基於GraphCuts圖割算法的圖像分割----OpenCV代碼與實現
轉載請註明出處：http://blog.csdn.net/wangyaninglm/article/details/44151213，node
來自：shiter編寫程序的藝術
ios
1.緒論
圖切割算法是組合圖論的經典算法之一。近年來，許多學者將其應用到圖像和視頻分割中，取得了很好的效果。本文簡單介紹了圖切算法和交互式圖像分割技術，以及圖切算法在交互式圖像分割中的應用。git
圖像分割指圖像分紅各具特性的區域並提取出感興趣目標的技術和過程，它是由圖像處理到圖像分析的關鍵步驟，是一種基本的計算機視覺技術。只有在圖像分割的基礎上才能對目標進行特徵提取和參數測量，使得更高層的圖像分析和理解成爲可能。所以對圖像分割方法的研究具備十分重要的意義。算法
圖像分割技術的研究已有幾十年的歷史，但至今人們並不能找到通用的方法可以適合於全部類型的圖像。經常使用的圖像分割技術可劃分爲四類：特徵閾值或聚類、邊緣檢測、區域生長或區域提取。雖然這些方法分割灰度圖像效果較好，但用於彩色圖像的分割每每達不到理想的效果。windows
交互式圖像分割是指，首先由用戶以某種交互手段指定圖像的部分前景與部分背景，而後算法以用戶的輸入做爲分割的約束條件自動地計算出知足約束條件下的最佳分割。典型的交互手段包括用一把畫刷在前景和背景處各畫幾筆（如[1][4]等）以及在前景的周圍畫一個方框（如[2]）等。app
基於圖切算法的圖像分割技術是近年來國際上圖像分割領域的一個新的研究熱點。該類方法將圖像映射爲賦權無向圖，把像素視做節點，利用最小切割獲得圖像的最佳分割。ide
2.幾種改進算法
- Graph Cut[1]算法是一種直接基於圖切算法的圖像分割技術。它僅須要在前景和背景處各畫幾筆做爲輸入，算法將創建各個像素點與前景背景類似度的賦權圖，並經過求解最小切割區分前景和背景。
- Grabcut[2]算法方法的用戶交互量不多，僅僅須要指定一個包含前景的矩形，隨後用基於圖切算法在圖像中提取前景。
- Lazy Snapping[4]系統則是對[1]的改進。經過預計算和聚類技術，該方法提供了一個即時反饋的平臺，方便用戶進行交互分割。
文檔說明：函數
http://download.csdn.net/detail/wangyaninglm/8484301ui
3.代碼實現效果
graphcuts代碼：this
http://download.csdn.net/detail/wangyaninglm/8484243
ICCV'2001論文"Interactive graph cuts for optimal boundary and region segmentation of objects in N-D images"。
Graph Cut方法是基於顏色統計採樣的方法，所以對前背景相差較大的圖像效果較佳。
同時，比例係數lambda的調節直接影響到最終的分割效果。
grabcut代碼：
// Grabcut.cpp : 定義控制檯應用程序的入口點。
//
#include "stdafx.h"
#include "opencv2/highgui/highgui.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include <iostream>
#include "ComputeTime.h"
#include "windows.h"
using namespace std;
using namespace cv;
static void help()
{
cout << "\nThis program demonstrates GrabCut segmentation -- select an object in a region\n"
"and then grabcut will attempt to segment it out.\n"
"Call:\n"
"./grabcut <image_name>\n"
"\nSelect a rectangular area around the object you want to segment\n" <<
"\nHot keys: \n"
"\tESC - quit the program\n"
"\tr - restore the original image\n"
"\tn - next iteration\n"
"\n"
"\tleft mouse button - set rectangle\n"
"\n"
"\tCTRL+left mouse button - set GC_BGD pixels\n"
"\tSHIFT+left mouse button - set CG_FGD pixels\n"
"\n"
"\tCTRL+right mouse button - set GC_PR_BGD pixels\n"
"\tSHIFT+right mouse button - set CG_PR_FGD pixels\n" << endl;
}
const Scalar RED = Scalar(0,0,255);
const Scalar PINK = Scalar(230,130,255);
const Scalar BLUE = Scalar(255,0,0);
const Scalar LIGHTBLUE = Scalar(255,255,160);
const Scalar GREEN = Scalar(0,255,0);
const int BGD_KEY = CV_EVENT_FLAG_CTRLKEY; //Ctrl鍵
const int FGD_KEY = CV_EVENT_FLAG_SHIFTKEY; //Shift鍵
static void getBinMask( const Mat& comMask, Mat& binMask )
{
if( comMask.empty() || comMask.type()!=CV_8UC1 )
CV_Error( CV_StsBadArg, "comMask is empty or has incorrect type (not CV_8UC1)" );
if( binMask.empty() || binMask.rows!=comMask.rows || binMask.cols!=comMask.cols )
binMask.create( comMask.size(), CV_8UC1 );
binMask = comMask & 1; //獲得mask的最低位,其實是隻保留肯定的或者有可能的前景點當作mask
}
class GCApplication
{
public:
enum{ NOT_SET = 0, IN_PROCESS = 1, SET = 2 };
static const int radius = 2;
static const int thickness = -1;
void reset();
void setImageAndWinName( const Mat& _image, const string& _winName );
void showImage() const;
void mouseClick( int event, int x, int y, int flags, void* param );
int nextIter();
int getIterCount() const { return iterCount; }
private:
void setRectInMask();
void setLblsInMask( int flags, Point p, bool isPr );
const string* winName;
const Mat* image;
Mat mask;
Mat bgdModel, fgdModel;
uchar rectState, lblsState, prLblsState;
bool isInitialized;
Rect rect;
vector<Point> fgdPxls, bgdPxls, prFgdPxls, prBgdPxls;
int iterCount;
};
/*給類的變量賦值*/
void GCApplication::reset()
{
if( !mask.empty() )
mask.setTo(Scalar::all(GC_BGD));
bgdPxls.clear(); fgdPxls.clear();
prBgdPxls.clear(); prFgdPxls.clear();
isInitialized = false;
rectState = NOT_SET; //NOT_SET == 0
lblsState = NOT_SET;
prLblsState = NOT_SET;
iterCount = 0;
}
/*給類的成員變量賦值而已*/
void GCApplication::setImageAndWinName( const Mat& _image, const string& _winName )
{
if( _image.empty() || _winName.empty() )
return;
image = &_image;
winName = &_winName;
mask.create( image->size(), CV_8UC1);
reset();
}
/*顯示4個點，一個矩形和圖像內容，由於後面的步驟不少地方都要用到這個函數，因此單獨拿出來*/
void GCApplication::showImage() const
{
if( image->empty() || winName->empty() )
return;
Mat res;
Mat binMask;
if( !isInitialized )
image->copyTo( res );
else
{
getBinMask( mask, binMask );
image->copyTo( res, binMask ); //按照最低位是0仍是1來複制，只保留跟前景有關的圖像，好比說可能的前景，可能的背景
}
vector<Point>::const_iterator it;
/*下面4句代碼是將選中的4個點用不一樣的顏色顯示出來*/
for( it = bgdPxls.begin(); it != bgdPxls.end(); ++it ) //迭代器能夠當作是一個指針
circle( res, *it, radius, BLUE, thickness );
for( it = fgdPxls.begin(); it != fgdPxls.end(); ++it ) //肯定的前景用紅色表示
circle( res, *it, radius, RED, thickness );
for( it = prBgdPxls.begin(); it != prBgdPxls.end(); ++it )
circle( res, *it, radius, LIGHTBLUE, thickness );
for( it = prFgdPxls.begin(); it != prFgdPxls.end(); ++it )
circle( res, *it, radius, PINK, thickness );
/*畫矩形*/
if( rectState == IN_PROCESS || rectState == SET )
rectangle( res, Point( rect.x, rect.y ), Point(rect.x + rect.width, rect.y + rect.height ), GREEN, 2);
imshow( *winName, res );
}
/*該步驟完成後，mask圖像中rect內部是3，外面全是0*/
void GCApplication::setRectInMask()
{
assert( !mask.empty() );
mask.setTo( GC_BGD ); //GC_BGD == 0
rect.x = max(0, rect.x);
rect.y = max(0, rect.y);
rect.width = min(rect.width, image->cols-rect.x);
rect.height = min(rect.height, image->rows-rect.y);
(mask(rect)).setTo( Scalar(GC_PR_FGD) ); //GC_PR_FGD == 3，矩形內部,爲可能的前景點
}
void GCApplication::setLblsInMask( int flags, Point p, bool isPr )
{
vector<Point> *bpxls, *fpxls;
uchar bvalue, fvalue;
if( !isPr ) //肯定的點
{
bpxls = &bgdPxls;
fpxls = &fgdPxls;
bvalue = GC_BGD; //0
fvalue = GC_FGD; //1
}
else //機率點
{
bpxls = &prBgdPxls;
fpxls = &prFgdPxls;
bvalue = GC_PR_BGD; //2
fvalue = GC_PR_FGD; //3
}
if( flags & BGD_KEY )
{
bpxls->push_back(p);
circle( mask, p, radius, bvalue, thickness ); //該點處爲2
}
if( flags & FGD_KEY )
{
fpxls->push_back(p);
circle( mask, p, radius, fvalue, thickness ); //該點處爲3
}
}
/*鼠標響應函數，參數flags爲CV_EVENT_FLAG的組合*/
void GCApplication::mouseClick( int event, int x, int y, int flags, void* )
{
// TODO add bad args check
switch( event )
{
case CV_EVENT_LBUTTONDOWN: // set rect or GC_BGD(GC_FGD) labels
{
bool isb = (flags & BGD_KEY) != 0,
isf = (flags & FGD_KEY) != 0;
if( rectState == NOT_SET && !isb && !isf )//只有左鍵按下時
{
rectState = IN_PROCESS; //表示正在畫矩形
rect = Rect( x, y, 1, 1 );
}
if ( (isb || isf) && rectState == SET ) //按下了alt鍵或者shift鍵，且畫好了矩形，表示正在畫前景背景點
lblsState = IN_PROCESS;
}
break;
case CV_EVENT_RBUTTONDOWN: // set GC_PR_BGD(GC_PR_FGD) labels
{
bool isb = (flags & BGD_KEY) != 0,
isf = (flags & FGD_KEY) != 0;
if ( (isb || isf) && rectState == SET ) //正在畫可能的前景背景點
prLblsState = IN_PROCESS;
}
break;
case CV_EVENT_LBUTTONUP:
if( rectState == IN_PROCESS )
{
rect = Rect( Point(rect.x, rect.y), Point(x,y) ); //矩形結束
rectState = SET;
setRectInMask();
assert( bgdPxls.empty() && fgdPxls.empty() && prBgdPxls.empty() && prFgdPxls.empty() );
showImage();
}
if( lblsState == IN_PROCESS ) //已畫了先後景點
{
setLblsInMask(flags, Point(x,y), false); //畫出前景點
lblsState = SET;
showImage();
}
break;
case CV_EVENT_RBUTTONUP:
if( prLblsState == IN_PROCESS )
{
setLblsInMask(flags, Point(x,y), true); //畫出背景點
prLblsState = SET;
showImage();
}
break;
case CV_EVENT_MOUSEMOVE:
if( rectState == IN_PROCESS )
{
rect = Rect( Point(rect.x, rect.y), Point(x,y) );
assert( bgdPxls.empty() && fgdPxls.empty() && prBgdPxls.empty() && prFgdPxls.empty() );
showImage(); //不斷的顯示圖片
}
else if( lblsState == IN_PROCESS )
{
setLblsInMask(flags, Point(x,y), false);
showImage();
}
else if( prLblsState == IN_PROCESS )
{
setLblsInMask(flags, Point(x,y), true);
showImage();
}
break;
}
}
/*該函數進行grabcut算法，而且返回算法運行迭代的次數*/
int GCApplication::nextIter()
{
if( isInitialized )
//使用grab算法進行一次迭代，參數2爲mask，裏面存的mask位是：矩形內部除掉那些多是背景或者已經肯定是背景後的全部的點，且mask同時也爲輸出
//保存的是分割後的前景圖像
grabCut( *image, mask, rect, bgdModel, fgdModel, 1 );
else
{
if( rectState != SET )
return iterCount;
if( lblsState == SET || prLblsState == SET )
grabCut( *image, mask, rect, bgdModel, fgdModel, 1, GC_INIT_WITH_MASK );
else
grabCut( *image, mask, rect, bgdModel, fgdModel, 1, GC_INIT_WITH_RECT );
isInitialized = true;
}
iterCount++;
bgdPxls.clear(); fgdPxls.clear();
prBgdPxls.clear(); prFgdPxls.clear();
return iterCount;
}
GCApplication gcapp;
static void on_mouse( int event, int x, int y, int flags, void* param )
{
gcapp.mouseClick( event, x, y, flags, param );
}
int main( int argc, char** argv )
{
string filename;
cout<<" Grabcuts ! \n";
cout<<"input image name: "<<endl;
cin>>filename;
Mat image = imread( filename, 1 );
if( image.empty() )
{
cout << "\n Durn, couldn't read image filename " << filename << endl;
return 1;
}
help();
const string winName = "image";
cvNamedWindow( winName.c_str(), CV_WINDOW_AUTOSIZE );
cvSetMouseCallback( winName.c_str(), on_mouse, 0 );
gcapp.setImageAndWinName( image, winName );
gcapp.showImage();
for(;;)
{
int c = cvWaitKey(0);
switch( (char) c )
{
case '\x1b':
cout << "Exiting ..." << endl;
goto exit_main;
case 'r':
cout << endl;
gcapp.reset();
gcapp.showImage();
break;
case 'n':
ComputeTime ct ;
ct.Begin();
int iterCount = gcapp.getIterCount();
cout << "<" << iterCount << "... ";
int newIterCount = gcapp.nextIter();
if( newIterCount > iterCount )
{
gcapp.showImage();
cout << iterCount << ">" << endl;
cout<<"運行時間: "<<ct.End()<<endl;
}
else
cout << "rect must be determined>" << endl;
break;
}
}
exit_main:
cvDestroyWindow( winName.c_str() );
return 0;
}
lazy snapping代碼實現：
// LazySnapping.cpp : 定義控制檯應用程序的入口點。
//
/* author: zhijie Lee
* home page: lzhj.me
* 2012-02-06
*/
#include "stdafx.h"
#include <cv.h>
#include <highgui.h>
#include "graph.h"
#include <vector>
#include <iostream>
#include <cmath>
#include <string>
using namespace std;
typedef Graph<float,float,float> GraphType;
class LasySnapping
{
public :
LasySnapping();
~LasySnapping()
{
if(graph)
{
delete graph;
}
};
private :
vector<CvPoint> forePts;
vector<CvPoint> backPts;
IplImage* image;
// average color of foreground points
unsigned char avgForeColor[3];
// average color of background points
unsigned char avgBackColor[3];
public :
void setImage(IplImage* image)
{
this->image = image;
graph = new GraphType(image->width*image->height,image->width*image->height*2);
}
// include-pen locus
void setForegroundPoints(vector<CvPoint> pts)
{
forePts.clear();
for(int i =0; i< pts.size(); i++)
{
if(!isPtInVector(pts[i],forePts))
{
forePts.push_back(pts[i]);
}
}
if(forePts.size() == 0)
{
return;
}
int sum[3] = {0};
for(int i =0; i < forePts.size(); i++)
{
unsigned char* p = (unsigned char*)image->imageData + forePts[i].x * 3
+ forePts[i].y*image->widthStep;
sum[0] += p[0];
sum[1] += p[1];
sum[2] += p[2];
}
cout<<sum[0]<<" " <<forePts.size()<<endl;
avgForeColor[0] = sum[0]/forePts.size();
avgForeColor[1] = sum[1]/forePts.size();
avgForeColor[2] = sum[2]/forePts.size();
}
// exclude-pen locus
void setBackgroundPoints(vector<CvPoint> pts)
{
backPts.clear();
for(int i =0; i< pts.size(); i++)
{
if(!isPtInVector(pts[i],backPts))
{
backPts.push_back(pts[i]);
}
}
if(backPts.size() == 0)
{
return;
}
int sum[3] = {0};
for(int i =0; i < backPts.size(); i++)
{
unsigned char* p = (unsigned char*)image->imageData + backPts[i].x * 3 +
backPts[i].y*image->widthStep;
sum[0] += p[0];
sum[1] += p[1];
sum[2] += p[2];
}
avgBackColor[0] = sum[0]/backPts.size();
avgBackColor[1] = sum[1]/backPts.size();
avgBackColor[2] = sum[2]/backPts.size();
}
// return maxflow of graph
int runMaxflow();
// get result, a grayscale mast image indicating forground by 255 and background by 0
IplImage* getImageMask();
private :
float colorDistance(unsigned char* color1, unsigned char* color2);
float minDistance(unsigned char* color, vector<CvPoint> points);
bool isPtInVector(CvPoint pt, vector<CvPoint> points);
void getE1(unsigned char* color,float* energy);
float getE2(unsigned char* color1,unsigned char* color2);
GraphType *graph;
};
LasySnapping::LasySnapping()
{
graph = NULL;
avgForeColor[0] = 0;
avgForeColor[1] = 0;
avgForeColor[2] = 0;
avgBackColor[0] = 0;
avgBackColor[1] = 0;
avgBackColor[2] = 0;
}
float LasySnapping::colorDistance(unsigned char* color1, unsigned char* color2)
{
return sqrt(((float)color1[0]-(float)color2[0])*((float)color1[0]-(float)color2[0])+
((float)color1[1]-(float)color2[1])*((float)color1[1]-(float)color2[1])+
((float)color1[2]-(float)color2[2])*((float)color1[2]-(float)color2[2]));
}
float LasySnapping::minDistance(unsigned char* color, vector<CvPoint> points)
{
float distance = -1;
for(int i =0 ; i < points.size(); i++)
{
unsigned char* p = (unsigned char*)image->imageData + points[i].y * image->widthStep +
points[i].x * image->nChannels;
float d = colorDistance(p,color);
if(distance < 0 )
{
distance = d;
}
else
{
if(distance > d)
{
distance = d;
}
}
}
return distance;
}
bool LasySnapping::isPtInVector(CvPoint pt, vector<CvPoint> points)
{
for(int i =0 ; i < points.size(); i++)
{
if(pt.x == points[i].x && pt.y == points[i].y)
{
return true;
}
}
return false;
}
void LasySnapping::getE1(unsigned char* color,float* energy)
{
// average distance
float df = colorDistance(color,avgForeColor);
float db = colorDistance(color,avgBackColor);
// min distance from background points and forground points
// float df = minDistance(color,forePts);
// float db = minDistance(color,backPts);
energy[0] = df/(db+df);
energy[1] = db/(db+df);
}
float LasySnapping::getE2(unsigned char* color1,unsigned char* color2)
{
const float EPSILON = 0.01;
float lambda = 100;
return lambda/(EPSILON+
(color1[0]-color2[0])*(color1[0]-color2[0])+
(color1[1]-color2[1])*(color1[1]-color2[1])+
(color1[2]-color2[2])*(color1[2]-color2[2]));
}
int LasySnapping::runMaxflow()
{
const float INFINNITE_MAX = 1e10;
int indexPt = 0;
for(int h = 0; h < image->height; h ++)
{
unsigned char* p = (unsigned char*)image->imageData + h *image->widthStep;
for(int w = 0; w < image->width; w ++)
{
// calculate energe E1
float e1[2]={0};
if(isPtInVector(cvPoint(w,h),forePts))
{
e1[0] =0;
e1[1] = INFINNITE_MAX;
}
else if
(isPtInVector(cvPoint(w,h),backPts))
{
e1[0] = INFINNITE_MAX;
e1[1] = 0;
}
else
{
getE1(p,e1);
}
// add node
graph->add_node();
graph->add_tweights(indexPt, e1[0],e1[1]);
// add edge, 4-connect
if(h > 0 && w > 0)
{
float e2 = getE2(p,p-3);
graph->add_edge(indexPt,indexPt-1,e2,e2);
e2 = getE2(p,p-image->widthStep);
graph->add_edge(indexPt,indexPt-image->width,e2,e2);
}
p+= 3;
indexPt ++;
}
}
return graph->maxflow();
}
IplImage* LasySnapping::getImageMask()
{
IplImage* gray = cvCreateImage(cvGetSize(image),8,1);
int indexPt =0;
for(int h =0; h < image->height; h++)
{
unsigned char* p = (unsigned char*)gray->imageData + h*gray->widthStep;
for(int w =0 ;w <image->width; w++)
{
if (graph->what_segment(indexPt) == GraphType::SOURCE)
{
*p = 0;
}
else
{
*p = 255;
}
p++;
indexPt ++;
}
}
return gray;
}
// global
vector<CvPoint> forePts;
vector<CvPoint> backPts;
int currentMode = 0;// indicate foreground or background, foreground as default
CvScalar paintColor[2] = {CV_RGB(0,0,255),CV_RGB(255,0,0)};
IplImage* image = NULL;
char* winName = "lazySnapping";
IplImage* imageDraw = NULL;
const int SCALE = 4;
void on_mouse( int event, int x, int y, int flags, void* )
{
if( event == CV_EVENT_LBUTTONUP )
{
if(backPts.size() == 0 && forePts.size() == 0)
{
return;
}
LasySnapping ls;
IplImage* imageLS = cvCreateImage(cvSize(image->width/SCALE,image->height/SCALE),
8,3);
cvResize(image,imageLS);
ls.setImage(imageLS);
ls.setBackgroundPoints(backPts);
ls.setForegroundPoints(forePts);
ls.runMaxflow();
IplImage* mask = ls.getImageMask();
IplImage* gray = cvCreateImage(cvGetSize(image),8,1);
cvResize(mask,gray);
// edge
cvCanny(gray,gray,50,150,3);
IplImage* showImg = cvCloneImage(imageDraw);
for(int h =0; h < image->height; h ++)
{
unsigned char* pgray = (unsigned char*)gray->imageData + gray->widthStep*h;
unsigned char* pimage = (unsigned char*)showImg->imageData + showImg->widthStep*h;
for(int width =0; width < image->width; width++)
{
if(*pgray++ != 0 )
{
pimage[0] = 0;
pimage[1] = 255;
pimage[2] = 0;
}
pimage+=3;
}
}
cvSaveImage("t.bmp",showImg);
cvShowImage(winName,showImg);
cvReleaseImage(&imageLS);
cvReleaseImage(&mask);
cvReleaseImage(&showImg);
cvReleaseImage(&gray);
}
else if( event == CV_EVENT_LBUTTONDOWN )
{
}
else if( event == CV_EVENT_MOUSEMOVE && (flags & CV_EVENT_FLAG_LBUTTON))
{
CvPoint pt = cvPoint(x,y);
if(currentMode == 0)
{//foreground
forePts.push_back(cvPoint(x/SCALE,y/SCALE));
}
else
{//background
backPts.push_back(cvPoint(x/SCALE,y/SCALE));
}
cvCircle(imageDraw,pt,2,paintColor[currentMode]);
cvShowImage(winName,imageDraw);
}
}
int main(int argc, char** argv)
{
//if(argc != 2)
//{
// cout<<"command : lazysnapping inputImage"<<endl;
// return 0;
// }
string image_name;
cout<<"input image name: "<<endl;
cin>>image_name;
cvNamedWindow(winName,1);
cvSetMouseCallback( winName, on_mouse, 0);
image = cvLoadImage(image_name.c_str(),CV_LOAD_IMAGE_COLOR);
imageDraw = cvCloneImage(image);
cvShowImage(winName, image);
for(;;)
{
int c = cvWaitKey(0);
c = (char)c;
if(c == 27)
{//exit
break;
}
else if(c == 'r')
{//reset
image = cvLoadImage(image_name.c_str(),CV_LOAD_IMAGE_COLOR);
imageDraw = cvCloneImage(image);
forePts.clear();
backPts.clear();
currentMode = 0;
cvShowImage(winName, image);
}
else if(c == 'b')
{//change to background selection
currentMode = 1;
}else if(c == 'f')
{//change to foreground selection
currentMode = 0;
}
}
cvReleaseImage(&image);
cvReleaseImage(&imageDraw);
return 0;
}
參考文獻
[1] Y. Boykov, and M. P. Jolly, 「Interactive graph cuts for optimal boundary and region segmentation ofobjects in N-D images」,Proceeding ofIEEE International Conference on Computer Vision, 1:105~112, July 2001.
[2] C. Rother, A. Blake, and V. Kolmogorov, 「Grabcut – interactive foreground extractionusing iterated graph cuts」,Proceedingsof ACM SIGGRAPH 2004, 23(3):307~312, August 2004.
[3] A. Agarwala, M. Dontcheva, M. Agrawala,et al, 「Interactive digital photomontage」,Proceedings of ACM SIGGRAPH 2004, 23(3):294~302, August 2004.
[4] Y. Li, J. Sun, C. Tang,et al, 「Interacting withimages: Lazy snapping」,Proceedingsof ACM SIGGRAPH 2004, 23(3):303~308, August 2004.
[5] A. Blake, C. Rother, M. Brown,et al, 「Interactive ImageSegmentation using an adaptive GMMRF model」.Proceedings of European Conference on Computer Vision, pp. 428~441,May 2004.
[6] V. Kwatra, A. Schodl, I. Essa,et al, 「Graphcut Textures:Image and Video Synthesis Using Graph Cuts」.Proceedings of ACM Siggraph 2003, pp.277~286, Augst 2003.
部分代碼與文檔是早些時候收集的，出處找不到了，還請原做者看到後聯繫註明。
相關文章
- 1. 基於GraphCuts圖割算法的圖像分割----OpenCV代碼與實現
- 2. opencv 圖像分割，掩碼分割，對象分割，grabCut圖割算法
- 3. 基於GraphCuts的醫學影像分割
- 4. 基於 OpenCV 的圖像分割
- 5. 基於OpenCV的圖像分割處理！
- 6. opencv圖像分割
- 7. 圖像分割—基於圖的圖像分割（Graph-Based Image Segmentation）
- 8. C++和opencv實現圖像分割（二）
- 9. C++和opencv實現圖像分割（一）
- 10. OpenCV-Python——第22章：分水嶺算法實現圖像分割
- 更多相關文章...
- • PHP imagecreate - 新建一個基於調色板的圖像 - PHP參考手冊
- • PHP 圖像處理 - PHP參考手冊
- • ☆基於Java Instrument的Agent實現
- • 算法總結-二分查找法
相關標籤/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
歡迎關注本站公眾號,獲取更多信息
