fastbf_main.c
/**
* @file fastbf_main.c
* @brief Main executable file.
*
* @author PRAVIN NAIR <sreehari1390@gmail.com>
* ANMOL POPLI <anmol.ap020@gmail.com>
*/
#include "headersreq.h"
int main(int argc, char* argv[]) {
int cores;
#ifdef __linux__
#ifdef _OPENMP
cores = core_affinity();
#endif
#else
#ifdef _OPENMP
cores = omp_get_num_procs(); /** \brief Number of Hyperthreads for MP */
#else
cores=0;
#endif
#endif
if (cores>8) cores=8;
/* Check if there is a right call for algorithm */
if (argc>9){
printf("Too many arguments. \nSyntax is: FBF input sigmas sigmar output eps sigmaref noise_indicator sigman \n");
return EXIT_FAILURE;
}
if (argc<6){
printf("Not enough arguments. \nSyntax is: FBF input sigmas sigmar output eps \n");
return EXIT_FAILURE;
}
program_params params; /** \brief Program parameters */
/* Declaration parameters */
double *input_image, *output_image;
int columns,rows;
struct timeval start,end;
int i,j,sigmas;
double sigmar,eps,sigmaref;
/* Read image, image filename input as argv[1] */
input_image = (double *)read_image(&columns, &rows, argv[1], IMAGEIO_DOUBLE | IMAGEIO_PLANAR | IMAGEIO_GRAYSCALE);
sscanf(argv[2], "%d", &sigmas); /* SigmaS input as argv[2] */
sscanf(argv[3], "%lf", &sigmar); /* SigmaR input as argv[3] */
sscanf(argv[5], "%lf", &eps); /* parameter epsilon input as argv[5] */
if (argc>=7)
sscanf(argv[6], "%lf", &sigmaref); /* Parameter to control stretching of the difference images as argv[6] */
else
sigmaref=32;
double** image = alloc_array(rows,columns); /** \brief Matrix to store input image */
double** image_out = alloc_array(rows,columns); /** \brief Matrix to store output image */
for (i=0; i<rows; i++) {
for (j=0; j<columns; j++) {
image[i][j] = input_image[i*columns+j]*255.0;
}
}
/* Check if noise is added */
bool addnoise;
if (argc>7)
addnoise = (bool)atof(argv[7]);
else
addnoise = 0;
if (addnoise==1) {
double sigman;
if (argc==9)
sscanf(argv[8], "%lf", &sigman); /*Noise standard deviation as argv[8] */
else{
printf("Specify standard deviation of noise. \nSyntax is: FBF input sigmas sigmar output eps sigmaref noise_indicator sigman \n");
return EXIT_FAILURE;
}
if (addgaussiannoise(image,rows,columns,sigman)!=EXIT_SUCCESS) {/* adding gaussian noise */
printf("Adding gaussian noise failed \n");
return EXIT_FAILURE;
}
/* Clipping the noisy image */
for (i=0; i<rows; i++) {
for (j=0; j<columns; j++) {
if (image[i][j] < 0.0) image[i][j]=0.0;
if (image[i][j] > 255.0) image[i][j]=255.0;
}
}
}
gettimeofday(&start, NULL);
/* Shiftable Bilateral Filter applied to image and result stored in image_out */
if (shiftableBF(rows,columns,sigmas,sigmar,image,image_out,cores,¶ms,eps)!=EXIT_SUCCESS){
printf("Fast bilateral filter algorithm failed \n");
return EXIT_FAILURE;
}
gettimeofday(&end, NULL);
printf("Number of DFT coefficients used for approximating range kernel is %d \n",params.K);
printf("Execution time: %f ms\n", ((end.tv_sec-start.tv_sec)*1000.0) + ((end.tv_usec-start.tv_usec)/1000.0));
/* Write image, image filename input as argv[4] */
output_image = (double*) calloc(rows*columns,sizeof(double));
for (i=0; i<rows; i++) {
for (j=0; j<columns; j++) {
output_image[i*columns+j] = image_out[i][j]/255.0;
}
}
if (write_image(output_image, columns, rows, argv[4], IMAGEIO_DOUBLE | IMAGEIO_PLANAR | IMAGEIO_GRAYSCALE, 100)!=1){
printf("Writing image failed \n");
return EXIT_FAILURE;
}
dealloc_array_fl(image_out,rows);
/* No noise condition */
if (addnoise == 0){
/* Computing the difference image */
double* diff_image = (double*) calloc(rows*columns,sizeof(double));
for (i=0; i<rows*columns; i++)
diff_image[i] = (input_image[i]-output_image[i]);
/* Stretching and clipping the difference image */
double std=calculatestd(diff_image,rows*columns);
for (i=0; i<rows*columns; i++) {
diff_image[i] = (128.0/255.0)+(diff_image[i]*sigmaref/(std*255));
if (diff_image[i]<0.0) diff_image[i]=0.0;
if (diff_image[i]>1.0) diff_image[i]=1.0;
}
/* Writing the difference image */
if (write_image(diff_image, columns, rows, "difference.png", IMAGEIO_DOUBLE | IMAGEIO_PLANAR | IMAGEIO_GRAYSCALE, 100)!=1){
printf("Writing image failed \n");
return EXIT_FAILURE;
}
dealloc_array_fl(image,rows);
free(diff_image);
free(output_image);
}
/* Noisy conditions */
else{
double* noisy_image = (double*) calloc(rows*columns,sizeof(double));
for (i=0; i<rows; i++) {
for (j=0; j<columns; j++) {
noisy_image[i*columns+j] = image[i][j]/255.0;
}
}
dealloc_array_fl(image,rows);
/* Generating the difference images */
double* diff_image = (double*) calloc(rows*columns,sizeof(double));
double* diff_noisyimage = (double*) calloc(rows*columns,sizeof(double));
for (i=0; i<rows*columns; i++){
diff_image[i] = (input_image[i]-output_image[i]);
diff_noisyimage[i] = (noisy_image[i]-output_image[i]);
}
/* Stretching and clipping the difference images */
double std=calculatestd(diff_image,rows*columns);
double stdnoisy=calculatestd(diff_noisyimage,rows*columns);
for (i=0; i<rows*columns; i++) {
diff_image[i] = (128.0/255.0)+(diff_image[i]*sigmaref/(std*255));
if (diff_image[i]<0.0) diff_image[i]=0.0;
if (diff_image[i]>1.0) diff_image[i]=1.0;
diff_noisyimage[i] = (128.0/255.0)+(diff_noisyimage[i]*sigmaref/(stdnoisy*255));
if (diff_noisyimage[i]<0.0) diff_noisyimage[i]=0.0;
if (diff_noisyimage[i]>1.0) diff_noisyimage[i]=1.0;
}
/* Writing the noisy image */
if (write_image(noisy_image, columns, rows, "noisy.png", IMAGEIO_DOUBLE | IMAGEIO_PLANAR | IMAGEIO_GRAYSCALE, 100)!=1){
printf("Writing image failed \n");
return EXIT_FAILURE;
}
free(noisy_image);
/* Writing the difference image : original - filtered */
if (write_image(diff_image, columns, rows, "diff_noiseless.png", IMAGEIO_DOUBLE | IMAGEIO_PLANAR | IMAGEIO_GRAYSCALE, 100)!=1){
printf("Writing image failed \n");
return EXIT_FAILURE;
}
free(diff_image);
/* Writing the difference image : noisy - filtered */
if (write_image(diff_noisyimage, columns, rows, "diff_noisy.png", IMAGEIO_DOUBLE | IMAGEIO_PLANAR | IMAGEIO_GRAYSCALE, 100)!=1){
printf("Writing image failed \n");
return EXIT_FAILURE;
}
free(diff_noisyimage);
free(output_image);
}
free(input_image);
#ifdef __linux__
/** \brief Arrays to store range kernels for plotting */
double* intensity = (double*) calloc(2*params.T+1,sizeof(double));
double* rkernel = (double*) calloc(2*params.T+1,sizeof(double));
double* rkernel_approx = (double*) calloc(2*params.T+1,sizeof(double));
for (i=0; i<=2*params.T; i++) {
intensity[i] = i-params.T;
rkernel[i] = exp(-0.5*intensity[i]*intensity[i]/(sigmar*sigmar));
}
/* Plot target and approximate range kernels */
double w0 = (2*M_PI)/(2*params.T+1);
for (i=0; i<=2*params.T; i++) {
rkernel_approx[i] = 0;
for (j=0; j<params.K; j++) {
rkernel_approx[i] += params.coeff[j]*cos(j*w0*intensity[i]);
}
}
gnuplot_ctrl *h1;
h1 = gnuplot_init() ;
gnuplot_cmd(h1, "set terminal png");
gnuplot_cmd(h1,"set title \"Range Kernel Plot\"");
gnuplot_setstyle(h1,"lines");
gnuplot_cmd(h1, "set output \"rangekernel_plot.png\"");
gnuplot_plot_xy(h1, intensity, rkernel, 2*params.T+1, "Target Gaussian");
gnuplot_cmd(h1, "set output \"rangekernel_plot.png\"");
gnuplot_plot_xy(h1, intensity, rkernel_approx, 2*params.T+1, "Fourier Approximation");
gnuplot_close(h1);
free(intensity);
free(rkernel);
free(rkernel_approx);
#endif // __linux__
return EXIT_SUCCESS;
}
