int _smoothmain(int w, int h, int color, imgtype *imgdata)
{
fill_img_info(w, h, color, imgdata);
meanshift();
writeoutimg();
free_mem();
printf("done.\n");
return 0;
}
/*! @brief Mean Shift (MS) and Centroid Finding Function
*
* This function runs through a binary image and finds the blobs
* centroids using the mean shift method
*
* @param binaryImage the binary either (0-255) or (0-1) input image to be MS processed
* @param blobs 2D list of 2D points
* @param blobCentres list of 2D points representing centroids of each blob
*/
void ossimSDFilter::findBlobsMS(cv::Mat& binaryImage, vector< vector< cv::Point2i > >& blobs, vector< cv::Point2i >& blobCentres)
{
//Clear all blobs
blobs.clear();
blobCentres.clear();
//Ensure input image is 8CU1
binaryImage.convertTo(binaryImage,CV_8UC1);
//Find nonzero coordinates in binary image
find(binaryImage, blobCentres);
int rows = 2;
int cols = blobCentres.size();
std::vector<double> flatBlobs;
std::vector<double> labelledClusters;
std::vector<double> meansFinal;
flatBlobs.resize(rows*cols);
labelledClusters.resize(rows*cols);
int counter = 0;
for(std::vector<cv::Point2i>::iterator it = blobCentres.begin(); it != blobCentres.end(); ++it, counter++)
{
flatBlobs[counter*rows + 0] = it->x;
flatBlobs[counter*rows + 1] = it->y;
}
meanshift(flatBlobs, rows, cols, bw, descendRate, iterMax, labelledClusters, meansFinal);
cv::Mat test = cv::Mat(binaryImage.rows, binaryImage.cols, CV_8UC1, cv::Scalar::all(0));
blobCentres.clear();
for(int i = 0; i < counter; i++)
blobCentres.push_back(cv::Point2i(floor(meansFinal[i*rows]+0.5),floor(meansFinal[i*rows+1]+0.5)));
paintCentres(test, blobCentres);
blobCentres.clear();
find(test, blobCentres);
}
int main(int argc, char** argv) {
FILE* log = fopen("./logs/log", "a");
//default configuration
char* input_file_path;
char output_filename[100];
double sigma_meanshift = 0.4;
double sigma_affmat = 0.005;
double diffuse_t = 5000.0;
int nev = 20;
Mat source_img;
Mat output_img;
time_t start_time;
time_t current_time;
time_t end_time;
//parse command line arguments
for(int i = 0; i < argc; i++) {
if(strcmp(argv[i],"-f")==0) {
i++;
input_file_path = argv[i];
printf("-f: %s\n", input_file_path);
}
if(strcmp(argv[i],"-sa")==0) {
i++;
sigma_affmat = atof(argv[i]);
printf("-sa: %f\n", sigma_affmat);
}
if(strcmp(argv[i],"-sm")==0) {
i++;
sigma_meanshift = atof(argv[i]);
printf("-sm: %f\n", sigma_meanshift);
}
if(strcmp(argv[i],"-t")==0) {
i++;
diffuse_t = atof(argv[i]);
printf("-t: %f\n", diffuse_t);
}
if(strcmp(argv[i],"-n")==0) {
i++;
nev = atoi(argv[i]);
printf("-n: %d\n", nev);
}
}
if(access(input_file_path, F_OK) == -1) {
printf("Invalid input file path: %s\n", input_file_path);
arg_error();
}
printf("Args parsing completed!\n");
printf("%s %s -sa %f -sm %f -t %f -n %d\n", argv[0], input_file_path, sigma_affmat, sigma_meanshift, diffuse_t, nev);
start_time = time(NULL);
/* Load image data */
source_img = loadDoubleImage(input_file_path);
output_img = Mat::zeros(source_img.size(), CV_8UC3);
//printMatdouble(source_img);
/* Log record */
fprintf(log,"[%lu-%s]\nnev=%d, sigma_affmat=%f, sigma_meanshift=%f, diffuse_t=%f", start_time, basename(input_file_path), nev, sigma_affmat, sigma_meanshift, diffuse_t);
sprintf(output_filename, "./out/%lu_sa%d_sm%d_nev%d_difft%d-%s", start_time, (int)(sigma_affmat*1000),(int)(sigma_meanshift*1000), nev,(int)diffuse_t, basename(input_file_path));
/* Compute affinity matrix */
CSCMat affinityMat = getCSCAffinityMatrix(source_img, sigma_affmat);
//printCSCMatrix(affinityMat);
/* Compute laplacian matrix */
CSCMat laplacianMat = getCSCLaplacianSym(affinityMat);
//printCSCMatrix(laplacianMat);
current_time = time(NULL);
/* Compute eigen vectors*/
vector<vector<double > > dataPoints = spectralDecomposition(laplacianMat, nev, diffuse_t);
//print2DVecArray(dataPoints);
/* Log record */
fprintf(log,"Spectral decomposition time: %lu\n", time(NULL) - current_time);
current_time = time(NULL);
/*Compute clusters*/
vector<vector<double > > datameans = meanshift(dataPoints, sigma_meanshift);
//vecPrint2DArray(datameans);
/* Log record */
fprintf(log,"Clustering time: %lu\n", time(NULL) - current_time);
vector<vector<int > > clusters = cluster2(datameans, 1.0);
/* Colorize segments*/
visualiseSegments(output_img, clusters);
/* Log record */
fprintf(log, "Total time: %lu\n\n", time(NULL) - start_time);
imwrite(output_filename, output_img);
fclose(log);
printf("Done!\n");
waitKey(0);
return 0;
}
