Detecting Contour of bubble

I am doing image processing on pictures of a droplet. However on quite a few images, the contour of the droplet is not detected when I try to read the image from left to right. I want all the white boundary to be detected from left to right. I have attached the necessary images too. Kindly help me out in this regard.

[G.N.V.SUDEEPTA VAISHNAVI]

This is the original image.

[Image Analyst]

No images are attached. I'm attaching one I got from KSSV.

I have no idea what you mean. What/where is the "contour"?

What do you mean by "I want all the white boundary to be detected from left to right."? To get boundaries (perimeter of white region) you can use bwperim() or bwboundaries().

Answer 1

[Image Analyst]

Again, not sure what you want since the description is not precise, but here are some options:

grayImage = imread('Vaishnavi.png');
binaryImage = grayImage(:, :, 1) > 128; % Binarize

% Method 1
perimImage = bwperim(binaryImage); % Get perimeter
rgbImage = imoverlay(binaryImage, perimImage, 'r'); % Show perimeter in the overlay.
imshow(rgbImage);

% Alternatively you can use Method 2
boundaries = bwboundaries(binaryImage); % Get a list of (x,y) coordinates of the perimeter.
x = boundaries{1}(:, 2);
y = boundaries{1}(:, 1);
hold on;
plot(x, y, 'r-', 'LineWidth', 3)

[G.N.V.SUDEEPTA VAISHNAVI]

The alternative method 2 worked out. Thanks for the answer.

I have another doubt with respect to this method. I need to find out the points where the actual bubble is present. In more specific terms, I need to find out where there is a change in the slope when I differentiate (Remove the rectanglular portion of the image). Any idea how can I do it?

I have tried the Roger Stafford forumula that you mentioned in one of the threads. The link to the thread is attached below.

https://in.mathworks.com/matlabcentral/answers/57194-how-to-find-the-sharp-turn-in-a-2d-line-curve#answer_69185

[Image Analyst]

To find all the coordinates of the white points in a binary image you can do this:

[whiteRows, whiteColumns] = find(binaryImage);  % or [y, x] =    (NOT [x, y]!!!)

Usually though you don't do that. The image itself is a map of what pixels are white.

I'm not quite sure what the rectangular portion is. Is it that white "tail" that shoots mostly upwards from the 2 O'Clock location on the round blob? If so there is a variety of things you can do to remove it depending on how much or little you're willing to alter the shape of the round part of the blob. The easiest thing to do is to just do a morphological opening with imopen().

You can also try to identify sharp turns in the (x,y) using Roger's method, or now there is a built-in findchangepts() function that you can try. Another option is to use fitpolynomialRANSAC() (if you have the Computer Vision Toolbox), or use movmedian() to identify points far away from the median. You might also try rmoutliers(). Like I said there are several ways it could be done, all with results that are mostly the same but may have small differences.

[G.N.V.SUDEEPTA VAISHNAVI]

Yes, it is the tail that shoots upwards.

I will definitely look into the methods you suggested.

Answer 2

[KSSV]

I = imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/814649/image.jpeg') ;
% Remove boundary/ crop 
I = I(10:end-10,10:end-10,:) ; 
I1 = imbinarize(rgb2gray(I)) ; 
% option 1
c = contour(I1) ; 
% option 2 
[y,x] = find(I1) ; 
idx = boundary(x,y) ; 
plot(x(idx),y(idx),'k')

Answer 3

[yanqi liu]

clc; clear all; close all;
I = imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/814649/image.jpeg') ;
I = imcrop(I, [20 20 size(I,2)-40 size(I,1)-40]);
I1 = imbinarize(rgb2gray(I)) ; 
figure; imshow(I1,[]); hold on;
[M,c] = contour(bwperim(I1)) ; 
% set line 
set(c, 'Color', 'r')
set(c, 'Color', 'r', 'LineWidth', 2)

% all the white boundary to be detected from left to right
be = bwperim(I1); [r,c] = find(be);
[~,ind] = min(c);
contour2 = bwtraceboundary(be,[r(ind) c(ind)],'W');
figure; imshow(I1,[]); hold on;
plot(contour2(:,2),contour2(:,1),'c','LineWidth',2)

[Image Analyst]

Like I said, you can use bwboundaries(). It's advantage is that you do not need to specify a starting point like you do with bwtraceboundary().