cv::Mat TrackFace::cartoonifyImageColor(cv::Mat srcColor)
{
cv::Mat smallImg;
cv::resize(srcColor, smallImg, Size(srcColor.cols/2, srcColor.rows/2), 0, 0, INTER_LINEAR);
cv::Mat tmp=Mat(Size(srcColor.cols/2, srcColor.rows/2), CV_8UC3);
int repetition=7;
for (int i=0;i<repetition;i++)
{
int ksize=9;
double sigmaColor=9;
double sigmaSpace=7;
bilateralFilter(smallImg, tmp, ksize, sigmaColor, sigmaSpace);
bilateralFilter(tmp, smallImg, ksize, sigmaColor, sigmaSpace);
}
cv::Mat bigImg;
cv::resize(smallImg, bigImg, Size(srcColor.cols,srcColor.rows), 0, 0, INTER_LINEAR);
cv::Mat dst;
bigImg.copyTo(dst, cartoonifyImageSketch(srcColor));
return dst;
}
/******************************************************************************
* Input argument(s) : QImage *inputImage - Input image for the filter
* const int &radius - Radius of structuring element
* const bool &roiFlag - Flag to enable region of interest
* QRect *roiRect - Region of interest rectangle
* Return type : QImage* - Output image after bilateral filtering
* Functionality : Function to apply bilateral filter on the given
* input image and returns the processed image
******************************************************************************/
QImage *ImageSmootheningFilter::applyBilateralFilter(QImage *inputImage, const int &radius, const bool &roiFlag, QRect *roiRect) const
{
cv::Mat inputMat = qtOpenCVBridge->QImage2Mat(inputImage);
cv::Mat outputMat = inputMat.clone();
if(roiFlag == true){
cv::Rect regionOfInterest(roiRect->x(), roiRect->y(), roiRect->width(), roiRect->height());
cv::Mat croppedInputMat(inputMat, regionOfInterest);
cv::Mat croppedOutputMat(outputMat, regionOfInterest);
bilateralFilter(croppedInputMat, croppedOutputMat, radius, radius*2.0, radius/2.0);
}
else{
bilateralFilter(inputMat, outputMat, radius, radius*2.0, radius/2.0);
}
return qtOpenCVBridge->Mat2QImage(outputMat);
}
/*
src: input image
method: name of noise reduction method that shall be performed
"average" ==> moving average
"median" ==> median filter
"adaptive" ==> edge preserving average filter
"bilateral" ==> bilateral filter
kSize: (spatial) kernel size
param: if method == "adaptive" : threshold ; if method == "bilateral" standard-deviation of radiometric kernel
can be ignored otherwise (default value = 0)
return: output image
*/
Mat Dip2::noiseReduction(Mat& src, string method, int kSize, double param){
// apply moving average filter
if (method.compare("average") == 0){
return averageFilter(src, kSize);
}
// apply median filter
if (method.compare("median") == 0){
return medianFilter(src, kSize);
}
// apply adaptive average filter
if (method.compare("adaptive") == 0){
return adaptiveFilter(src, kSize, param);
}
// apply bilateral filter
if (method.compare("bilateral") == 0){
return bilateralFilter(src, kSize, param);
}
// if none of above, throw warning and return copy of original
cout << "WARNING: Unknown filtering method! Returning original" << endl;
cout << "Press enter to continue" << endl;
cin.get();
return src.clone();
}
Mat preProcess(Mat &img){
Mat res = Mat(img.rows, img.cols, CV_8UC1);
blur( img, res, Size( 3, 3 ), Point(-1,-1) );
Mat aux = res.clone();
bilateralFilter ( aux, res, 5, 5*2, 5/2 );
aux = res.clone();
cv::GaussianBlur(aux, res, cv::Size(0, 0), 3);
cv::addWeighted(aux, 1.5, res, -0.5, 0, res);
//Filtro de Wiener
cvWiener2ADP(res, res, 5, 5);
//Binarizacao e Afinamento
threshold(res, res, mediana(res), 255, THRESH_BINARY_INV);
//Esqueletização
thinning(res);
/*namedWindow("Preprocess", CV_WINDOW_AUTOSIZE);
imshow("Preprocess", res);
waitKey(0);*/
return res;
}
void process(InputArray _src, OutputArray _dst)
{
Mat src = _src.getMat();
CV_Assert(!src.empty());
_dst.create(src.size(), CV_32FC3);
Mat img = _dst.getMat();
Ptr<Tonemap> linear = createTonemap(1.0f);
linear->process(src, img);
Mat gray_img;
cvtColor(img, gray_img, COLOR_RGB2GRAY);
Mat log_img;
log(gray_img, log_img);
Mat map_img;
bilateralFilter(log_img, map_img, -1, sigma_color, sigma_space);
double min, max;
minMaxLoc(map_img, &min, &max);
float scale = contrast / static_cast<float>(max - min);
exp(map_img * (scale - 1.0f) + log_img, map_img);
log_img.release();
mapLuminance(img, img, gray_img, map_img, saturation);
pow(img, 1.0f / gamma, img);
}
KDvoid Smoothing ( KDint nIdx )
{
Mat tSrc;
Mat tDst;
KDint MAX_KERNEL_LENGTH = 9; // 31;
KDint i;
// Load the source image
tSrc = imread ( "/res/image/lena.jpg", 1 );
tDst = tSrc.clone ( );
for ( i = 1; i < MAX_KERNEL_LENGTH; i = i + 2 )
{
switch ( nIdx )
{
case 0 : blur ( tSrc, tDst, Size ( i, i ), Point ( -1, -1 ) ); break; // Applying Homogeneous blur
case 1 : GaussianBlur ( tSrc, tDst, Size ( i, i ), 0, 0 ); break; // Applying Gaussian blur
case 2 : medianBlur ( tSrc, tDst, i ); break; // Applying Median blur
case 3 : bilateralFilter ( tSrc, tDst, i, i * 2, i / 2 ); break; // Applying Bilateral Filter
}
}
g_pController->setFrame ( 1, tSrc );
g_pController->setFrame ( 2, tDst );
}
ImageImPro* OpenImProLib_OpenCvImpl::filterBilateral(ImageImPro* ptrInput, int winDiameter, double sigmaRange, double sigmaSpace){
Mat* ptrMatInput = ptrInput->getMat();
Mat* ptrMatOutput = ptrInput->getMat();
bilateralFilter( *ptrMatInput, *ptrMatOutput, winDiameter, sigmaRange, sigmaSpace);
ImageImPro* ptrOutput = new ImageImPro_OpenCvImpl(ptrMatOutput);
delete ptrMatInput;
delete ptrMatOutput;
return ptrOutput;
}
void bilateral(Mat& frame) {
Mat tmp;
for (int i = 0; i < 2; ++i) {
pyrDown(frame, frame, Size(frame.cols / 2, frame.rows / 2));
}
bilateralFilter(tmp, frame, 3, 3, 3);
for (int i = 0; i < 2; ++i) {
pyrUp(frame, frame, Size(frame.cols * 2, frame.rows * 2));
}
}
void bilateralSatured(Mat& frame) {
Mat tmp;
for (int i = 0; i < 2; ++i) {
pyrDown(frame, frame, Size(frame.cols / 2, frame.rows / 2));
}
saturar(frame, tmp, 55);
bilateralFilter(tmp, frame, 3, 3, 3);
for (int i = 0; i < 2; ++i) {
pyrUp(frame, frame, Size(frame.cols * 2, frame.rows * 2));
}
}
bool BlurBlock::run(bool oneShot){
Mat imgSrc = _myInputs["BLOCK__BLUR_IN_IMG"].get<cv::Mat>(),
imgOut;
switch (_myInputs["BLOCK__BLUR_IN_METHOD"].get<int>())
{
case 0://Mean
{
blur(imgSrc, imgOut,
cv::Size(_mySubParams["BLOCK__BLUR_IN_METHOD.Mean.kernel size X"].get<int>(),
_mySubParams["BLOCK__BLUR_IN_METHOD.Mean.kernel size Y"].get<int>()),
cv::Point(_mySubParams["BLOCK__BLUR_IN_METHOD.Mean.anchor point X"].get<int>(),
_mySubParams["BLOCK__BLUR_IN_METHOD.Mean.anchor point Y"].get<int>()),
cv::BORDER_DEFAULT);
break;
}
case 1://Gaussian
{
cv::Size ksize(_mySubParams["BLOCK__BLUR_IN_METHOD.Gaussian.kernel size X"].get<int>(),
_mySubParams["BLOCK__BLUR_IN_METHOD.Gaussian.kernel size Y"].get<int>());
if (ksize.width <= 0) ksize.width = 1;
if (ksize.width % 2 == 0) ksize.width += 1;
if (ksize.height <= 0) ksize.height = 1;
if (ksize.height % 2 == 0) ksize.height += 1;
GaussianBlur(imgSrc, imgOut, ksize,
_mySubParams["BLOCK__BLUR_IN_METHOD.Gaussian.Sigma X"].get<double>(),
_mySubParams["BLOCK__BLUR_IN_METHOD.Gaussian.Sigma Y"].get<double>(),
cv::BORDER_DEFAULT);
break;
}
case 2://Median
{
int medianSize = _mySubParams["BLOCK__BLUR_IN_METHOD.Median.kernel size"].get<int>();
if (medianSize % 2 != 1)
medianSize += 1;
medianBlur(imgSrc, imgOut, medianSize);
break;
}
case 3://Bilateral
{
bilateralFilter(imgSrc, imgOut, _mySubParams["BLOCK__BLUR_IN_METHOD.Bilateral.Diameter"].get<int>(),
_mySubParams["BLOCK__BLUR_IN_METHOD.Bilateral.Sigma color"].get<double>(),
_mySubParams["BLOCK__BLUR_IN_METHOD.Bilateral.Sigma space"].get<double>());
break;
}
default:
return false;//nothing to do as we don't support this type of operation
break;
}
_myOutputs["BLOCK__BLUR_OUT_IMAGE"] = imgOut;
return true;
};
void Filter::applyFilter()
{
configureSpinBox();
switch(currentFilter){
case FILTER_HOMOGENEOUS:
blur(originalImage, image, ksize);
ui->filteredImage->setPixmap(ImageHandler::getQPixmap(image));
break;
case FILTER_GAUSSIAN:
GaussianBlur(originalImage, image, ksize, 0, 0);
ui->filteredImage->setPixmap(ImageHandler::getQPixmap(image));
break;
case FILTER_MEDIAN:
medianBlur(originalImage, image, ksize.height);
ui->filteredImage->setPixmap(ImageHandler::getQPixmap(image));
break;
case FILTER_BILATERAL:
bilateralFilter(originalImage, image, 5, sigma, sigma);
break;
}
}
int runFilter(int argc, char *argv[])
{
char * filenameInput=argv[1];
char * filenameOutput=argv[2];
unsigned int inputType = guessFilenameTypeStupid(filenameInput);
struct Image * inputImage = readImage(filenameInput,inputType,0);
struct Image * outputImage = 0; //This will get allocated when and if needed
if (inputImage!=0)
{
unsigned int outputType = guessFilenameTypeStupid(filenameOutput);
unsigned int i=0;
for (i=0; i<argc; i++)
{
if ( strcmp(argv[i],"--learn")==0 )
{
destroyImage(inputImage);
learnImage(filenameInput,atoi(argv[i+1]),atoi(argv[i+2]));
exit(0);
} else
if ( strcmp(argv[i],"--rgbcube")==0 )
{
unsigned int dim=32;
unsigned char R = (char) atoi(argv[i]+1);
unsigned char G = (char) atoi(argv[i]+2);
unsigned char B = (char) atoi(argv[i]+3);
outputImage = createImage( dim , dim , 3 , 8 );
bitbltColorRGB(outputImage->pixels , 0 , 0 , dim , dim , R , G , B , dim-1 , dim-1);
writeImageFile(outputImage,PPM_CODEC ,"new_mX.pnm");
writeImageFile(outputImage,PPM_CODEC ,"new_pX.pnm");
writeImageFile(outputImage,PPM_CODEC ,"new_mY.pnm");
writeImageFile(outputImage,PPM_CODEC ,"new_pY.pnm");
writeImageFile(outputImage,PPM_CODEC ,"new_mZ.pnm");
writeImageFile(outputImage,PPM_CODEC ,"new_pZ.pnm");
destroyImage(outputImage);
} else
if ( strcmp(argv[i],"--envcube")==0 )
{
fprintf(stdout,"Converting Environment Cube \n");
unsigned int outputType = guessFilenameTypeStupid(filenameOutput);
//outputImage = createSameDimensionsImage(inputImage);
unsigned int outputWidth = inputImage->width;
unsigned int outputHeight = (unsigned int ) (3*inputImage->width)/4;
outputImage = createImage( outputWidth , outputHeight , 3 , 8 );
createCubeMapFace( outputImage->pixels , outputImage->width , outputImage->height , outputImage->channels , outputImage->bitsperpixel ,
inputImage->pixels , inputImage->width , inputImage->height , inputImage->channels , inputImage->bitsperpixel
);
struct Image * partImg=0;
unsigned int outX=0 , outY=0 , outWidth=0 , outHeight=0;
getCubeMap2DCoords(outputWidth,outputHeight, /*x*/ -1 , /*y*/ 0 , /*z*/ 0 , &outX , &outY , &outWidth , &outHeight );
partImg=createImageBitBlt( outputImage , outX , outY , outWidth , outHeight );
writeImageFile(partImg,PPM_CODEC ,"new_mX.pnm"); destroyImage(partImg);
getCubeMap2DCoords(outputWidth,outputHeight, /*x*/ 1 , /*y*/ 0 , /*z*/ 0 , &outX , &outY , &outWidth , &outHeight );
partImg=createImageBitBlt( outputImage , outX , outY , outWidth , outHeight );
writeImageFile(partImg,PPM_CODEC ,"new_pX.pnm"); destroyImage(partImg);
getCubeMap2DCoords(outputWidth,outputHeight, /*x*/ 0 , /*y*/ -1 , /*z*/ 0 , &outX , &outY , &outWidth , &outHeight );
partImg=createImageBitBlt( outputImage , outX , outY , outWidth , outHeight );
writeImageFile(partImg,PPM_CODEC ,"new_mY.pnm"); destroyImage(partImg);
getCubeMap2DCoords(outputWidth,outputHeight, /*x*/ 0 , /*y*/ 1 , /*z*/ 0 , &outX , &outY , &outWidth , &outHeight );
partImg=createImageBitBlt( outputImage , outX , outY , outWidth , outHeight );
writeImageFile(partImg,PPM_CODEC ,"new_pY.pnm"); destroyImage(partImg);
getCubeMap2DCoords(outputWidth,outputHeight, /*x*/ 0 , /*y*/ 0 , /*z*/ -1 , &outX , &outY , &outWidth , &outHeight );
partImg=createImageBitBlt( outputImage , outX , outY , outWidth , outHeight );
writeImageFile(partImg,PPM_CODEC ,"new_mZ.pnm"); destroyImage(partImg);
getCubeMap2DCoords(outputWidth,outputHeight, /*x*/ 0 , /*y*/ 0 , /*z*/ 1 , &outX , &outY , &outWidth , &outHeight );
partImg=createImageBitBlt( outputImage , outX , outY , outWidth , outHeight );
writeImageFile(partImg,PPM_CODEC ,"new_pZ.pnm"); destroyImage(partImg);
} else
if ( strcmp(argv[i],"--compare")==0 )
{
unsigned int outputType = guessFilenameTypeStupid(filenameOutput);
outputImage = readImage(filenameOutput,outputType ,0);
float noise = calculatePSNR( outputImage->pixels , outputImage->width , outputImage->height , outputImage->channels ,
inputImage->pixels , inputImage->width , inputImage->height , inputImage->channels );
fprintf(stdout,"Compared Detected Noise is %0.4f dB \n",noise);
exit(0);
} else
if ( strcmp(argv[i],"--gaussian")==0 )
{
monochrome(inputImage);
outputImage = createSameDimensionsImage(inputImage);
unsigned int normalizeGaussianKernel=1;
unsigned int kernelWidth=5;
unsigned int kernelHeight=5;
float * convolutionMatrix=allocateGaussianKernel(kernelWidth,kernelHeight,normalizeGaussianKernel);
float divisor=1.0;
float * inF = copyUCharImage2Float(inputImage->pixels , inputImage->width , inputImage->height , inputImage->channels );
float * outF = (float*) malloc(sizeof(float) * outputImage->width * outputImage->height * outputImage->channels );
convolutionFilter1ChF(
outF , outputImage->width , outputImage->height ,
inF, inputImage->width , inputImage->height ,
convolutionMatrix , kernelWidth , kernelHeight , &divisor
);
free(convolutionMatrix);
castFloatImage2UChar(outputImage->pixels, outF, outputImage->width , outputImage->height , outputImage->channels );
free(inF);
free(outF);
} else
if ( strcmp(argv[i],"--ctbilateral")==0 )
{
monochrome(inputImage);
outputImage = createSameDimensionsImage(inputImage);
float sigma = atof(argv[i+1]);
constantTimeBilateralFilter(
inputImage->pixels , inputImage->width , inputImage->height , inputImage->channels ,
outputImage->pixels , outputImage->width , outputImage->height
,&sigma //sigma
,atoi(argv[i+2]) //bins
,atoi(argv[i+3]) //useDeriche
);
} else
if ( strcmp(argv[i],"--deriche")==0 )
{
monochrome(inputImage);
outputImage = createSameDimensionsImage(inputImage);
float sigma = atof(argv[i+1]);
dericheRecursiveGaussianGray( outputImage->pixels , outputImage->width , outputImage->height , inputImage->channels ,
inputImage->pixels , inputImage->width , inputImage->height ,
&sigma , atoi(argv[i+2])
);
} else
if ( strcmp(argv[i],"--dericheF")==0 )
{
fprintf(stderr,"This is a test call for casting code , this shouldnt be normally used..\n");
monochrome(inputImage);
outputImage = createSameDimensionsImage(inputImage);
float sigma = atof(argv[i+1]);
//outputImage = copyImage(inputImage);
float * inF = copyUCharImage2Float(inputImage->pixels , inputImage->width , inputImage->height , inputImage->channels );
float * outF = (float*) malloc(sizeof(float) * outputImage->width * outputImage->height * outputImage->channels );
dericheRecursiveGaussianGrayF( outF , outputImage->width , outputImage->height , inputImage->channels ,
inF , inputImage->width , inputImage->height ,
&sigma , atoi(argv[i+2])
);
castFloatImage2UChar(outputImage->pixels, outF, outputImage->width , outputImage->height , outputImage->channels );
free(inF);
free(outF);
} else
if ( strcmp(argv[i],"--median")==0 )
{
outputImage = copyImage(inputImage);
medianFilter3ch(
outputImage->pixels , outputImage->width , outputImage->height ,
inputImage->pixels , inputImage->width , inputImage->height ,
atoi(argv[i+1]) , atoi(argv[i+2])
);
} else
if ( strcmp(argv[i],"--meansat")==0 )
{
outputImage = copyImage(inputImage);
meanFilterSAT(
outputImage->pixels , outputImage->width , outputImage->height , outputImage->channels ,
inputImage->pixels , inputImage->width , inputImage->height , inputImage->channels ,
atoi(argv[i+1]) , atoi(argv[i+2])
);
} else
if ( strcmp(argv[i],"--monochrome")==0 )
{
outputImage = copyImage(inputImage);
monochrome(outputImage);
} else
if ( strcmp(argv[i],"--bilateral")==0 )
{
outputImage = copyImage(inputImage);
bilateralFilter( outputImage->pixels , outputImage->width , outputImage->height ,
inputImage->pixels , inputImage->width , inputImage->height ,
atof(argv[i+1]) , atof(argv[i+2]) , atoi(argv[i+3])
);
} else
if ( strcmp(argv[i],"--contrast")==0 )
{
outputImage = copyImage(inputImage);
contrast(outputImage,atof(argv[i+1]));
} else
if ( strcmp(argv[i],"--sattest")==0 )
{
float * tmp = allocateGaussianKernel(3,5.0,1);
if (tmp!=0) { free(tmp); }
tmp = allocateGaussianKernel(9,5.0,1);
if (tmp!=0) { free(tmp); }
tmp = allocateGaussianKernel(15,5.0,1);
if (tmp!=0) { free(tmp); }
summedAreaTableTest();
unsigned int * integralImageOutput = 0;
integralImageOutput = generateSummedAreaTableRGB(inputImage->pixels , inputImage->width , inputImage->height);
if (integralImageOutput!=0)
{
free(integralImageOutput);
fprintf(stderr,"integralImage test was successful\n");
}
}
}
writeImageFile(outputImage,outputType ,filenameOutput);
destroyImage(outputImage);
destroyImage(inputImage);
return 1;
}
return 0;
}
void RGBDCamera::update(const RawFrame* this_frame) {
//Check the timestamp, and skip if we have already seen this frame
if (this_frame->timestamp <= latest_stamp_) {
return;
} else {
latest_stamp_ = this_frame->timestamp;
}
//Apply bilateral filter to incoming depth
uint16_t* filtered_depth;
cudaMalloc((void**)&filtered_depth, this_frame->width*this_frame->height*sizeof(uint16_t));
bilateralFilter(this_frame->depth, filtered_depth, this_frame->width, this_frame->height);
//Convert the input color data to intensity
float* temp_intensity;
cudaMalloc((void**)&temp_intensity, this_frame->width*this_frame->height*sizeof(float));
colorToIntensity(this_frame->color, temp_intensity, this_frame->width*this_frame->height);
//Create pyramids
for (int i = 0; i < PYRAMID_DEPTH; i++) {
//Fill in sizes the first two times through
if (pass_ < 2) {
current_icp_frame_[i] = new ICPFrame(this_frame->width/pow(2,i), this_frame->height/pow(2,i));
current_rgbd_frame_[i] = new RGBDFrame(this_frame->width/pow(2,i), this_frame->height/pow(2,i));
}
//Add ICP data
generateVertexMap(filtered_depth, current_icp_frame_[i]->vertex, current_icp_frame_[i]->width, current_icp_frame_[i]->height, focal_length_, make_int2(this_frame->width, this_frame->height));
generateNormalMap(current_icp_frame_[i]->vertex, current_icp_frame_[i]->normal, current_icp_frame_[i]->width, current_icp_frame_[i]->height);
//Add RGBD data
cudaMemcpy(current_rgbd_frame_[i]->vertex, current_icp_frame_[i]->vertex, current_rgbd_frame_[i]->width*current_rgbd_frame_[i]->height*sizeof(glm::vec3), cudaMemcpyDeviceToDevice);
cudaMemcpy(current_rgbd_frame_[i]->intensity, temp_intensity, current_rgbd_frame_[i]->width*current_rgbd_frame_[i]->height*sizeof(float), cudaMemcpyDeviceToDevice);
//Downsample depth and color if not the last iteration
if (i != (PYRAMID_DEPTH-1)) {
subsampleDepth(filtered_depth, current_icp_frame_[i]->width, current_icp_frame_[i]->height);
subsample(temp_intensity, current_rgbd_frame_[i]->width, current_rgbd_frame_[i]->height);
cudaDeviceSynchronize();
}
}
//Clear the filtered depth and temporary color since they are no longer needed
cudaFree(filtered_depth);
cudaFree(temp_intensity);
if (pass_ >= 1) {
glm::mat4 update_trans(1.0f);
//Loop through pyramids backwards (coarse first)
for (int i = PYRAMID_DEPTH - 1; i >= 0; i--) {
//Get a copy of the ICP frame for this pyramid level
ICPFrame icp_f(current_icp_frame_[i]->width, current_icp_frame_[i]->height);
cudaMemcpy(icp_f.vertex, current_icp_frame_[i]->vertex, icp_f.width*icp_f.height*sizeof(glm::vec3), cudaMemcpyDeviceToDevice);
cudaMemcpy(icp_f.normal, current_icp_frame_[i]->normal, icp_f.width*icp_f.height*sizeof(glm::vec3), cudaMemcpyDeviceToDevice);
//Get a copy of the RGBD frame for this pyramid level
//RGBDFrame rgbd_f(current_rgbd_frame_[i]->width, current_rgbd_frame_[i]->height);
//cudaMemcpy(rgbd_f.vertex, current_rgbd_frame_[i]->vertex, rgbd_f.width*rgbd_f.height*sizeof(glm::vec3), cudaMemcpyDeviceToDevice);
//cudaMemcpy(rgbd_f.intensity, current_rgbd_frame_[i]->intensity, rgbd_f.width*rgbd_f.height*sizeof(float), cudaMemcpyDeviceToDevice);
//Apply the most recent update to the points/normals
if (i < (PYRAMID_DEPTH-1)) {
transformVertexMap(icp_f.vertex, update_trans, icp_f.width*icp_f.height);
transformNormalMap(icp_f.normal, update_trans, icp_f.width*icp_f.height);
cudaDeviceSynchronize();
}
//Loop through iterations
for (int j = 0; j < PYRAMID_ITERS[i]; j++) {
//Get the Geometric ICP cost values
float A1[6 * 6];
float b1[6];
computeICPCost2(last_icp_frame_[i], icp_f, A1, b1);
//Get the Photometric RGB-D cost values
//float A2[6*6];
//float b2[6];
//compueRGBDCost(last_rgbd_frame_, rgbd_f, A2, b2);
//Combine the two
//for (size_t k = 0; k < 6; k++) {
//for (size_t l = 0; l < 6; l++) {
//A1[6 * k + l] += A2[6 * k + l];
//}
//b1[k] += b2[k];
//}
//Solve for the optimized camera transformation
float x[6];
solveCholesky(6, A1, b1, x);
//Check for NaN/divergence
if (isnan(x[0]) || isnan(x[1]) || isnan(x[2]) || isnan(x[3]) || isnan(x[4]) || isnan(x[5])) {
printf("Camera tracking is lost.\n");
break;
}
//Update position/orientation of the camera
glm::mat4 this_trans =
glm::rotate(glm::mat4(1.0f), -x[2] * 180.0f / 3.14159f, glm::vec3(0.0f, 0.0f, 1.0f))
* glm::rotate(glm::mat4(1.0f), -x[1] * 180.0f / 3.14159f, glm::vec3(0.0f, 1.0f, 0.0f))
* glm::rotate(glm::mat4(1.0f), -x[0] * 180.0f / 3.14159f, glm::vec3(1.0f, 0.0f, 0.0f))
* glm::translate(glm::mat4(1.0f), glm::vec3(x[3], x[4], x[5]));
update_trans = this_trans * update_trans;
//Apply the update to the points/normals
if (j < (PYRAMID_ITERS[i] - 1)) {
transformVertexMap(icp_f.vertex, this_trans, icp_f.width*icp_f.height);
transformNormalMap(icp_f.normal, this_trans, icp_f.width*icp_f.height);
cudaDeviceSynchronize();
}
}
}
//Update the global transform with the result
position_ = glm::vec3(glm::vec4(position_, 1.0f) * update_trans);
orientation_ = glm::mat3(glm::mat4(orientation_) * update_trans);
}
if (pass_ < 2) {
pass_++;
}
//Swap current and last frames
for (int i = 0; i < PYRAMID_DEPTH; i++) {
ICPFrame* temp = current_icp_frame_[i];
current_icp_frame_[i] = last_icp_frame_[i];
last_icp_frame_[i] = temp;
//TODO: Longterm, only RGBD should do this. ICP should not swap, as last_frame should be updated by a different function
RGBDFrame* temp2 = current_rgbd_frame_[i];
current_rgbd_frame_[i] = last_rgbd_frame_[i];
last_rgbd_frame_[i] = temp2;
}
}
int opencv::ProcessImg(Mat& src)
{
try{
if (src.empty()) return 1;
Mat color, gray, binimg, bingray;
Convert2BGR(src, color);
Convert2GRAY(src, gray); Convert2GRAY(src, bingray);
exColor_Invert(gray);
vector<vector<Point>> contours;
int trd = get_OTSU_value(gray);
threshold(bingray, binimg, trd, 255, THRESH_BINARY);
Mat Fiter;
#if 1
//双边滤波
bilateralFilter(gray, Fiter, 15, 50, 2.0, BORDER_DEFAULT);
#else
//中值滤波
int kenerlsize = gray.rows / 150;
if (0 == kenerlsize % 2) kenerlsize += 1;
kenerlsize = kenerlsize > 2 ? kenerlsize : 3;
medianBlur(gray, gray, kenerlsize);
#endif
//Histogram(gray, gray);
findContours(binimg, contours, CV_RETR_CCOMP, CV_CHAIN_APPROX_NONE, cvPoint(0, 0));
Mat imgmask(binimg.rows, binimg.cols, CV_8UC1, Scalar(255));
//drawContours(imgmask, contours, -1, Scalar(255), CV_FILLED/*2*/); // -1 表示所有轮廓
//drawContours(imgmask, contours, -1, Scalar(0), 2); // -1 表示所有轮廓
#ifdef _DEBUG
for (size_t i = 0; i < contours.size(); i++)
cout << contours[i].size()<<endl;
#endif
get_mask_image(src, imgmask, contours, -1,1,2);
src = imgmask.clone();
return 1;
Convert2GRAY(gray, gray);
Mat edge;
Canny(gray, edge, 0, 255, 5);
src = edge.clone();
return NoError;
//模式识别
// ANN_MLP ann;
// ann.train();
}
catch (...)
{
#ifdef _DEBUG
cout << "img deal wrongly" << endl;
#endif
return -1;
}
}
void tmo_durand02(pfs::Array2Df& R, pfs::Array2Df& G, pfs::Array2Df& B,
float sigma_s, float sigma_r, float baseContrast, int downsample,
bool color_correction,
pfs::Progress &ph)
{
int w = R.getCols();
int h = R.getRows();
int size = w*h;
pfs::Array2Df I(w,h); // intensities
pfs::Array2Df BASE(w,h); // base layer
pfs::Array2Df DETAIL(w,h); // detail layer
float min_pos = 1e10f; // minimum positive value (to avoid log(0))
for (int i = 0 ; i < size ; i++)
{
I(i) = 1.0f/61.0f * ( 20.0f*R(i) + 40.0f*G(i) + B(i) );
if ( I(i) < min_pos && I(i) > 0.0f )
{
min_pos = I(i);
}
}
for (int i = 0 ; i < size ; i++)
{
float L = I(i);
if ( L <= 0.0f )
{
L = min_pos;
}
R(i) /= L;
G(i) /= L;
B(i) /= L;
I(i) = std::log( L );
}
#ifdef HAVE_FFTW3F
fastBilateralFilter( I, BASE, sigma_s, sigma_r, downsample, ph );
#else
bilateralFilter( &I, &BASE, sigma_s, sigma_r, ph );
#endif
//!! FIX: find minimum and maximum luminance, but skip 1% of outliers
float maxB;
float minB;
findMaxMinPercentile(&BASE, 0.01f, 0.99f, minB, maxB);
float compressionfactor = baseContrast / (maxB - minB);
// Color correction factor
const float k1 = 1.48f;
const float k2 = 0.82f;
const float s = ( (1 + k1)*pow(compressionfactor,k2) )/( 1 + k1*pow(compressionfactor,k2) );
for (int i = 0 ; i < size ; i++)
{
DETAIL(i) = I(i) - BASE(i);
I(i) = BASE(i) * compressionfactor + DETAIL(i);
//!! FIX: this to keep the output in normalized range 0.01 - 1.0
//intensitites are related only to minimum luminance because I
//would say this is more stable over time than using maximum
//luminance and is also robust against random peaks of very high
//luminance
I(i) -= 4.3f+minB*compressionfactor;
if ( color_correction )
{
R(i) = decode( std::pow( R(i), s ) * std::exp( I(i) ) );
G(i) = decode( std::pow( G(i), s ) * std::exp( I(i) ) );
B(i) = decode( std::pow( B(i), s ) * std::exp( I(i) ) );
}
else
{
R(i) *= decode( std::exp( I(i) ) );
G(i) *= decode( std::exp( I(i) ) );
B(i) *= decode( std::exp( I(i) ) );
}
}
if (!ph.canceled())
{
ph.setValue( 100 );
}
}
void PlateLines::processImage(Mat inputImage, vector<TextLine> textLines, float sensitivity)
{
if (this->debug)
cout << "PlateLines findLines" << endl;
timespec startTime;
getTimeMonotonic(&startTime);
// Ignore input images that are pure white or pure black
Scalar avgPixelIntensity = mean(inputImage);
if (avgPixelIntensity[0] >= 252)
return;
else if (avgPixelIntensity[0] <= 3)
return;
// Do a bilateral filter to clean the noise but keep edges sharp
Mat smoothed(inputImage.size(), inputImage.type());
bilateralFilter(inputImage, smoothed, 3, 45, 45);
int morph_elem = 2;
int morph_size = 2;
Mat element = getStructuringElement( morph_elem, Size( 2*morph_size + 1, 2*morph_size+1 ), Point( morph_size, morph_size ) );
Mat edges(inputImage.size(), inputImage.type());
Canny(smoothed, edges, 66, 133);
// Create a mask that is dilated based on the detected characters
Mat mask = Mat::zeros(inputImage.size(), CV_8U);
for (unsigned int i = 0; i < textLines.size(); i++)
{
vector<vector<Point> > polygons;
polygons.push_back(textLines[i].textArea);
fillPoly(mask, polygons, Scalar(255,255,255));
}
dilate(mask, mask, getStructuringElement( 1, Size( 1 + 1, 2*1+1 ), Point( 1, 1 ) ));
bitwise_not(mask, mask);
// AND canny edges with the character mask
bitwise_and(edges, mask, edges);
vector<PlateLine> hlines = this->getLines(edges, sensitivity, false);
vector<PlateLine> vlines = this->getLines(edges, sensitivity, true);
for (unsigned int i = 0; i < hlines.size(); i++)
this->horizontalLines.push_back(hlines[i]);
for (unsigned int i = 0; i < vlines.size(); i++)
this->verticalLines.push_back(vlines[i]);
// if debug is enabled, draw the image
if (this->debug)
{
Mat debugImgHoriz(edges.size(), edges.type());
Mat debugImgVert(edges.size(), edges.type());
edges.copyTo(debugImgHoriz);
edges.copyTo(debugImgVert);
cvtColor(debugImgHoriz,debugImgHoriz,CV_GRAY2BGR);
cvtColor(debugImgVert,debugImgVert,CV_GRAY2BGR);
for( size_t i = 0; i < this->horizontalLines.size(); i++ )
{
line( debugImgHoriz, this->horizontalLines[i].line.p1, this->horizontalLines[i].line.p2, Scalar(0,0,255), 1, CV_AA);
}
for( size_t i = 0; i < this->verticalLines.size(); i++ )
{
line( debugImgVert, this->verticalLines[i].line.p1, this->verticalLines[i].line.p2, Scalar(0,0,255), 1, CV_AA);
}
vector<Mat> images;
images.push_back(debugImgHoriz);
images.push_back(debugImgVert);
Mat dashboard = drawImageDashboard(images, debugImgVert.type(), 1);
displayImage(pipelineData->config, "Hough Lines", dashboard);
}
if (pipelineData->config->debugTiming)
{
timespec endTime;
getTimeMonotonic(&endTime);
cout << "Plate Lines Time: " << diffclock(startTime, endTime) << "ms." << endl;
}
}
// Create a grayscale face image that has a standard size and contrast & brightness.
// "srcImg" should be a copy of the whole color camera frame, so that it can draw the eye positions onto.
// If 'doLeftAndRightSeparately' is true, it will process left & right sides seperately,
// so that if there is a strong light on one side but not the other, it will still look OK.
// Performs Face Preprocessing as a combination of:
// - geometrical scaling, rotation and translation using Eye Detection,
// - smoothing away image noise using a Bilateral Filter,
// - standardize the brightness on both left and right sides of the face independently using separated Histogram Equalization,
// - removal of background and hair using an Elliptical Mask.
// Returns either a preprocessed face square image or NULL (ie: couldn't detect the face and 2 eyes).
// If a face is found, it can store the rect coordinates into 'storeFaceRect' and 'storeLeftEye' & 'storeRightEye' if given,
// and eye search regions into 'searchedLeftEye' & 'searchedRightEye' if given.
cv::Mat preprocessFace::getPreprocessedFace(cv::Mat &srcImg, int desiredFaceWidth, cv::CascadeClassifier &faceCascade, cv::CascadeClassifier &eyeCascade1, cv::CascadeClassifier &eyeCascade2, bool doLeftAndRightSeparately, cv::Rect *storeFaceRect, cv::Point *storeLeftEye, cv::Point *storeRightEye, cv::Rect *searchedLeftEye, cv::Rect *searchedRightEye)
{
// Use square faces.
int desiredFaceHeight = desiredFaceWidth;
// Mark the detected face region and eye search regions as invalid, in case they aren't detected.
if (storeFaceRect)
storeFaceRect->width = -1;
if (storeLeftEye)
storeLeftEye->x = -1;
if (storeRightEye)
storeRightEye->x = -1;
if (searchedLeftEye)
searchedLeftEye->width = -1;
if (searchedRightEye)
searchedRightEye->width = -1;
// Find the largest face.
cv::Rect faceRect;
detector.detectLargestObject(srcImg, faceCascade, faceRect);
// Check if a face was detected.
if (faceRect.width > 0) {
// Give the face rect to the caller if desired.
if (storeFaceRect)
*storeFaceRect = faceRect;
cv::Mat faceImg = srcImg(faceRect); // Get the detected face image.
// If the input image is not grayscale, then convert the BGR or BGRA color image to grayscale.
cv::Mat gray;
if (faceImg.channels() == 3) {
cvtColor(faceImg, gray, CV_BGR2GRAY);
}
else if (faceImg.channels() == 4) {
cvtColor(faceImg, gray, CV_BGRA2GRAY);
}
else {
// Access the input image directly, since it is already grayscale.
gray = faceImg;
}
// Search for the 2 eyes at the full resolution, since eye detection needs max resolution possible!
cv::Point leftEye, rightEye;
detectBothEyes(gray, eyeCascade1, eyeCascade2, leftEye, rightEye, searchedLeftEye, searchedRightEye);
// Give the eye results to the caller if desired.
if (storeLeftEye)
*storeLeftEye = leftEye;
if (storeRightEye)
*storeRightEye = rightEye;
// Check if both eyes were detected.
if (leftEye.x >= 0 && rightEye.x >= 0)
{
inputFaceCount++;
// Make the face image the same size as the training images.
// Since we found both eyes, lets rotate & scale & translate the face so that the 2 eyes
// line up perfectly with ideal eye positions. This makes sure that eyes will be horizontal,
// and not too far left or right of the face, etc.
// Get the center between the 2 eyes.
cv::Point2f eyesCenter = cv::Point2f((leftEye.x + rightEye.x) * 0.5f, (leftEye.y + rightEye.y) * 0.5f);
// Get the angle between the 2 eyes.
double dy = (rightEye.y - leftEye.y);
double dx = (rightEye.x - leftEye.x);
double len = sqrt(dx*dx + dy*dy);
double angle = atan2(dy, dx) * 180.0 / CV_PI; // Convert from radians to degrees.
// Hand measurements shown that the left eye center should ideally be at roughly (0.19, 0.14) of a scaled face image.
const double DESIRED_RIGHT_EYE_X = (1.0f - DESIRED_LEFT_EYE_X);
// Get the amount we need to scale the image to be the desired fixed size we want.
double desiredLen = (DESIRED_RIGHT_EYE_X - DESIRED_LEFT_EYE_X) * desiredFaceWidth;
double scale = desiredLen / len;
// Get the transformation matrix for rotating and scaling the face to the desired angle & size.
cv::Mat rot_mat = getRotationMatrix2D(eyesCenter, angle, scale);
// Shift the center of the eyes to be the desired center between the eyes.
rot_mat.at<double>(0, 2) += desiredFaceWidth * 0.5f - eyesCenter.x;
rot_mat.at<double>(1, 2) += desiredFaceHeight * DESIRED_LEFT_EYE_Y - eyesCenter.y;
// Rotate and scale and translate the image to the desired angle & size & position!
// Note that we use 'w' for the height instead of 'h', because the input face has 1:1 aspect ratio.
cv::Mat warped = cv::Mat(desiredFaceHeight, desiredFaceWidth, CV_8U, cv::Scalar(128)); // Clear the output image to a default grey.
warpAffine(gray, warped, rot_mat, warped.size());
//imshow("warped", warped);
// Give the image a standard brightness and contrast, in case it was too dark or had low contrast.
if (!doLeftAndRightSeparately) {
// Do it on the whole face.
equalizeHist(warped, warped);
}
else {
// Do it seperately for the left and right sides of the face.
equalizeLeftAndRightHalves(warped);
}
//imshow("equalized", warped);
// Use the "Bilateral Filter" to reduce pixel noise by smoothing the image, but keeping the sharp edges in the face.
cv::Mat filtered = cv::Mat(warped.size(), CV_8U);
bilateralFilter(warped, filtered, 0, 20.0, 2.0);
//imshow("filtered", filtered);
// Filter out the corners of the face, since we mainly just care about the middle parts.
// Draw a filled ellipse in the middle of the face-sized image.
cv::Mat mask = cv::Mat(warped.size(), CV_8U, cv::Scalar(0)); // Start with an empty mask.
cv::Point faceCenter = cv::Point(desiredFaceWidth / 2, cvRound(desiredFaceHeight * FACE_ELLIPSE_CY));
cv::Size size = cv::Size(cvRound(desiredFaceWidth * FACE_ELLIPSE_W), cvRound(desiredFaceHeight * FACE_ELLIPSE_H));
ellipse(mask, faceCenter, size, 0, 0, 360, cv::Scalar(255), CV_FILLED);
//imshow("mask", mask);
// Use the mask, to remove outside pixels.
cv::Mat dstImg = cv::Mat(warped.size(), CV_8U, cv::Scalar(128)); // Clear the output image to a default gray.
/*
namedWindow("filtered");
imshow("filtered", filtered);
namedWindow("dstImg");
imshow("dstImg", dstImg);
namedWindow("mask");
imshow("mask", mask);
*/
// Apply the elliptical mask on the face.
filtered.copyTo(dstImg, mask); // Copies non-masked pixels from filtered to dstImg.
//imshow("dstImg", dstImg);
/*if (leftEye.x >= 0 && rightEye.x >= 0)
{
eyeRegion.x = 0;
eyeRegion.y = searchedLeftEye.y;
eyeRegion.width = preprocessedFace.rows;
eyeRegion.height = 30;
cv::rectangle(frame, eyeRegion, CV_RGB(255, 255, 0), 3);
}*/
if (needResults)
{
writeResults(srcImg, gray, dstImg);
}
return dstImg;
}
/*
else {
// Since no eyes were found, just do a generic image resize.
resize(gray, tmpImg, Size(w,h));
}
*/
}
return cv::Mat();
}
void CV_BilateralFilterTest::run_func()
{
bilateralFilter(_src, _parallel_dst, _d, _sigma_color, _sigma_space);
}
