Detecting coordinates of line edges

12 次查看（过去 30 天）
Pankaj
Pankaj 2020-11-30
评论： Pankaj 2020-12-1
I have a matrix containing 1s and 0s. The elements of this matrix are pixels of an image, where 1&0 represents black and white. The 1s form non-intersecting line segmennts. I want to know pixel coordinates of line edges. Following is an example matrix-
binaryImage= [0 1 0 0;...
0 0 1 0; ...
0 0 0 1; ...
0 0 0 0]
The desired output is-
[1, 2], [3, 4]
Note that there could be multiple line segments. Is Hough transform useful here?
Thanks
--------------------
Update: I am also attaching another example matrix and corrosponding `imagesc` for better visualization of problem.
  3 个评论
显示 1更早的评论
sushanth govinahallisathyanarayana
sushanth govinahallisathyanarayana 2020-11-30
Yes, Hough might be useful, since you can describe a line with 2 points, however, you need an approximate idea of the scales at which you want to look, and how the lines may be oriented.

请先登录，再进行评论。

请先登录，再回答此问题。

采纳的回答

Image Analyst
Image Analyst 2020-11-30
Use bwmorph() to get an image of just the endpoints, then use find() to get the rows and columns:
endPointsImage = bwmorph(binaryImage, 'endpoints');
[rows, columns] = find(endPointsImage);
rows and columns will be a list of the rows and columns of each endpoints. They are synced up so for index N, rows(N) has the row of the Nth endpoint and columns(N) has the column of the Nth endpoint.
  6 个评论
显示 4更早的评论
Image Analyst
Image Analyst 2020-11-30
Yes. But you have to label the binary image so as to know which endpoint belongs to which blob. Keep in mind that is the blob is shaped like an X then it would have 4 endpoints, or 3 if it were shaped like a Y. So you could have different numbers of endpoints for each blob. So you'd do something like
[labeledImage, numRegions] = bwlabel(binaryImage);
endPointsImage = bwmorph(binaryImage, 'endpoints');
[rows, columns] = find(endPointsImage);
endpoints = cell(numRegions, 1; % At least 1 endpoint for each region
endpointsPerLabel = zeros(numRegions, 1);
for k = 1 : length(rows)
blobLabel = labeledImage(rows(k), columns(k));
endpointsPerLabel(blobLabel) = endpointsPerLabel(blobLabel) + 1;
x = columns(k);
y = rows(k);
endpoints{blobLabel, endpointsPerLabel(blobLabel)} = [x, y];
end
If this doesn't work, re-read my last comment, especially the last sentence.
Pankaj
Pankaj 2020-12-1
Thanks!! This is exactly what I wanted. I have also modified my question, slightly.

请先登录，再进行评论。

更多回答（1 个）

KSSV
KSSV 2020-11-30
A= [0 1 0 0;...
0 0 1 0; ...
0 0 0 1; ...
0 0 0 0] ;
idx = find(A) ;
[i,j] = ind2sub(size(A),idx) ;
[i j]
  3 个评论
显示 1更早的评论
KSSV
KSSV 2020-11-30
The above ives the locations of 1 in the matrix A.
Image Analyst
Image Analyst 2020-11-30
Pankaj, the code I gave you in my Answer gives the endpoints only. Did you overlook it?

请先登录，再进行评论。

类别

在 Help Center 和 File Exchange 中查找有关 Hough Transform 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by