매칭률 구하기

매칭률 구하기를 하기 위한 무작정 자료수집

이게 관련 함수인가 봄

CV_TM_SQDIFF
CV_TM_SQDIFF_NORMED

 

CV_TM_CCOEFF
CV_TM_CCOEFF_NORMED

CV_TM_CCORR

CV_TM_CCORR_NORMED

opencv제대로 배우기 p.298 - p.303 템플릿 매칭

템플릿 매칭(Template Matching) 

= 큰 이미지에서 특정 작은 이미지를 찾는 알고리즘

(http://www.obg.co.kr/?mid=programming&document_srl=8840&listStyle=viewer&page=8)

cvMatchTemplate()

입력 영상을 주어진 영상 패치로 스캔하면서 매칭을 수행

void cvMatchTemplate(

const    CvArr*    image,

const    CvArr*    templ,

CvArr*                result,

int                       method

);

하나의 8비트 또는 실수형 영상, 또는 트루컬러 영상을 입력으로 사용.

제곱차 매칭 방법(method = CV_TM_SQDIFF)

픽셀값을 제곱차를 이용.

완벽하게 일치하면 0을 반환, 일치하지 않으면 값이 커진다.

상관관계 매칭 방법(method = CV_TM_CCORR)

템플릿과 입력 영상의 곱을 제곱하여 모두 더한다. 

완벽하게 일치하면 값이 크게 나오고, 일치하지 않으면 작은 값이 나오거나 0이 나온다.

상관관계 매칭 방법(method = CV_TM_CCOEFF)

템플렛과 입력 영상 각각의 평균을 고려한 매칭을 수행

완벽하게 일치하면 1을 반환하고, 완전히 불일치하면 -1을 반환

이 값이 0을 반환하는 경우는 두 영상 사이에 전혀 연관성이 없음을 의미

http://www.larmor.com/projects/JavaOpenCVMatchTemplate/doc/com/larmor/opencv/MatchTemplate.html

(CV_TM_SQDIFF) : 제곱의 차이 

(CV_TM_SQDIFF_NORMED) : 정규화 제곱 차이 

(CV_TM_CCORR) : 교차 상관 

(CV_TM_CCORR_NORMED) : 정규화 된 상호 상관 관계 

(CV_TM_CCOEFF) : 상관 계수 

(CV_TM_CCOEFF_NORMED) : 정규화 상관 계수 (NCC - 빠른 정상화 간 상관 관계)

---

---

com.larmor.opencv
Class MatchTemplate

java.lang.Object
  com.larmor.opencv.MatchTemplate

---

public class MatchTemplate
extends Object

This Class wraps OpenCV (the Open Source Computer Vision Library http://www.intel.com/technology/computing/opencv/index.htm) function cvMatchTemplate and implements other methods for utilities result visualization.

The class uses JAI (Java Advanced Imaging) API and JAI-Java Image I/O Technology (for dowload: https://jai.dev.java.net/ and https://jai-imageio.dev.java.net/).

The OpenCV function cvMatchTemplate compares overlapped image of size w�h with a template using the specified method and stores the comparison results to result image. The possible methods for matching are:

• (CV_TM_SQDIFF): squared difference
• (CV_TM_SQDIFF_NORMED): normalized squared difference
• (CV_TM_CCORR): cross correlation
• (CV_TM_CCORR_NORMED): normalized cross correlation
• (CV_TM_CCOEFF): correlation coefficient
• (CV_TM_CCOEFF_NORMED): normalized correlation coefficient (NCC - Fast Normalized Cross-Correlation)
  see http://www.idiom.com/~zilla/Papers/nvisionInterface/nip.html
  and http://www.cs.ubc.ca/~deaton/remarks_ncc.html

Here are the formulae for the different comparison methods one may use (I denotes image, T - template, R - result. The summation is done over template and/or the image patch: x'=0..w-1, y'=0..h-1):

 method=CV_TM_SQDIFF:
 R(x,y)=sumx',y'[T(x',y')-I(x+x',y+y')]2
 
 method=CV_TM_SQDIFF_NORMED:
 R(x,y)=sumx',y'[T(x',y')-I(x+x',y+y')]2/sqrt[sumx',y'T(x',y')2븉umx',y'I(x+x',y+y')2]
 
 method=CV_TM_CCORR:
 R(x,y)=sumx',y'[T(x',y')뷞(x+x',y+y')]
 
 method=CV_TM_CCORR_NORMED:
 R(x,y)=sumx',y'[T(x',y')뷞(x+x',y+y')]/sqrt[sumx',y'T(x',y')2븉umx',y'I(x+x',y+y')2]
 
 method=CV_TM_CCOEFF:
 R(x,y)=sumx',y'[T'(x',y')뷞'(x+x',y+y')],
 where T'(x',y')=T(x',y') - 1/(w뷻)븉umx",y"T(x",y")
 I'(x+x',y+y')=I(x+x',y+y') - 1/(w뷻)븉umx",y"I(x+x",y+y")
 
 method=CV_TM_CCOEFF_NORMED:
 R(x,y)=sumx',y'[T'(x',y')뷞'(x+x',y+y')]/sqrt[sumx',y'T'(x',y')2븉umx',y'I'(x+x',y+y')2]
 
 

Version:

Vers. 0.5 beta (March.2008)

Author:

Pier Paolo Ciarravano Larmor License: GNU General Public License

---

Field Summary
static int	CV_TM_CCOEFF           CV_TM_CCOEFF: correlation coefficient
static int	CV_TM_CCOEFF_NORMED           CV_TM_CCOEFF_NORMED: normalized correlation coefficient (NCC - Fast Normalized Cross-Correlation) see http://www.idiom.com/~zilla/Papers/nvisionInterface/nip.html and http://www.cs.ubc.ca/~deaton/remarks_ncc.html
static int	CV_TM_CCORR           CV_TM_CCORR: cross correlation
static int	CV_TM_CCORR_NORMED           CV_TM_CCORR_NORMED: normalized cross correlation
static int	CV_TM_SQDIFF           CV_TM_SQDIFF: squared difference
static int	CV_TM_SQDIFF_NORMED           CV_TM_SQDIFF_NORMED: normalized squared difference

 

Constructor Summary
MatchTemplate(BufferedImage source, BufferedImage template)           Init source and template image for matchTemplate... methods, these images using java.awt.color.ColorSpace.CS_GRAY.

 

Method Summary
static void	displayBufferedImage(BufferedImage image, String title)           Utility method for display a BufferedImage.
static void	displayBufferedImageCrossPoint(BufferedImage image, String title, Point center, int x, int y)           Utility method for display a BufferedImage and draw cross point on it.
float[]	getCvMatchTemplateFloat()           You can call this method only after matchTemplate(...) methods
static BufferedImage	getGrayBufferedImage(RenderedImage image)           Use this method to costruct image and template source images for costructor with java.awt.color.ColorSpace.CS_GRAY.
static BufferedImage	getImageGrayFromImage(BufferedImage image, int x, int y, int w, int h)           Use this method to costruct TYPE_BYTE_GRAY image rectangle of w x h pixels (starting from x and y of original image) from input image.
BufferedImage	getMatchImage()           You can call this method only after matchTemplate(...) methods
int	getResultHeight()           You can call this method only after matchTemplate(...) methods
int	getResultWidth()           You can call this method only after matchTemplate(...) methods
Point	matchTemplate(int method)           This method match entire image with template using the specified method and return point of best matches.
Point	matchTemplate(int x, int y, int w, int h, int method)           This method match rectangle image of w x h pixels (starting from x and y of original image) with template using the specified method and return point of best matches.
Point	matchTemplateBestPoint(int method)           This method matchs entire image with template using the specified method and return point of best matches.
Point	matchTemplateBestPoint(int x, int y, int w, int h, int method)           This method match rectangle image of w x h pixels (starting from x and y of original image) with template using the specified method and return point of best matches.

 

Methods inherited from class java.lang.Object
equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

 

Field Detail

CV_TM_CCOEFF

public static final int CV_TM_CCOEFF

CV_TM_CCOEFF: correlation coefficient

See Also:

Constant Field Values

---

CV_TM_CCOEFF_NORMED

public static final int CV_TM_CCOEFF_NORMED

CV_TM_CCOEFF_NORMED: normalized correlation coefficient (NCC - Fast Normalized Cross-Correlation)
see http://www.idiom.com/~zilla/Papers/nvisionInterface/nip.html
and http://www.cs.ubc.ca/~deaton/remarks_ncc.html

See Also:

Constant Field Values

---

CV_TM_CCORR

public static final int CV_TM_CCORR

CV_TM_CCORR: cross correlation

See Also:

Constant Field Values

---

CV_TM_CCORR_NORMED

public static final int CV_TM_CCORR_NORMED

CV_TM_CCORR_NORMED: normalized cross correlation

See Also:

Constant Field Values

---

CV_TM_SQDIFF

public static final int CV_TM_SQDIFF

CV_TM_SQDIFF: squared difference

See Also:

Constant Field Values

---

CV_TM_SQDIFF_NORMED

public static final int CV_TM_SQDIFF_NORMED

CV_TM_SQDIFF_NORMED: normalized squared difference

See Also:

Constant Field Values

Constructor Detail

MatchTemplate

public MatchTemplate(BufferedImage source,
                     BufferedImage template)

Init source and template image for matchTemplate... methods, these images using java.awt.color.ColorSpace.CS_GRAY.
For costruct these images use:

        java.awt.image.RenderedImage image = javax.imageio.ImageIO.read(new File(source));
        java.awt.image.BufferedImage imageBI = MatchTemplate.getGrayBufferedImage(sourceImg);
 

Parameters:

source - source BufferedImage using java.awt.color.ColorSpace.CS_GRAY

template - template BufferedImage using java.awt.color.ColorSpace.CS_GRAY

Method Detail

displayBufferedImage

public static void displayBufferedImage(BufferedImage image,
                                        String title)

Utility method for display a BufferedImage.

Parameters:

image - input image

title - window title string

---

displayBufferedImageCrossPoint

public static void displayBufferedImageCrossPoint(BufferedImage image,
                                                  String title,
                                                  Point center,
                                                  int x,
                                                  int y)

Utility method for display a BufferedImage and draw cross point on it.

Parameters:

image - input image

title - window title string

center - relative center point returned from matchTemplate... methods

x - the x coordinates of the pixel used in matchTemplate... methods

y - the y coordinates of the pixel used in matchTemplate... methods

---

getCvMatchTemplateFloat

public float[] getCvMatchTemplateFloat()

You can call this method only after matchTemplate(...) methods

Returns:

result pixels matrix image

---

getGrayBufferedImage

public static BufferedImage getGrayBufferedImage(RenderedImage image)

Use this method to costruct image and template source images for costructor with java.awt.color.ColorSpace.CS_GRAY.
Use:

        java.awt.image.RenderedImage image = javax.imageio.ImageIO.read(new File(source));
        java.awt.image.BufferedImage imageBI = MatchTemplate.getGrayBufferedImage(sourceImg);
 

Parameters:

image - input java.awt.image.RenderedImage

Returns:

java.awt.color.ColorSpace.CS_GRAY image

---

getImageGrayFromImage

public static BufferedImage getImageGrayFromImage(BufferedImage image,
                                                  int x,
                                                  int y,
                                                  int w,
                                                  int h)

Use this method to costruct TYPE_BYTE_GRAY image rectangle of w x h pixels (starting from x and y of original image) from input image.

Parameters:

image - BufferedImage

x - the x coordinates of the pixel location

y - the y coordinates of the pixel location

w - Width of the pixel rectangle

h - Height of the pixel rectangle

Returns:

BufferedImage image

---

getMatchImage

public BufferedImage getMatchImage()

You can call this method only after matchTemplate(...) methods

Returns:

TYPE_BYTE_GRAY BufferedImage of result pixels matrix image

---

getResultHeight

public int getResultHeight()

You can call this method only after matchTemplate(...) methods

Returns:

result image Height

---

getResultWidth

public int getResultWidth()

You can call this method only after matchTemplate(...) methods

Returns:

result image Width

---

matchTemplate

public Point matchTemplate(int method)

This method match entire image with template using the specified method and return point of best matches. It Uses java implementation to find minimum and maximum element for best matches.
Using this method, you can call getCvMatchTemplateFloat(),getResultWidth() and getResultHeight() methods for return result float[] matrix and image dimension.

Parameters:

method - the comparison methods

Returns:

java.awt.Point of best matches

---

matchTemplate

public Point matchTemplate(int x,
                           int y,
                           int w,
                           int h,
                           int method)

This method match rectangle image of w x h pixels (starting from x and y of original image) with template using the specified method and return point of best matches. It Uses java implementation to find minimum and maximum element for best matches.
Using this method, you can call getCvMatchTemplateFloat(),getResultWidth() and getResultHeight() methods for return result float[] matrix and image dimension.

Parameters:

x - the x coordinates of the pixel location

y - the y coordinates of the pixel location

w - Width of the pixel rectangle

h - Height of the pixel rectangle

method - the comparison methods

Returns:

java.awt.Point of best matches

---

matchTemplateBestPoint

public Point matchTemplateBestPoint(int method)

This method matchs entire image with template using the specified method and return point of best matches. It Uses OpenCV internal function cvMinMaxLoc for find minimum and maximum element for best matches.

Parameters:

method - the comparison methods

Returns:

java.awt.Point of best matches

---

matchTemplateBestPoint

public Point matchTemplateBestPoint(int x,
                                    int y,
                                    int w,
                                    int h,
                                    int method)

This method match rectangle image of w x h pixels (starting from x and y of original image) with template using the specified method and return point of best matches. It Uses OpenCV internal function cvMinMaxLoc for find minimum and maximum element for best matches.

Parameters:

x - the x coordinates of the pixel location

y - the y coordinates of the pixel location

w - Width of the pixel rectangle

h - Height of the pixel rectangle

method - the comparison methods

Returns:

java.awt.Point of best matches

---

---

매칭률에 대한 질문과 답변 url

http://cafe.naver.com/opencv/26797

http://cafe.naver.com/opencv/26730

http://cafe.naver.com/opencv/26813

cvMatchTemplate

http://aidenjlee.tistory.com/116

cvMatchTemplate :)

OpenCV 학습, 

void CFollowDlg : : MatchTemplate (CString Path) { 

/ / = 0 더블.; 

IplImage * imgSrc = cvLoadImage (Path); 

IplImage * imgTemp = cvLoadImage ( "C  : \ \ t.bmp"); 

CvSize sizeSrc = cvGetSize (imgSrc); 

CvSize의 sizeTemp = cvGetSize (imgTemp); 

CvSize의 sizeResult = cvSize (sizeSrc.width - sizeTemp.width 1, sizeSrc.height - sizeTemp.height 1); 

IplImage * imgResult = cvCreateImage (sizeResult, IPL_DEPTH_32F, 1); 

(imgSrc, imgTemp, imgResult, CV_TM_CCORR_NORMED) cvMatchTemplate; 

/ / = cvMatchShapes (imgSrc, imgTemp, CV_CONTOURS_MATCH_I3); 

dMax = 0 .; 

CvPoint Point = cvPoint (0,0);

/ /if(!a) MessageBox ( "성공하지 못함!", MB_OK) ;

for (int cx=0 ; cx

{ 

for (int cy=0 ; cy

{ 

float fTemp = CV_IMAGE_ELEM(imgResult,float,cy,cx); 

if (dMax < fTemp) 

{ 

dMax = fTemp; 

point = cvPoint(cx,cy); 

} 

} 

} 

CvPoint point2 = cvPoint(point.x+sizeTemp.width,point.y+sizeTemp.height); 

cvRectangle(imgSrc,point,point2,cvScalar(255)); 

cvNamedWindow( "Test", CV_WINDOW_AUTOSIZE ); 

cvShowImage("Test",imgSrc); 

/*   for(;;) 

{ 

int k = cvWaitKey(5); 

if( k == 27 ) break; 

} 

*/

}

void CFollowDlg::OnButton4() 

{ 

// TODO: Add your control notification handler code here 

CString path; 

CFileDialog dlg(TRUE); 

if(dlg.DoModal()==IDOK) 

path=dlg.GetPathName(); 

MatchTemplate(path); 

}

소개 cvMatchTemplate () 설명서 :

프로토 타입 :

void cvMatchTemplate( const CvArr* image, const CvArr* templ,

                      CvArr* result, int method );

 

이미지

이미지를 검색할 수 있습니다. 비트 이미지를 포인트 또는 32 - 비트 부동 단일 그것은이어야 채널, 8 -

templ

검색 서식 파일, 이미지가 없습니다 입력보다 큰 이미지 및 입력 유형을 가지고있는 동일한 데이터

결과

이미지 비교 매핑 결과를 중. 단일 채널, 32 - 이미지는 승 비트 H는 부동 소수점 ×. 않으면과 templ 반장님입니다 w × 후, 결과) 1 +1) ×을 (H 조 - H를해야합니다 (- 승 w

방법

를 지정 검색 방법 : 기능 cvMatchTemplate과 함수 cvCalcBackProjectPatch 비슷한. 예금에 비해 그 결과 것입니다 결과가 슬라이드를 통해 전체 이미지 이미지와, 비교 메서드는 지정된와 templ과 영역을 오버랩의 × H 조 w 이미지 크기.

함수 또는 *) CV_TM_SQDIFF cvMinMaxLoc의 이용을 통하여 최소 글로벌 찾아 마무리 비교를 위해) 최대 (CV_TM_CCORR * 및 CV_TM_CCOEFF의 *.