MATLAB Answers

How to identify contiguous regions between two thresholds?

조회 수: 9(최근 30일)
K E 6 Feb 2013
If I have a variable z at mapped locations x and y, is there way to obtain the x-y locations of the contour(s) enclosing any continuous region in which z is between two thresholds minZ and maxZ? I do not have the Image Processing toolbox but this task may be similar to image segmentation. Below I use contourf to obtain the x,y locations of the region's outer edge, but perhaps there is a better approach.
% Make a fake dataset
z = peaks;
z = z(1:25, :); % Let's make different numbers of x and y elements for clarity
y = 1:size(z, 1);
x = 1:size(z, 2);
% Desired thresholds for the region
minZ = 2;
maxZ = 3;
[cc, hh] = contourf('v6', x, y, z, [minZ maxZ]); % Must use version 6
for iContour = 1:length(hh)
% Obtain the x and y location of the region's outer edges
% Does not account for cutouts inside the region
if get(hh(iContour), 'cdata') == minZ
xOutline = get(hh(iContour), 'xdata');
yOutline = get(hh(iContour), 'ydata');
% Find the points inside the region
isInside = inpolygon(x, y, xOutline, yOutline);
end
end
delete(hh); % Get rid of contours

댓글을 달려면 로그인하십시오.

채택된 답변

ChristianW 6 Feb 2013
편집: ChristianW 6 Feb 2013
[C,h] = contour(...)
The ContourMatrix C contains directly the infomation you are looking for. Just look up ContourMatrix in the Matlab help.
Here is an example plot for the first line:
Z = peaks;
subplot(211)
[C,h] = contour(Z,[2 3]); title('contour'); ax_c = gca;
subplot(212)
hold on; title('proof')
plot( C(1,(1:C(2,1))+1) , C(2,(1:C(2,1))+1) )
set(gca,'xlim',get(ax_c,'xlim'),'ylim',get(ax_c,'ylim'))
댓글 수: 1표시숨기기 없음
K E 7 Feb 2013
Great, now I not dependent on the v6 version of contour. Thanks!

댓글을 달려면 로그인하십시오.

추가 답변(1개)

Image Analyst 6 Feb 2013
Explain what "identify" means. You can of course just make a binary image of where that criteria is true like this:
zIsInRange = (z >= minZ) & (z <= maxZ);
But this binary image could have several blobs on it. That's why I ask you what "identify" means to you. Normally if you had the Image Processing Toolbox, you'd call bwlabel() or bwconncomp().
댓글 수: 3표시숨기기 이전 댓글 수: 2
K E 6 Feb 2013
Yes, I can do what I want with the 'v6' version of contourf. But that doesn't seem so stable!

댓글을 달려면 로그인하십시오.

Community Treasure Hunt

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

Start Hunting!

Translated by