what is exactly MultiScale hough transform (setting inputs and compare runtime between SHT & MHT)

asked 2017-10-20 04:37:49 -0500

saeid_masumi gravatar image

updated 2017-10-22 06:41:56 -0500

I'm working on Line detection using OpenCV and when I saw Multi-Scale hough transform I read some papers about this algorithm and I see Multi-Scale algorithm used to large image.The original paper said that when we working with a large image the SHT parameters can increase dramatically. So I wrote a program to test (proof) it. I resize my image to be a large one but I see that SHT( the standard form of HT) is so faster than MHT

 Mat image = imread("/home/saeed/Desktop/desktop/s.png",IMREAD_GRAYSCALE);
 resize(image,image,Size(1024,1024);

 //MHT form
 HoughLines(img,contours,1,CV_PI/180,100,0.03,1); //this line give me 2.45 sec

//SHT form 
HoughLines(img,contours,1,CV_PI/180,100);    //this line give me 0.01sec

And so I think if I work on a real-time application the SHT method is a better choice (is this correct?)

And my second question. How could we set MHT inputs (SRN and STN)? I guess we should randomly select them and if the output is good then keep them!! I know these parameters for refining results so I choose every possibilities of this two parameters but MHT and SHT give me same results except in detecting lines that them angle and rho is very close.

edit retag flag offensive close merge delete