv中的Image<TColor,TDepth>.Sobel方法来进行计算。需要注意的是，xorder和yorder中必须且只能有一个为非零值，即只能计算x方向或者y反向的导数；如果将方形滤波器的宽度设置为特殊值CV_SCHARR（-1），将使用Scharr滤波器代替Sobel滤波器。
使用Sobel滤波器的示例代码如下：
//
Sobel算子
private
string
SobelFeatureDetect()
{
//
获取参数
int
xOrder
=
int
.Parse((
string
)cmbSobelXOrder.SelectedItem);
int
yOrder
=
int
.Parse((
string
)cmbSobelYOrder.SelectedItem);
int
apertureSize
=
int
.Parse((
string
)cmbSobelApertureSize.SelectedItem);
if
((xOrder
==
0
&&
yOrder
==
0
)
||
(xOrder
!=
0
&&
yOrder
!=
0
))
return
"
Sobel算子，参数错误：xOrder和yOrder中必须且只能有一个非零。\r\n
"
;
//
计算
Stopwatch sw
=
new
Stopwatch();
sw.Start();
Image
<
Gray, Single
>
imageDest
=
imageSourceGrayscale.Sobel(xOrder, yOrder, apertureSize);
sw.Stop();
//
显示
pbResult.Image
=
imageDest.Bitmap;
//
释放资源
imageDest.Dispose();
//
返回
return
string
.Format(
"
·Sobel算子，用时{0:F05}毫秒，参数（x方向求导阶数：{1}，y方向求导阶数：{2}，方形滤波器宽度：{3}）\r\n
"
, sw.Elapsed.TotalMilliseconds, xOrder, yOrder, apertureSize);
}
拉普拉斯算子
拉普拉斯算子可以用作边缘检测；可以用OpenCv中的cvLaplace函数或者EmguCv中的Image<TColor,TDepth>.Laplace方法来进行拉普拉斯变换。需要注意的是：OpenCv的文档有点小错误，apertureSize参数值不能为CV_SCHARR（-1）。
使用拉普拉斯变换的示例代码如下：
//
拉普拉斯变换
private
string
LaplaceFeatureDetect()
{
//
获取参数
int
apertureSize
=
int
.Parse((
string
)cmbLaplaceApertureSize.SelectedItem);
//
计算
Stopwatch sw
=
new
Stopwatch();
sw.Start();
Image
<
Gray, Single
>
imageDest
=
imageSourceGrayscale.Laplace(apertureSize);
sw.Stop();
//
显示
pbResult.Image
=
imageDest.Bitmap;
//
释放资源
imageDest.Dispose();
//
返回
return
string
.Format(
"
·拉普拉斯变换，用时{0:F05}毫秒，参数（方形滤波器宽度：{1}）\r\n
"
, sw.Elapsed.TotalMilliseconds, apertureSize);
}
Canny算子
Canny算子也可以用作边缘检测；可以用OpenCv中的cvCanny函数或者EmguCv中的Image<TColor,TDepth>.Canny方法来进行Canny边缘检测。所不同的是，Image<TColor,TDepth>.Canny方法可以用于检测彩色图像的边缘，但是它只能使用apertureSize参数的默认值3；
而cvCanny只能处理灰度图像，不过可以自定义apertureSize。cvCanny和Canny的方法参数名有点点不同，下面是参数对照表。
Image<TColor,TDepth>.Canny CvInvoke.cvCanny
thresh lowThresh
threshLinking highThresh
3 apertureSize
值得注意的是，apertureSize只能取3，5或者7，这可以在cvcanny.cpp第87行看到：
aperture_size &= INT_MAX; if ( (aperture_size & 1 ) == 0 || aperture_size < 3 || aperture_size > 7 ) CV_ERROR( CV_StsBadFlag, "" );
使用Canny算子的示例代码如下：
//
Canny算子
private
string
CannyFeatureDetect()
{
//
获取参数
double
lowThresh
=
double
.Parse(txtCannyLowThresh.Text);
double
highThresh
=
double
.Parse(txtCannyHighThresh.Text);
int
apertureSize
=
int
.Parse((
string
)cmbCannyApertureSize.SelectedItem);
//
计算
Stopwatch sw
=
new
Stopwatch();
sw.Start();
Image
<
Gray, Byte
>
imageDest
=
null
;
Image
<
Bgr, Byte
>
imageDest2
=
null
;
if
(rbCannyUseCvCanny.Checked)
{
imageDest
=
new
Image
<
Gray,
byte
>
(imageSourceGrayscale.Size);
CvInvoke.cvCanny(imageSourceGrayscale.Ptr, imageDest.Ptr, lowThresh, highThresh, apertureSize);
}
else
imageDest2
=
imageSource.Canny(
new
Bgr(lowThresh, lowThresh, lowThresh),
new
Bgr(highThresh, highThresh, highThresh));
sw.Stop();
//
显示
pbResult.Image
=
rbCannyUseCvCanny.Checked
?
imageDest.Bitmap : imageDest2.Bitmap;
//
释放资源
if
(imageDest
!=
null
)
imageDest.Dispose();
if
(imageDest2
!=
null
)
imageDest2.Dispose();
//
返回
return
string
.Format(
"
·Canny算子，用时{0:F05}毫秒，参数（方式：{1}，阀值下限：{2}，阀值上限：{3}，方形滤波器宽度：{4}）\r\n
"
, sw.Elapsed.TotalMilliseconds, rbCannyUseCvCanny.Checked
?
"
cvCanny
"
:
"
Image<TColor, TDepth>.Canny
"
, lowThresh, highThresh, apertureSize);
}
另外，在http://www.china-vision.net/blog/user2/15975/archives/2007/804.html有一种自动获取Canny算子高低阀值的方法，作者提供了用C语言实现的代码。我将其改写成了C#版本，代码如下：
///
<summary>
///
计算图像的自适应Canny算子阀值
///
</summary>
///
<param name="imageSrc">
源图像，只能是256级灰度图像
</param>
///
<param name="apertureSize">
方形滤波器的宽度
</param>
///
<param name="lowThresh">
阀值下限
</param>
///
<param name="highThresh">
阀值上限
</param>
unsafe
void
AdaptiveFindCannyThreshold(Image
<
Gray, Byte
>
imageSrc,
int
apertureSize,
out
double
lowThresh,
out
double
highThresh)
{
//
计算源图像x方向和y方向的1阶Sobel算子
Size size
=
imageSrc.Size;
Image
<
Gray, Int16
>
imageDx
=
new
Image
<
Gray,
short
>
(size);
Image
<
Gray, Int16
>
imageDy
=
new
Image
<
Gray,
short
>
(size);
CvInvoke.cvSobel(imageSrc.Ptr, imageDx.Ptr,
1
,
0
, apertureSize);
CvInvoke.cvSobel(imageSrc.Ptr, imageDy.Ptr,
0
,
1
, apertureSize);
Image
<
Gray, Single
>
image
=
new
Image
<
Gray,
float
>
(size);
int
i, j;
DenseHistogram hist
=
null
;
int
hist_size
=
255
;
float
[] range_0
=
new
float
[] {
0
,
256
};
double
PercentOfPixelsNotEdges
=
0.7
;
//
计算边缘的强度，并保存于图像中
float
maxv
=
0
;
float
temp;
byte
*
imageDataDx
=
(
byte
*
)imageDx.MIplImage.imageData.ToPointer();
byte
*
imageDataDy
=
(
byte
*
)imageDy.MIplImage.imageData.ToPointer();
byte
*
imageData
=
(
byte
*
)image.MIplImage.imageData.ToPointer();
int
widthStepDx
=
imageDx.MIplImage.widthStep;
int
widthStepDy
=
widthStepDx;
int
widthStep
=
image.MIplImage.widthStep;
for
(i
=
0
; i
<
size.Height; i
++
)
{
short
*
_dx
=
(
short
*
)(imageDataDx
+
widthStepDx
*
i);
short
*
_dy
=
(
short
*
)(imageDataDy
+
widthStepDy
*
i);
float
*
_image
=
(
float
*
)(imageData
+
widthStep
*
i);
for
(j
=
0
; j
<
size.Width; j
++
)
{
temp
=
(
float
)(Math.Abs(
*
(_dx
+
j))
+
Math.Abs(
*
(_dy
+
j)));
*
(_image
+
j)
=
temp;
if
(maxv
<
temp)
maxv
=
temp;
}
}
//
计算直方图
range_0[
1
]
=
maxv;
hist_size
=
hist_size
>
maxv
?
(
int
)maxv : hist_size;
hist
=
new
DenseHistogram(hist_size,
new
RangeF(range_0[
0
], range_0[
1
]));
hist.Calculate
<
Single
>
(
new
Image
<
Gray, Single
>
[] { image },
false
,
null
);
int
total
=
(
int
)(size.Height
*
size.Width
*
PercentOfPixelsNotEdges);
double
sum
=
0
;
int
icount
=
hist.BinDimension[
0
].Size;
for
(i
=
0
; i
<
icount; i
++
)
{
sum
+=
hist[i];
if
(sum
>
total)
break
;
}
//
计算阀值
highThresh
=
(i
+
1
)
*
maxv
/
hist_size;
lowThresh
=
highThresh
*
0.4
;
//
释放资源
imageDx.Dispose();
imageDy.Dispose(); image.Dispose();
hist.Dispose();
}
霍夫变换
霍夫变换是一种在图像中寻找直线、圆及其他简单形状的方法，在OpenCv中实现了霍夫线变换和霍夫圆变换。值得注意的地方有以下几点：（1）HoughLines2需要先计算Canny边缘，然后再检测直线；（2）HoughLines2计算结果的获取随获取方式的不同而不同；（3）HoughCircles检测结果似乎不正确。
使用霍夫变换的示例代码如下所示：
//
霍夫线变换
private
string
HoughLinesFeatureDetect()
{
//
获取参数
HOUGH_TYPE method
=
rbHoughLinesSHT.Checked
?
HOUGH_TYPE.CV_HOUGH_STANDARD : (rbHoughLinesPPHT.Checked
?
HOUGH_TYPE.CV_HOUGH_PROBABILISTIC : HOUGH_TYPE.CV_HOUGH_MULTI_SCALE);
double
rho
=
double
.Parse(txtHoughLinesRho.Text);
double
theta
=
double
.Parse(txtHoughLinesTheta.Text);
int
threshold
=
int
.Parse(txtHoughLinesThreshold.Text);
double
param1
=
double
.Parse(txtHoughLinesParam1.Text);
double
param2
=
double
.Parse(txtHoughLinesParam2.Text);
MemStorage storage
=
new
MemStorage();
int
linesCount
=
0
;
StringBuilder sbResult
=
new
StringBuilder();
//
计算，先运行Canny边缘检测（参数来自Canny算子属性页），然后再用计算霍夫线变换
double
lowThresh
=
double
.Parse(txtCannyLowThresh.Text);
double
highThresh
=
double
.Parse(txtCannyHighThresh.Text);
int
apertureSize
=
int
.Parse((
string
)cmbCannyApertureSize.SelectedItem);
Image
<
Gray, Byte
>
imageCanny
=
new
Image
<
Gray,
byte
>
(imageSourceGrayscale.Size);
CvInvoke.cvCanny(imageSourceGrayscale.Ptr, imageCanny.Ptr, lowThresh, highThresh, apertureSize);
Stopwatch sw
=
new
Stopwatch();
sw.Start();
IntPtr ptrLines
=
CvInvoke.cvHoughLines2(imageCanny.Ptr, storage.Ptr, method, rho, theta, threshold, param1, param2);
Seq
<
LineSegment2D
>
linesSeq
=
null
;
Seq
<
PointF
>
linesSeq2
=
null
;
if
(method
==
HOUGH_TYPE.CV_HOUGH_PROBABILISTIC)
linesSeq
=
new
Seq
<
LineSegment2D
>
(ptrLines, storage);
else
linesSeq2
=
new
Seq
<
PointF
>
(ptrLines, storage);
sw.Stop();
//
显示
Image
<
Bgr, Byte
>
imageResult
=
imageSourceGrayscale.Convert
<
Bgr, Byte
>
();
if
(linesSeq
!=
null
)
{
linesCount
=
linesSeq.Total;
foreach
(LineSegment2D line
in
linesSeq)
{
imageResult.Draw(line,
new
Bgr(255d, 0d, 0d),
4
);
sbResult.AppendFormat(
"
{0}-{1}，
"
, line.P1, line.P2);
}
}
else
{
linesCount
=
linesSeq2.Total;
foreach
(PointF line
in
linesSeq2)
{
float
r
=
line.X;
float
t
=
line.Y;
double
a
=
Math.Cos(t), b
=
Math.Sin(t);
double
x0
=
a
*
r, y0
=
b
*
r;
int
x1
=
(
int
)(x0
+
1000
*
(
-
b));
int
y1
=
(
int
)(y0
+
1000
*
(a));
int
x2
=
(
int
)(x0
-
1000
*
(
-
b));
int
y2
=
(
int
)(y0
-
1000
*
(a));
Point pt1
=
new
Point(x1, y1);
Point pt2
=
new
Point(x2, y2);
imageResult.Draw(
new
LineSegment2D(pt1, pt2),
new
Bgr(255d, 0d, 0d),
4
);
sbResult.AppendFormat(
"
{0}-{1}，
"
, pt1, pt2);
}
}
pbResult.Image
=
imageResult.Bitmap;
//
释放资源
imageCanny.Dispose();
imageResult.Dispose();
storage.Dispose();
//
返回
return
string
.Format(
"
·霍夫线变换，用时{0:F05}毫秒，参数（变换方式：{1}，距离精度：{2}，弧度精度：{3}，阀值：{4}，参数1：{5}，参数2：{6}），找到{7}条直线\r\n{8}
"
,
sw.Elapsed.TotalMilliseconds, method.ToString(
"
G
"
), rho, theta, threshold, param1, param2, linesCount, linesCount
!=
0
?
(sbResult.ToString()
+
"
\r\n
"
) :
""
);
}
//
霍夫圆变换
private
string
HoughCirclesFeatureDetect()
{
//
获取参数
double
dp
=
double
.Parse(txtHoughCirclesDp.Text);
double
minDist
=
double
.Parse(txtHoughCirclesMinDist.Text);
double
param1
=
double
.Parse(txtHoughCirclesParam1.Text);
double
param2
=
double
.Parse(txtHoughCirclesParam2.Text);
int
minRadius
=
int
.Parse(txtHoughCirclesMinRadius.Text);
int
maxRadius
=
int
.Parse(txtHoughCirclesMaxRadius.Text);
StringBuilder sbResult
=
new
StringBuilder();
//
计算
Stopwatch sw
=
new
Stopwatch();
sw.Start();
CircleF[][] circles
=
imageSourceGrayscale.HoughCircles(
new
Gray(param1),
new
Gray(param2), dp, minDist, minRadius, maxRadius);
sw.Stop();
//
显示
Image
<
Bgr, Byte
>
imageResult
=
imageSourceGrayscale.Convert
<
Bgr, Byte
>
();
int
circlesCount
=
0
;
foreach
(CircleF[] cs
in
circles)
{
foreach
(CircleF circle
in
cs)
{
imageResult.Draw(circle,
new
Bgr(255d, 0d, 0d),
4
);
sbResult.AppendFormat(
"
圆心{0}半径{1}，
"
, circle.Center, circle.Radius);
circlesCount
++
;
}
}
pbResult.Image
=
imageResult.Bitmap;
//
释放资源
imageResult.Dispose();
//
返回
return
string
.Format(
"
·霍夫圆变换，用时{0:F05}毫秒，参数（累加器图像的最小分辨率：{1}，不同圆之间的最小距离：{2}，边缘阀值：{3}，累加器阀值：{4}，最小圆半径：{5}，最大圆半径：{6}），找到{7}个圆\r\n{8}
"
,
sw.Elapsed.TotalMilliseconds, dp, minDist, param1, param2, minRadius, maxRadius, circlesCount, sbResult.Length
>
0
?
(sbResult.ToString()
+
"
\r\n
"
) :
""
);
}
Harris角点
cvCornerHarris函数检测的结果实际上是一幅包含Harris角点的浮点型单通道图像，可以使用类似下面的代码来计算包含Harris角点的图像：
//
Harris角点
private
string
CornerHarrisFeatureDetect()
{
//
获取参数
int
blockSize
=
int
.Parse(txtCornerHarrisBlockSize.Text);
int
apertureSize
=
int
.Parse(txtCornerHarrisApertureSize.Text);
double
k
=
double
.Parse(txtCornerHarrisK.Text);
//
计算
Image
<
Gray, Single
>
imageDest
=
new
Image
<
Gray,
float
>
(imageSourceGrayscale.Size);
Stopwatch sw
=
new
Stopwatch();
sw.Start();
CvInvoke.cvCornerHarris(imageSourceGrayscale.Ptr, imageDest.Ptr, blockSize, apertureSize, k);
sw.Stop();
//
显示
pbResult.Image
=
imageDest.Bitmap;
//
释放资源
imageDest.Dispose();
//
返回
return
string
.Format(
"
·Harris角点，用时{0:F05}毫秒，参数（邻域大小：{1}，方形滤波器宽度：{2}，权重系数：{3}）\r\n
"
, sw.Elapsed.TotalMilliseconds, blockSize, apertureSize, k);
}
如果要计算Harris角点列表，需要使用cvGoodFeatureToTrack函数，并传递适当的参数。
ShiTomasi角点
在默认情况下，cvGoodFeatureToTrack函数计算ShiTomasi角点；不过如果将参数use_harris设置为非0值，那么它会计算harris角点。
使用cvGoodFeatureToTrack函数的示例代码如下：
//
ShiTomasi角点
private
string
CornerShiTomasiFeatureDetect()
{
//
获取参数
int
cornerCount
=
int
.Parse(txtGoodFeaturesCornerCount.Text);
double
qualityLevel
=
double
.Parse(txtGoodFeaturesQualityLevel.Text);
double
minDistance
=
double
.Parse(txtGoodFeaturesMinDistance.Text);
int
blockSize
=
int
.Parse(txtGoodFeaturesBlockSize.Text);
bool
useHarris
=
cbGoodFeaturesUseHarris.Checked;
double
k
=
double
.Parse(txtGoodFeaturesK.Text);
//
计算
Stopwatch sw
=
new
Stopwatch();
sw.Start();
PointF[][] corners
=
imageSourceGrayscale.GoodFeaturesToTrack(cornerCount, qualityLevel, minDistance, blockSize, useHarris, k);
sw.Stop();
//
显示
Image
<
Bgr, Byte
>
imageResult
=
imageSourceGrayscale.Convert
<
Bgr, Byte
>
();
int
cornerCount2
=
0
;
StringBuilder sbResult
=
new
StringBuilder();
int
radius
=
(
int
)(minDistance
/
2
)
+
1
;
int
thickness
=
(
int
)(minDistance
/
4
)
+
1
;
foreach
(PointF[] cs
in
corners)
{
foreach
(PointF p
in
cs)
{
imageResult.Draw(
new
CircleF(p, radius),
new
Bgr(255d, 0d, 0d), thickness);
cornerCount2
++
;
sbResult.AppendFormat(
"
{0}，
"
, p);
}
}
pbResult.Image
=
imageResult.Bitmap;
//
释放资源
imageResult.Dispose();
//
返回
return
string
.Format(
"
·ShiTomasi角点，用时{0:F05}毫秒，参数（最大角点数目：{1}，最小特征值：{2}，角点间的最小距离：{3}，邻域大小：{4}，角点类型：{5}，权重系数：{6}），检测到{7}个角点\r\n{8}
"
,
sw.Elapsed.TotalMilliseconds, cornerCount, qualityLevel, minDistance, blockSize, useHarris
?
"
Harris
"
:
"
ShiTomasi
"
, k, cornerCount2, cornerCount2
>
0
?
(sbResult.ToString()
+
"
\r\n
"
) :
""
);
}
亚像素级角点
在检测亚像素级角点前，需要提供角点的初始为止，这些初始位置可以用本文给出的其他的角点检测方式来获取，不过使用GoodFeaturesToTrack得到的结果最方便直接使用。
亚像素级角点检测的示例代码如下：
//
亚像素级角点
private
string
CornerSubPixFeatureDetect()
{
//
获取参数
int
winWidth
=
int
.Parse(txtCornerSubPixWinWidth.Text);
int
winHeight
=
int
.Parse(txtCornerSubPixWinHeight.Text);
Size win
=
new
Size(winWidth, winHeight);
int
zeroZoneWidth
=
int
.Parse(txtCornerSubPixZeroZoneWidth.Text);
int
zeroZoneHeight
=
int
.Parse(txtCornerSubPixZeroZoneHeight.Text);
Size zeroZone
=
new
Size(zeroZoneWidth, zeroZoneHeight);
int
maxIter
=
int
.Parse(txtCornerSubPixMaxIter.Text);
double
epsilon
=
double
.Parse(txtCornerSubPixEpsilon.Text);
MCvTermCriteria criteria
=
new
MCvTermCriteria(maxIter, epsilon);
//
先计算得到易于跟踪的点（ShiTomasi角点）
int
cornerCount
=
int
.Parse(txtGoodFeaturesCornerCount.Text);
double
qualityLevel
=
double
.Parse(txtGoodFeaturesQualityLevel.Text);
double
minDistance
=
double
.Parse(txtGoodFeaturesMinDistance.Text);
int
blockSize
=
int
.Parse(txtGoodFeaturesBlockSize.Text);
bool
useHarris
=
cbGoodFeaturesUseHarris.Checked;
double
k
=
double
.Parse(txtGoodFeaturesK.Text);
PointF[][] corners
=
imageSourceGrayscale.GoodFeaturesToTrack(cornerCount, qualityLevel, minDistance, blockSize, useHarris, k);
//
计算
Stopwatch sw
=
new
Stopwatch();
sw.Start();
imageSourceGrayscale.FindCornerSubPix(corners, win, zeroZone, criteria);
sw.Stop();
//
显示
Image
<
Bgr, Byte
>
imageResult
=
imageSourceGrayscale.Convert
<
Bgr, Byte
>
();
int
cornerCount2
=
0
;
StringBuilder sbResult
=
new
StringBuilder();
int
radius
=
(
int
)(minDistance
/
2
)
+
1
;
int
thickness
=
(
int
)(minDistance
/
4
)
+
1
;
foreach
(PointF[] cs
in
corners)
{
foreach
(PointF p
in
cs)
{
imageResult.Draw(
new
CircleF(p, radius),
new
Bgr(255d, 0d, 0d), thickness);
cornerCount2
++
;
sbResult.AppendFormat(
"
{0}，
"
, p);
}
}
pbResult.Image
=
imageResult.Bitmap;
//
释放资源
imageResult.Dispose();
//
返回
return
string
.Format(
"
·亚像素级角点，用时{0:F05}毫秒，参数（搜索窗口：{1}，死区：{2}，最大迭代次数：{3}，亚像素值的精度：{4}），检测到{5}个角点\r\n{6}
"
,
sw.Elapsed.TotalMilliseconds, win, zeroZone, maxIter, epsilon, cornerCount2, cornerCount2
>
0
?
(sbResult.ToString()
+
"
\r\n
"
) :
""
);
}
SURF角点
OpenCv中的cvExtractSURF函数和EmguCv中的Image<TColor,TDepth>.ExtractSURF方法用于检测SURF角点。
SURF角点检测的示例代码如下：
//
SURF角点
private
string
SurfFeatureDetect()
{
//
获取参数
bool
getDescriptors
=
cbSurfGetDescriptors.Checked;
MCvSURFParams surfParam
=
new
MCvSURFParams();
surfParam.extended
=
rbSurfBasicDescriptor.Checked
?
0
:
1
;
surfParam.hessianThreshold
=
double
.Parse(txtSurfHessianThreshold.Text);
surfParam.nOctaves
=
int
.Parse(txtSurfNumberOfOctaves.Text);
surfParam.nOctaveLayers
=
int
.Parse(txtSurfNumberOfOctaveLayers.Text);
//
计算
SURFFeature[] features
=
null
;
MKeyPoint[] keyPoints
=
null
;
Stopwatch sw
=
new
Stopwatch();
sw.Start();
if
(getDescriptors)
features
=
imageSourceGrayscale.ExtractSURF(
ref
surfParam);
else
keyPoints
=
surfParam.DetectKeyPoints(imageSourceGrayscale,
null
);
sw.Stop();
//
显示
bool
showDetail
=
cbSurfShowDetail.Checked;
Image
<
Bgr, Byte
>
imageResult
=
imageSourceGrayscale.Convert
<
Bgr, Byte
>
();
StringBuilder sbResult
=
new
StringBuilder();
int
idx
=
0
;
if
(getDescriptors)
{
foreach
(SURFFeature feature
in
features)
{
imageResult.Draw(
new
CircleF(feature.Point.pt,
5
),
new
Bgr(255d, 0d, 0d),
2
);
if
(showDetail)
{
sbResult.AppendFormat(
"
第{0}点（坐标：{1}，尺寸：{2}，方向：{3}°，hessian值：{4}，拉普拉斯标志：{5}，描述：[
"
,
idx, feature.Point.pt, feature.Point.size, feature.Point.dir, feature.Point.hessian, feature.Point.laplacian);
foreach
(
float
d
in
feature.Descriptor)
sbResult.AppendFormat(
"
{0},
"
, d);
sbResult.Append(
"
]），
"
);
}
idx
++
;
}
}
else
{
foreach
(MKeyPoint keypoint
in
keyPoints)
{
imageResult.Draw(
new
CircleF(keypoint.Point,
5
),
new
Bgr(255d, 0d, 0d),
2
);
if
(showDetail)
sbResult.AppendFormat(
"
第{0}点（坐标：{1}，尺寸：{2}，方向：{3}°，响应：{4}，octave：{5}），
"
,
idx, keypoint.Point, keypoint.Size, keypoint.Angle, keypoint.Response, keypoint.Octave);
idx
++
;
}
}
pbResult.Image
=
imageResult.Bitmap;
//
释放资源
imageResult.Dispose();
//
返回
return
string
.Format(
"
·SURF角点，用时{0:F05}毫秒，参数（描述：{1}，hessian阀值：{2}，octave数目：{3}，每个octave的层数：{4}，检测到{5}个角点\r\n{6}
"
,
sw.Elapsed.TotalMilliseconds, getDescriptors
?
(surfParam.extended
==
0
?
"
获取基本描述
"
:
"
获取扩展描述
"
) :
"
不获取描述
"
, surfParam.hessianThreshold,
surfParam.nOctaves, surfParam.nOctaveLayers, getDescriptors
?
features.Length : keyPoints.Length, showDetail
?
sbResult.ToString()
+
"
\r\n
"
:
""
);
}
Star关键点
OpenCv中的cvGetStarKeypoints函数和EmguCv中的Image<TColor,TDepth>.GetStarKeypoints方法用于检测“星型”附近的点。
Star关键点检测的示例代码如下：
//
Star关键点
private
string
StarKeyPointFeatureDetect()
{
//
获取参数
StarDetector starParam
=
new
StarDetector();
starParam.MaxSize
=
int
.Parse((
string
)cmbStarMaxSize.SelectedItem);
starParam.ResponseThreshold
=
int
.Parse(txtStarResponseThreshold.Text);
starParam.LineThresholdProjected
=
int
.Parse(txtStarLineThresholdProjected.Text);
starParam.LineThresholdBinarized
=
int
.Parse(txtStarLineThresholdBinarized.Text);
starParam.SuppressNonmaxSize
=
int
.Parse(txtStarSuppressNonmaxSize.Text);
//
计算
Stopwatch sw
=
new
Stopwatch();
sw.Start();
MCvStarKeypoint[] keyPoints
=
imageSourceGrayscale.GetStarKeypoints(
ref
starParam);
sw.Stop();
//
显示
Image
<
Bgr, Byte
>
imageResult
=
imageSourceGrayscale.Convert
<
Bgr, Byte
>
();
StringBuilder sbResult
=
new
StringBuilder();
int
idx
=
0
;
foreach
(MCvStarKeypoint keypoint
in
keyPoints)
{
imageResult.Draw(
new
CircleF(
new
PointF(keypoint.pt.X, keypoint.pt.Y), keypoint.size
/
2
),
new
Bgr(255d, 0d, 0d), keypoint.size
/
4
);
sbResult.AppendFormat(
"
第{0}点（坐标：{1}，尺寸：{2}，强度：{3}），
"
, idx, keypoint.pt, keypoint.size, keypoint.response);
idx
++
;
}
pbResult.Image
=
imageResult.Bitmap;
//
释放资源
imageResult.Dispose();
//
返回
return
string
.Format(
"
·Star关键点，用时{0:F05}毫秒，参数（MaxSize：{1}，ResponseThreshold：{2}，LineThresholdProjected：{3}，LineThresholdBinarized：{4}，SuppressNonmaxSize：{5}），检测到{6}个关键点\r\n{7}
"
,
sw.Elapsed.TotalMilliseconds, starParam.MaxSize, starParam.ResponseThreshold, starParam.LineThresholdProjected, starParam.LineThresholdBinarized, starParam.SuppressNonmaxSize, keyPoints.Length, keyPoints.Length
>
0
?
(sbResult.ToString()
+
"
\r\n
"
) :
""
);
}
FAST角点检测
FAST角点由E. Rosten教授提出，相比其他检测手段，这种方法的速度正如其名，相当的快。值得关注的是他所研究的理论都是属于实用类的，都很快。Rosten教授实现了FAST角点检测，并将其提供给了OpenCv，相当的有爱呀；不过OpenCv中的函数和Rosten教授的实现似乎有点点不太一样。遗憾的是，OpenCv中目前还没有FAST角点检测的文档。下面是我从Rosten的代码中找到的函数声明，可以看到粗略的参数说明。
/*
The references are:
* Machine learning for high-speed corner detection,
E. Rosten and T. Drummond, ECCV 2006
* Faster and better: A machine learning approach to corner detection
E. Rosten, R. Porter and T. Drummond, PAMI, 2009
*/
void cvCornerFast( const CvArr* image, int threshold, int N,
int nonmax_suppression, int* ret_number_of_corners,
CvPoint** ret_corners);
image: OpenCV image in which to detect corners. Must be 8 bit unsigned.
threshold: Threshold for detection (higher is fewer corners). 0--255
N: Arc length of detector, 9, 10, 11 or 12. 9 is usually best.
nonmax_suppression: Whether to perform nonmaximal suppression.
ret_number_of_corners: The number of detected corners is returned here.
ret_corners: The corners are returned here.
EmguCv中的Image<TColor,TDepth>.GetFASTKeypoints方法也实现了FAST角点检测，不过参数少了一些，只有threshold和nonmaxSupression，其中N我估计取的默认值9，但是返回的角点数目我不知道是怎么设置的。
使用FAST角点检测的示例代码如下：
//
FAST关键点
private
string
FASTKeyPointFeatureDetect()
{
//
获取参数
int
threshold
=
int
.Parse(txtFASTThreshold.Text);
bool
nonmaxSuppression
=
cbFASTNonmaxSuppression.Checked;
bool
showDetail
=
cbFASTShowDetail.Checked;
//
计算
Stopwatch sw
=
new
Stopwatch();
sw.Start();
MKeyPoint[] keyPoints
=
imageSourceGrayscale.GetFASTKeypoints(threshold, nonmaxSuppression);
sw.Stop();
//
显示
Image
<
Bgr, Byte
>
imageResult
=
imageSourceGrayscale.Convert
<
Bgr, Byte
>
();
StringBuilder sbResult
=
new
StringBuilder();
int
idx
=
0
;
foreach
(MKeyPoint keypoint
in
keyPoints)
{
imageResult.Draw(
new
CircleF(keypoint.Point, (
int
)(keypoint.Size
/
2
)),
new
Bgr(255d, 0d, 0d), (
int
)(keypoint.Size
/
4
));
if
(showDetail)
sbResult.AppendFormat(
"
第{0}点（坐标：{1}，尺寸：{2}，方向：{3}°，响应：{4}，octave：{5}），
"
,
idx, keypoint.Point, keypoint.Size, keypoint.Angle, keypoint.Response, keypoint.Octave);
idx
++
;
}
pbResult.Image
=
imageResult.Bitmap;
//
释放资源
imageResult.Dispose();
//
返回
return
string
.Format(
"
·FAST关键点，用时{0:F05}毫秒，参数（阀值：{1}，nonmaxSupression：{2}），检测到{3}个关键点\r\n{4}
"
,
sw.Elapsed.TotalMilliseconds, threshold, nonmaxSuppression, keyPoints.Length, showDetail
?
(sbResult.ToString()
+
"
\r\n
"
) :
""
);
}
Lepetit关键点
Lepetit关键点由Vincent Lepetit提出，可以在他的网站（http://cvlab.epfl.ch/~vlepetit/）上看到相关的论文等资料。EmguCv中的类LDetector实现了Lepetit关键点的检测。
使用Lepetit关键点检测的示例代码如下：
//
Lepetit关键点
private
string
LepetitKeyPointFeatureDetect()
{
//
获取参数
LDetector lepetitDetector
=
new
LDetector();
lepetitDetector.BaseFeatureSize
=
double
.Parse(txtLepetitBaseFeatureSize.Text);
lepetitDetector.ClusteringDistance
=
double
.Parse(txtLepetitClasteringDistance.Text);
lepetitDetector.NOctaves
=
int
.Parse(txtLepetitNumberOfOctaves.Text);
lepetitDetector.NViews
=
int
.Parse(txtLepetitNumberOfViews.Text);
lepetitDetector.Radius
=
int
.Parse(txtLepetitRadius.Text);
lepetitDetector.Threshold
=
int
.Parse(txtLepetitThreshold.Text);
lepetitDetector.Verbose
=
cbLepetitVerbose.Checked;
int
maxCount
=
int
.Parse(txtLepetitMaxCount.Text);
bool
scaleCoords
=
cbLepetitScaleCoords.Checked;
bool
showDetail
=
cbLepetitShowDetail.Checked;
//
计算
Stopwatch sw
=
new
Stopwatch();
sw.Start();
MKeyPoint[] keyPoints
=
lepetitDetector.DetectKeyPoints(imageSourceGrayscale, maxCount, scaleCoords);
sw.Stop();
//
显示
Image
<
Bgr, Byte
>
imageResult
=
imageSourceGrayscale.Convert
<
Bgr, Byte
>
();
StringBuilder sbResult
=
new
StringBuilder();
int
idx
=
0
;
foreach
(MKeyPoint keypoint
in
keyPoints)
{
//
imageResult.Draw(new CircleF(keypoint.Point, (int)(keypoint.Size / 2)), new Bgr(255d, 0d, 0d), (int)(keypoint.Size / 4));
imageResult.Draw(
new
CircleF(keypoint.Point,
4
),
new
Bgr(255d, 0d, 0d),
2
);
if
(showDetail)
sbResult.AppendFormat(
"
第{0}点（坐标：{1}，尺寸：{2}，方向：{3}°，响应：{4}，octave：{5}），
"
,
idx, keypoint.Point, keypoint.Size, keypoint.Angle, keypoint.Response, keypoint.Octave);
idx
++
;
}
pbResult.Image
=
imageResult.Bitmap;
//
释放资源
imageResult.Dispose();
//
返回
return
string
.Format(
"
·Lepetit关键点，用时{0:F05}毫秒，参数（基础特征尺寸：{1}，集群距离：{2}，阶数：{3}，视图数：{4}，半径：{5}，阀值：{6}，计算详细结果：{7}，最大关键点数目：{8}，缩放坐标：{9}），检测到{10}个关键点\r\n{11}
"
,
sw.Elapsed.TotalMilliseconds, lepetitDetector.BaseFeatureSize, lepetitDetector.ClusteringDistance, lepetitDetector.NOctaves, lepetitDetector.NViews,
lepetitDetector.Radius, lepetitDetector.Threshold, lepetitDetector.Verbose, maxCount, scaleCoords, keyPoints.Length, showDetail
?
(sbResult.ToString()
+
"
\r\n
"
) :
""
);
}
SIFT角点
SIFT角点是一种广泛使用的图像特征，可用于物体跟踪、图像匹配、图像拼接等领域，然而奇怪的是它并未被OpenCv实现。提出SIFT角点的David Lowe教授已经用C和matlab实现了SIFT角点的检测，并开放了源代码，不过他的实现不方便直接使用。您可以在http://www.cs.ubc.ca/~lowe/keypoints/看到SIFT的介绍、相关论文及David Lowe教授的实现代码。下面我要介绍由Andrea Vedaldi和Brian Fulkerson先生创建的vlfeat开源图像处理库，vlfeat库有C和matlab两种实现，其中包含了SIFT检测。您可以在http://www.vlfeat.org/下载到vlfeat库的代码、文档及可执行文件。
使用vlfeat检测SIFT角点需要以下步骤：
（1）用函数vl_sift_new()初始化SIFT过滤器对象，该过滤器对象可以反复用于多幅尺寸相同的图像；
（2）用函数vl_sift_first_octave()及vl_sift_process_next()遍历缩放空间的每一阶，直到返回VL_ERR_EOF为止；
（3）对于缩放空间的每一阶，用函数vl_sift_detect()来获取关键点；
（4）对每个关键点，用函数vl_sift_calc_keypoint_orientations()来获取该点的方向；
（5）对关键点的每个方向，用函数vl_sift_calc_keypoint_descriptor()来获取该方向的描述；
（6）使用完之后，用函数vl_sift_delete()来释放资源；
（7）如果要计算某个自定义关键点的描述，可以使用函数vl_sift_calc_raw_descriptor()。
直接使用vlfeat中的SIFT角点检测示例代码如下：
//
通过P/Invoke调用vlfeat函数来进行SIFT检测
unsafe
private
string
SiftFeatureDetectByPinvoke(
int
noctaves,
int
nlevels,
int
o_min,
bool
showDetail)
{
StringBuilder sbResult
=
new
StringBuilder();
//
初始化
IntPtr ptrSiftFilt
=
VlFeatInvoke.vl_sift_new(imageSource.Width, imageSource.Height, noctaves, nlevels, o_min);
if
(ptrSiftFilt
==
IntPtr.Zero)
return
"
Sift特征检测：初始化失败。
"
;
//
处理
Image
<
Gray, Single
>
imageSourceSingle
=
imageSourceGrayscale.ConvertScale
<
Single
>
(1d, 0d);
Image
<
Bgr, Byte
>
imageResult
=
imageSourceGrayscale.Convert
<
Bgr, Byte
>
();
int
pointCount
=
0
;
int
idx
=
0
;
//
依次遍历每一组
if
(VlFeatInvoke.vl_sift_process_first_octave(ptrSiftFilt, imageSourceSingle.MIplImage.imageData)
!=
VlFeatInvoke.VL_ERR_EOF)
{
while
(
true
)
{
//
计算每组中的关键点
VlFeatInvoke.vl_sift_detect(ptrSiftFilt);
//
遍历并绘制每个点
VlSiftFilt siftFilt
=
(VlSiftFilt)Marshal.PtrToStructure(ptrSiftFilt,
typeof
(VlSiftFilt));
pointCount
+=
siftFilt.nkeys;
VlSiftKeypoint
*
pKeyPoints
=
(VlSiftKeypoint
*
)siftFilt.keys.ToPointer();
for
(
int
i
=
0
; i
<
siftFilt.nkeys; i
++
)
{
VlSiftKeypoint keyPoint
=
*
pKeyPoints;
pKeyPoints
++
;
imageResult.Draw(
new
CircleF(
new
PointF(keyPoint.x, keyPoint.y), keyPoint.sigma
/
2
),
new
Bgr(255d, 0d, 0d),
2
);
if
(showDetail)
sbResult.AppendFormat(
"
第{0}点，坐标：({1},{2})，阶：{3}，缩放：{4}，s：{5}，
"
, idx, keyPoint.x, keyPoint.y, keyPoint.o, keyPoint.sigma, keyPoint.s);
idx
++
;
//
计算并遍历每个点的方向
double
[] angles
=
new
double
[
4
];
int
angleCount
=
VlFeatInvoke.vl_sift_calc_keypoint_orientations(ptrSiftFilt, angles,
ref
keyPoint);
if
(showDetail)
sbResult.AppendFormat(
"
共{0}个方向，
"
, angleCount);
for
(
int
j
=
0
; j
<
angleCount; j
++
)
{
double
angle
=
angles[j];
if
(showDetail)
sbResult.AppendFormat(
"
【方向：{0}，描述：
"
, angle);
//
计算每个方向的描述
IntPtr ptrDescriptors
=
Marshal.AllocHGlobal(
128
*
sizeof
(
float
));
VlFeatInvoke.vl_sift_calc_keypoint_descriptor(ptrSiftFilt, ptrDescriptors,
ref
keyPoint, angle);
float
*
pDescriptors
=
(
float
*
)ptrDescriptors.ToPointer();
for
(
int
k
=
0
; k
<
128
; k
++
)
{
float
descriptor
=
*
pDescriptors;
pDescriptors
++
;
if
(showDetail)
sbResult.AppendFormat(
"
{0},
"
, descriptor);
}
sbResult.Append(
"
】，
"
);
Marshal.FreeHGlobal(ptrDescriptors);
}
}
//
下一阶
if
(VlFeatInvoke.vl_sift_process_next_octave(ptrSiftFilt)
==
VlFeatInvoke.VL_ERR_EOF)
break
;
}
}
//
显示
pbResult.Image
=
imageResult.Bitmap;
//
释放资源
VlFeatInvoke.vl_sift_delete(ptrSiftFilt);
imageSourceSingle.Dispose();
imageResult.Dispose();
//
返回
return
string
.Format(
"
·SIFT特征检测(P/Invoke)，用时：未统计，参数（阶数：{0}，每阶层数：{1}，最小阶索引：{2}），{3}个关键点\r\n{4}
"
,
noctaves, nlevels, o_min, pointCount, showDetail
?
(sbResult.ToString()
+
"
\r\n
"
) :
""
);
}
要在.net中使用vlfeat还是不够方便，为此我对vlfeat中的SIFT角点检测部分进行了封装，将相关操作放到了类SiftDetector中。
使用SiftDetector需要两至三步：
（1）用构造函数初始化SiftDetector对象；
（2）用Process方法计算特征；
（3）视需要调用Dispose方法释放资源，或者等待垃圾回收器来自动释放资源。
使用SiftDetector的示例代码如下：
//
通过dotnet封装的SiftDetector类来进行SIFT检测
private
string
SiftFeatureDetectByDotNet(
int
noctaves,
int
nlevels,
int
o_min,
bool
showDetail)
{
//
初始化对象
SiftDetector siftDetector
=
new
SiftDetector(imageSource.Size, noctaves, nlevels, o_min);
//
计算
Image
<
Gray, Single
>
imageSourceSingle
=
imageSourceGrayscale.Convert
<
Gray, Single
>
();
Stopwatch sw
=
new
Stopwatch();
sw.Start();
List
<
SiftFeature
>
features
=
siftDetector.Process(imageSourceSingle, showDetail
?
SiftDetectorResultType.Extended : SiftDetectorResultType.Basic);
sw.Stop();
//
显示结果
Image
<
Bgr, Byte
>
imageResult
=
imageSourceGrayscale.Convert
<
Bgr, Byte
>
();
StringBuilder sbResult
=
new
StringBuilder();
int
idx
=
0
;
foreach
(SiftFeature feature
in
features)
{
imageResult.Draw(
new
CircleF(
new
PointF(feature.keypoint.x, feature.keypoint.y), feature.keypoint.sigma
/
2
),
new
Bgr(255d, 0d, 0d),
2
);
if
(showDetail)
{
sbResult.AppendFormat(
"
第{0}点，坐标：({1},{2})，阶：{3}，缩放：{4}，s：{5}，
"
,
idx, feature.keypoint.x, feature.keypoint.y, feature.keypoint.o, feature.keypoint.sigma, feature.keypoint.s);
sbResult.AppendFormat(
"
共{0}个方向，
"
, feature.keypointOrientations
!=
null
?
feature.keypointOrientations.Length :
0
);
if
(feature.keypointOrientations
!=
null
)
{
foreach
(SiftKeyPointOrientation orientation
in
feature.keypointOrientations)
{
if
(orientation.descriptors
!=
null
)
{
sbResult.AppendFormat(
"
【方向：{0}，描述：
"
, orientation.angle);
foreach
(
float
descriptor
in
orientation.descriptors)
sbResult.AppendFormat(
"
{0},
"
, descriptor);
}
else
sbResult.AppendFormat(
"
【方向：{0}，
"
, orientation.angle);
sbResult.Append(
"
】，
"
);
}
}
}
}
pbResult.Image
=
imageResult.Bitmap;
//
释放资源
siftDetector.Dispose();
imageSourceSingle.Dispose();
imageResult.Dispose();
//
返回
return
string
.Format(
"
·SIFT特征检测(.net)，用时：{0:F05}毫秒，参数（阶数：{1}，每阶层数：{2}，最小阶索引：{3}），{4}个关键点\r\n{5}
"
,
sw.Elapsed.TotalMilliseconds, noctaves, nlevels, o_min, features.Count, showDetail
?
(sbResult.ToString()
+
"
\r\n
"
) :
""
);
}
对vlfeat库中的SIFT部分封装代码如下所示：
定义SiftDetector类
MSER区域
OpenCv中的函数cvExtractMSER以及EmguCv中的Image<TColor,TDepth>.ExtractMSER方法实现了MSER区域的检测。由于OpenCv的文档中目前还没有cvExtractMSER这一部分，大家如果要看文档的话，可以先去看EmguCv的文档。
需要注意的是MSER区域的检测结果是区域中所有的点序列。例如检测到3个区域，其中一个区域是从(0,0)到(2,1)的矩形，那么结果点序列为：(0,0),(1,0),(2,0),(2,1),(1,1),(0,1)。
MSER区域检测的示例代码如下：
//
MSER（区域）特征检测
private
string
MserFeatureDetect()
{
//
获取参数
MCvMSERParams mserParam
=
new
MCvMSERParams();
mserParam.delta
=
int
.Parse(txtMserDelta.Text);
mserParam.maxArea
=
int
.Parse(txtMserMaxArea.Text);
mserParam.minArea
=
int
.Parse(txtMserMinArea.Text);
mserParam.maxVariation
=
float
.Parse(txtMserMaxVariation.Text);
mserParam.minDiversity
=
float
.Parse(txtMserMinDiversity.Text);
mserParam.maxEvolution
=
int
.Parse(txtMserMaxEvolution.Text);
mserParam.areaThreshold
=
double
.Parse(txtMserAreaThreshold.Text);
mserParam.minMargin
=
double
.Parse(txtMserMinMargin.Text);
mserParam.edgeBlurSize
=
int
.Parse(txtMserEdgeBlurSize.Text);
bool
showDetail
=
cbMserShowDetail.Checked;
//
计算
Stopwatch sw
=
new
Stopwatch();
sw.Start();
MemStorage storage
=
new
MemStorage();
Seq
<
Point
>
[] regions
=
imageSource.ExtractMSER(
null
,
ref
mserParam, storage);
sw.Stop();
//
显示
Image
<
Bgr, Byte
>
imageResult
=
imageSourceGrayscale.Convert
<
Bgr, Byte
>
();
StringBuilder sbResult
=
new
StringBuilder();
int
idx
=
0
;
foreach
(Seq
<
Point
>
region
in
regions)
{
imageResult.DrawPolyline(region.ToArray(),
true
,
new
Bgr(255d, 0d, 0d),
2
);
if
(showDetail)
{
sbResult.AppendFormat(
"
第{0}区域，包含{1}个顶点（
"
, idx, region.Total);
foreach
(Point pt
in
region)
sbResult.AppendFormat(
"
{0},
"
, pt);
sbResult.Append(
"
）\r\n
"
);
}
idx
++
;
}
pbResult.Image
=
imageResult.Bitmap;
//
释放资源
imageResult.Dispose();
storage.Dispose();
//
返回
return
string
.Format(
"
·MSER区域，用时{0:F05}毫秒，参数（delta：{1}，maxArea：{2}，minArea：{3}，maxVariation：{4}，minDiversity：{5}，maxEvolution：{6}，areaThreshold：{7}，minMargin：{8}，edgeBlurSize：{9}），检测到{10}个区域\r\n{11}
"
,
sw.Elapsed.TotalMilliseconds, mserParam.delta, mserParam.maxArea, mserParam.minArea, mserParam.maxVariation, mserParam.minDiversity,
mserParam.maxEvolution, mserParam.areaThreshold, mserParam.minMargin, mserParam.edgeBlurSize, regions.Length, showDetail
?
sbResult.ToString() :
""
);
}
各种特征检测方法性能对比
上面介绍了这么多的特征检测方法，那么它们的性能到底如何呢？因为它们的参数设置对处理时间及结果的影响很大，我们在这里基本都使用默认参数处理同一幅图像。在我机器上的处理结果见下表：
| 特征 | 用时（毫秒） | 特征数目 |
| Sobel算子 | 5.99420 | n/a |
| 拉普拉斯算子 | 3.13440 | n/a |
| Canny算子 | 3.41160 | n/a |
| 霍夫线变换 | 13.70790 | 10 |
| 霍夫圆变换 | 78.07720 | 0 |
| Harris角点 | 9.41750 | n/a |
| ShiTomasi角点 | 16.98390 | 18 |
| 亚像素级角点 | 3.63360 | 18 |
| SURF角点 | 266.27000 | 151 |
| Star关键点 | 14.82800 | 56 |
| FAST角点 | 31.29670 | 159 |
| SIFT角点 | 287.52310 | 54 |
| MSER区域 | 40.62970 | 2 |
（图片尺寸：583x301，处理器：AMD ATHLON IIx2 240，内存：DDR3 4G，显卡：GeForce 9500GT，操作系统：Windows 7）
感谢您耐心看完本文，希望对您有所帮助。
下一篇文章我们将一起看看如何来跟踪本文讲到的特征点（角点）。
另外，如果需要本文的源代码，请点击这里下载。
1371
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
