facerecognition.cpp

face recognition using hmm. first version.

	IplImage* img;
	char filename[255];
	const string &PATH_IMAGES = PATH_FACES_IMG_DB+"Images\\";

	ImgObj person;
	int index = -1;
	vector< float > likelihood;
	vector< float >::iterator it;

	ehmmDb.Load( ehmmDbName, pathEhmmDb );
	img = cvLoadImage(image.c_str());
	cvShowImage("Video", img);

	cout << endl << "RecognizeEHMMDb: Database is loaded." << endl << endl;
	cout << ehmmDb;

	person.Create( "Test" );

	person.AddImage( image );

	recog.Create( IMG_WIDTH, IMG_HEIGHT, OBS_WIDTH, OBS_HEIGHT, NO_DCT_COEFF_X, NO_DCT_COEFF_Y, 
		STEP_X, STEP_Y, SUPPRESS_INTESITY );

	img = person.GetGrayScaleImage( 0, 100, 0 );

	index = (int)recog.ComputeLikelihood( *img, ehmmDb, likelihood );

	cvReleaseImage( &img );

	cout << endl << "All likelihoods" << endl << endl;

	for ( it = likelihood.begin( ); it < likelihood.end( ); it++ )
	{
		cout << *it << " ";
	}

	cout << endl << endl;

	if ( index >= 0 )
	{
		cout << "Person is " << ehmmDb.GetObj( index ).GetName( ) << " ( "
			<< likelihood[ index ] << " )." << endl;
	}
	sprintf(filename, "%s\\%s.jpg", 
		ehmmDb.GetObj(index).GetName().c_str(),ehmmDb.GetObj(index).GetName().c_str() );
	printf("recog result filename: %s\n", filename);
	img = cvLoadImage((PATH_IMAGES+filename).c_str());
	cvShowImage("ID", img);
	cvWaitKey(0);
	cvReleaseImage(&img);
}

//*****************************************************************************

void RecognizeEHMMDbFromCam( IplImage *imgCam,
					 const string &pathEhmmDb, 
					 const string &ehmmDbName)
{
	EHMMObjRecognition recog;
	EHMMObjDatabase ehmmDb;
	IplImage* img;
	char filename[255];
	const string &PATH_IMAGES = PATH_FACES_IMG_DB+"Images\\";

	ImgObj person;
	int index = -1;
	vector< float > likelihood;
	vector< float >::iterator it;

	ehmmDb.Load( ehmmDbName, pathEhmmDb );

	cout << endl << "RecognizeEHMMDb: Database is loaded." << endl << endl;
	cout << ehmmDb;

	recog.Create( IMG_WIDTH, IMG_HEIGHT, OBS_WIDTH, OBS_HEIGHT, NO_DCT_COEFF_X, NO_DCT_COEFF_Y, 
		STEP_X, STEP_Y, SUPPRESS_INTESITY );

	img = CamToGray(imgCam, 100, 0);

	index = (int)recog.ComputeLikelihood( *img, ehmmDb, likelihood );

	cvReleaseImage( &img );

	cout << endl << "All likelihoods" << endl << endl;

	for ( it = likelihood.begin( ); it < likelihood.end( ); it++ )
	{
		cout << *it << " ";
	}

	cout << endl << endl;

	if ( index >= 0 )
	{
		cout << "Person is " << ehmmDb.GetObj( index ).GetName( ) << " ( "
			<< likelihood[ index ] << " )." << endl;
	}
	sprintf(filename, "%s\\%s.jpg", 
		ehmmDb.GetObj(index).GetName().c_str(),ehmmDb.GetObj(index).GetName().c_str() );
	printf("recog result filename: %s\n", filename);
	img = cvLoadImage((PATH_IMAGES+filename).c_str());
	cvShowImage("ID", img);
	cvReleaseImage(&img);
//	cvReleaseImage(&imgCam);
}

IplImage *ClipFace( IplImage* src, int x, int y, int w, int h){
    IplImage *target;
	int i,j,x2,y2;

	if(!w || !h) return 0;

	//adjust window size to include more of the face
	x2=x+w; y2=y+h;
	if(x-MARGIN>=0)
		x=x-MARGIN;
	else
		x=0;
	if(y-MARGIN>=0)
		y=y-MARGIN;
	else
		y=0;
	if(x2+MARGIN<src->width)
		x2=x2+MARGIN;
	else
		x2=src->width-1;
	if(y2+MARGIN<src->height)
		y2=y2+MARGIN;
	else
		y2=src->height-1;
	w = x2-x;
	h = y2-y;
	printf("face w:%d h:%d\n", w, h);

	// create target image
	target = cvCreateImage( cvSize(w,h), 8, 3 );

	for (i=y; i<y+h; i++){
		for (j=x; j<x+w; j++){
			target->imageData[(i-y)*target->widthStep+(j-x)*3]=
				src->imageData[i*src->widthStep+j*3];
			target->imageData[(i-y)*target->widthStep+(j-x)*3+1]=
				src->imageData[i*src->widthStep+j*3+1];
			target->imageData[(i-y)*target->widthStep+(j-x)*3+2]=
				src->imageData[i*src->widthStep+j*3+2];
		}
	}
	return target;
}

IplImage* FindFace( IplImage* img )
{
    int scale = 1;
    CvPoint pt1, pt2;
    int i;
	IplImage *imgFace=0;

    cvClearMemStorage( storage );

    if( cascade )
    {
        CvSeq* faces = cvHaarDetectObjects( img, cascade, storage,
                                            1.1, 2, CV_HAAR_DO_CANNY_PRUNING,
                                            cvSize(40, 40) );
        for( i = 0; i < (faces ? faces->total : 0); i++ )
        {
            CvRect* r = (CvRect*)cvGetSeqElem( faces, i );
            pt1.x = r->x*scale;
            pt2.x = (r->x+r->width)*scale;
            pt1.y = r->y*scale;
            pt2.y = (r->y+r->height)*scale;
			imgFace = ClipFace(img, pt1.x, pt1.y, r->width, r->height);
            cvRectangle( img, pt1, pt2, CV_RGB(255,0,0), 3, 8, 0 );
			//just handle one face per frame,ignore the others
			cvShowImage("Search",imgFace);
			return imgFace;
		}
	}
	return imgFace;
}


//*****************************************************************************

int main(int argc, char* argv[])
{
    CvCapture* capture = 0;
    IplImage *frame, *frame_copy = 0;
	int c;

	cout << endl << "HMM Face Reconition" << endl;
	cout << "============================================" <<endl;
	cout << "Press esc to exit program" << endl;
	cout << "press any other key to recognize face" << endl;

	cvNamedWindow("Video",1);
	cvNamedWindow("ID",1);
	cvNamedWindow("Search",1);
	cvMoveWindow("Video",0,0);
	cvMoveWindow("ID", 400,0);
	cvMoveWindow("Search",0, 300); 

	if ( argc > 1 )
	{
		int choice;
		choice = atoi(argv[1]);
		if(choice==1){
			//recognize from a picture
			RecognizeEHMMDb( argv[2], PATH_FACES_EHMM_DB, "faces.db"  );
		}
		else if(choice==2 || choice==3){
			int optlen = strlen("--cascade=");
			cascade_name = cascade_name+optlen;
			cascade = (CvHaarClassifierCascade*)cvLoad( cascade_name, 0, 0, 0 );
		    
			if( !cascade )
			{
				fprintf( stderr, "ERROR: Could not load classifier cascade\n" );
				return -1;
			}
			storage = cvCreateMemStorage(0);

			//recognize from live video stream
			if(choice == 3)
				capture = cvCaptureFromCAM(0);
			else
				//test with video clip
				capture = cvCaptureFromAVI(argv[2]);
			
			if(capture){
				for(;;)
				{
					if( !cvGrabFrame( capture ))
						break;
					frame = cvRetrieveFrame( capture );
					if( !frame )
						break;
					if( !frame_copy )
						frame_copy = cvCreateImage( cvSize(frame->width,frame->height),
													IPL_DEPTH_8U, frame->nChannels );
					if( frame->origin == IPL_ORIGIN_TL )
						cvCopy( frame, frame_copy, 0 );
					else
						cvFlip( frame, frame_copy, 0 );
					cvShowImage("Video", frame_copy);

					c = cvWaitKey(10);
					if (c == '\x1b'){
						printf("Exiting ...\n");
						goto exit_main;
					}
					else if(c>=0){
						IplImage *imgFace;
						imgFace = FindFace(frame_copy);
						if(imgFace){
							RecognizeEHMMDbFromCam(imgFace, PATH_FACES_EHMM_DB, "faces.db");
							cvReleaseImage(&imgFace);
						}
					}
				}
			}
		}
		else{
			cout<<"Can neither find live video camera nor input static image for Face Recognition"<<endl;
			cout<<"Syntax: FaceRecognition [choice [input_image_file_name]]"<<endl;
			cout<<"        choice=1: recognize face from static input image, input_image_filename REQURIED"<<endl;
			cout<<"        choice=2: recognize face from pre-recorded video sequence"<<endl;
			cout<<"        choice=3: recognize face from LIVE cam"<<endl;
		}
		goto exit_main;
	}

	//create face database, training
	CreateFacesImgDb( PATH_FACES_IMG_DB, "FACES.db" );
	LoadImgDb( PATH_FACES_IMG_DB, "FACES.db" );
	CreateFacesEHMMDb( PATH_FACES_EHMM_DB, "FACES.db" );
	TrainEHMMDb( PATH_FACES_IMG_DB, "faces.db", PATH_FACES_EHMM_DB, "faces.db" );

exit_main:
    cvDestroyWindow("Video");
	cvDestroyWindow("ID");

//	if(frame)
//		cvReleaseImage(&frame);
	if(frame_copy)
		cvReleaseImage(&frame_copy);
	if(capture)
		cvReleaseCapture(&capture);
	return 0;
}