How to call LSD (LineSegmentDetector) from a c language program? - c

Question

i'm using LSD to detect straight lines in an image, the code that i have downloaded contains a Minimal example of calling LSD but it's static (i.e it outputs only the value in the main function) i want to apply the code on a video, that's the minimal example that outputs static results.
#include <stdio.h>
#include "lsd.h"
int main(void)
{
image_double image;
ntuple_list out;
unsigned int x,y,i,j;
unsigned int X = 512; /* x image size */
unsigned int Y = 512; /* y image size */
/* create a simple image: left half black, right half gray */
image = new_image_double(X,Y);
for(x=0;x<X;x++)
for(y=0;y<Y;y++)
image->data[ x + y * image->xsize ] = x<X/2 ? 0.0 : 64.0; /* image(x,y) */
IplImage* imgInTmp = cvLoadImage("C:\Documents and Settings\Eslam farag\My Documents\Visual Studio 2008\Projects\line\hand.JPEG", 0);
/* call LSD */
out = lsd(image);
/* print output */
printf("%u line segments found:\n",out->size);
for(i=0;i<out->size;i++)
{
for(j=0;j<out->dim;j++)
printf("%f ",out->values[ i * out->dim + j ]);
printf("\n");
}
/* free memory */
free_image_double(image);
free_ntuple_list(out);
return 0;
}
if anyone can help me to apply the code on video i will be pleased.thanks
best regards,

Answer 1

Since I couldn't find a complete example, I'm sharing a code I wrote that uses OpenCV to load a video file from the disk and perform some image processing on it.
The application takes a filename as input (on the cmd line) and converts each frame of the video to it's grayscale equivalent using OpenCV built-in function cvCvtColor() to do this.
I added some comments on the code to help you understand the basic tasks.
read_video.cpp:
#include <stdio.h>
#include <highgui.h>
#include <cv.h>
int main(int argc, char* argv[])
{
cvNamedWindow("video", CV_WINDOW_AUTOSIZE);
CvCapture *capture = cvCaptureFromAVI(argv[1]);
if(!capture)
{
printf("!!! cvCaptureFromAVI failed (file not found?)\n");
return -1;
}
IplImage* frame;
char key = 0;
while (key != 'q') // Loop for querying video frames. Pressing Q will quit
{
frame = cvQueryFrame( capture );
if( !frame )
{
printf("!!! cvQueryFrame failed\n");
break;
}
/* Let's do a grayscale conversion just 4 fun */
// A grayscale image has only one channel, and most probably the original
// video works with 3 channels (RGB). So, for the conversion to work, we
// need to allocate an image with only 1 channel to store the result of
// this operation.
IplImage* gray_frame = 0;
gray_frame = cvCreateImage(cvSize(frame->width, frame->height), frame->depth, 1);
if (!gray_frame)
{
printf("!!! cvCreateImage failed!\n" );
return -1;
}
cvCvtColor(frame, gray_frame, CV_RGB2GRAY); // The conversion itself
// Display processed frame on window
cvShowImage("video", gray_frame);
// Release allocated resources
cvReleaseImage(&gray_frame);
key = cvWaitKey(33);
}
cvReleaseCapture(&capture);
cvDestroyWindow("video");
}
Compiled with:
g++ read_video.cpp -o read `pkg-config --cflags --libs opencv`
If you want to know how to iterate through the pixels of the frame to do your custom processing, you need to check the following answer because it shows how to do a manual grayscale conversion. There you go: OpenCV cvSet2d.....what does this do

Answer 2

here is example of the code using LSD with opencv
#include "lsd.h"
void Test_LSD(IplImage* img)
{
IplImage* grey = cvCreateImage(cvGetSize(img), IPL_DEPTH_8U, 1);
cvCvtColor(img, grey, CV_BGR2GRAY);
image_double image;
ntuple_list out;
unsigned int x,y,i,j;
image = new_image_double(img->width,img->height);
for(x=0;x<grey->width;x++)
for(y=0;y<grey->height;y++)
{
CvScalar s= cvGet2D(grey,y,x);
double pix= s.val[0];
image->data[ x + y * image->xsize ]= pix; /* image(x,y) */
}
/* call LSD */
out = lsd(image);
//out= lsd_scale(image,1);
/* print output */
printf("%u line segments found:\n",out->size);
vector<Line> vec;
for(i=0;i<out->size;i++)
{
//for(j=0;j<out->dim;j++)
{
//printf("%f ",out->values[ i * out->dim + j ]);
Line line;
line.x1= out->values[ i * out->dim + 0];
line.y1= out->values[ i * out->dim + 1];
line.x2= out->values[ i * out->dim + 2];
line.y2= out->values[ i * out->dim + 3];
vec.push_back(line);
}
//printf("\n");
}
IplImage* black= cvCreateImage(cvGetSize(img), IPL_DEPTH_8U, 3);
cvZero(black);
draw_lines(vec,black);
/*cvNamedWindow("img", 0);
cvShowImage("img", img);*/
cvSaveImage("lines_detect.png",black/*img*/);
/* free memory */
free_image_double(image);
free_ntuple_list(out);
}
or this way
IplImage* get_lines(IplImage* img,vector<Line>& vec_lines)
{
//to grey
//IplImage* grey = cvCreateImage(cvGetSize(img), IPL_DEPTH_8U, 1);
//cvCvtColor(img, grey, CV_BGR2GRAY);
image_double image;
ntuple_list out;
unsigned int x,y,i,j;
image = new_image_double(img->width,img->height);
for(x=0;x</*grey*/img->width;x++)
for(y=0;y</*grey*/img->height;y++)
{
CvScalar s= cvGet2D(/*grey*/img,y,x);
double pix= s.val[0];
image->data[ x + y * image->xsize ]= pix;
}
/* call LSD */
out = lsd(image);
//out= lsd_scale(image,1);
/* print output */
//printf("%u line segments found:\n",out->size);
//vector<Line> vec;
for(i=0;i<out->size;i++)
{
//for(j=0;j<out->dim;j++)
{
//printf("%f ",out->values[ i * out->dim + j ]);
Line line;
line.x1= out->values[ i * out->dim + 0];
line.y1= out->values[ i * out->dim + 1];
line.x2= out->values[ i * out->dim + 2];
line.y2= out->values[ i * out->dim + 3];
/*vec*/vec_lines.push_back(line);
}
//printf("\n");
}
IplImage* black= cvCreateImage(cvGetSize(img), IPL_DEPTH_8U, 1);
cvZero(black);
for(int i=0;i<vec_lines.size();++i)
{
//if(vec[i].x1==vec[i].x2||vec[i].y1==vec[i].y2)
cvLine(black,cvPoint(vec_lines[i].x1,vec_lines[i].y1),cvPoint(vec_lines[i].x2,vec_lines[i].y2),CV_RGB(255,255,255),1, CV_AA);
}
/*cvNamedWindow("img", 0);
cvShowImage("img", img);*/
//cvSaveImage("lines_detect.png",black/*img*/);
/* free memory */
//cvReleaseImage(&grey);
free_image_double(image);
free_ntuple_list(out);
return black;
}