morphological operations in matlab

I am having an image uploaded at: http://tinypic.com/view.php?pic=1h2n8g&s=6 After doing some processing on it I found the following results uploaded at: http://tinypic.com/view.php?pic=1jqffa&s=6 and http://tinypic.com/view.php?pic=2l6sk1&s=6

The code used for processing is:

    img = imread('newroad1.jpg'); 
    hsv = rgb2hsv(img); 
    I1 = hsv(:,:,3); 
    figure,imshow(I1) 
    se = strel('square',5); 
    mgrad = imdilate(I1, se) - imerode(I1, se);
    figure,imshow(mgrad) 
    I=im2bw(mgrad,0.1);

I want to get the following results uploaded at: http://tinypic.com/view.php?pic=orp2rn&s=6 and http://tinypic.com/view.php?pic=14dga2s&s=6

What morphological operations should I use so that I can get the above mentioned results.

Answer 1

[Image Analyst]

Those would be the 'skel' and 'spur' options of bwmorph().

[azan]

I have used these commands. By using the 'skel' command I get the result uploaded at: http://tinypic.com/view.php?pic=359iwes&s=6 and by using the 'spur' command I get the result uploaded at: http://tinypic.com/view.php?pic=359iwes&s=6 These are not the exact results which I require. I want the white lanes to be extracted at the end http://tinypic.com/view.php?pic=14dga2s&s=6 along with the correct results.

[Image Analyst]

That is not a skeleton. Evidently you did not use the "inf" option with the skel.

[azan]

By using this I got the result http://tinypic.com/view.php?pic=nq2pvc&s=6

[Image Analyst]

OK, you're getting there. You just need to zero out the edges of the top image and the bottom image because they are giving spurious edges.

skelImage(row1:row2, col1:col2) = 0;

[azan]

I have done like this:

      img = imread('road1.jpg'); 
      hsv = rgb2hsv(img); 
      I1 = hsv(:,:,3);
      se = strel('square',5); 
      mgrad = imdilate(I1, se) - imerode(I1, se);
      I=im2bw(mgrad,0.1); 
       BW1=bwmorph(I,'skel',inf);
      BW1(1:2,1:2) = 0;
      figure,imshow(BW1)
      BW2=bwmorph(BW1,'spur');
      figure,imshow(BW2)

Same procedure is done for second image.But there is no significant change in result. Guide me

[Image Analyst]

What are you thinking when you do this:

BW1(1:2,1:2) = 0;

??? Are you thinking that erasing the upper left corner (a 2 by 2 patch) is going to erase the edges of the images - those lines that you do not want? Those lines exist way outside the 1:2,1:2 upper left box! After you use imshow(), call "axis on" to see what rows and columns you need to zero out. You need to zero out all of lines 150, 200, and 250 (or something like that) and all of column 800 (or something around there).

BW1(148:152,:) = false; % use the actual numbers.
BW1(:, 798:802) = false; % use the actual numbers.

[azan]

ok Sir now I got it..I am having images so I can fix the dimensions to erase edges. What if I am having a video. How to achieve the same result for the video.

[Image Analyst]

Just do it a frame at a time. Of course there are fancier ways involving time projection where you predict positions based on past frames but that's a lot more advanced. Just start with a frame at a time.

[azan]

ok sir I will try it and then will let you know if I will be having some problem.