How to track the boundary of a drop over successive frames?

Hi,

I am trying to track the boundary of drop that is spreading over a liquid. The drop is circular initially but get deformed over subsequent frames. A few frames from the video are attached in this Hyperlink. I tried using regionpros and hough transform in Matlab but neither of them worked. The major problem is due to the background which is a grid placed in order to enhance the contrast of the edge (without it the edge is not very clear) while recording the video. How can I measure the area/ perimeter of the region enclosed by the interface?

Any help on how to proceed would be really appreciated.

Thanks,

Rajesh

Answer 1

[Image Analyst]

Try thresholding and calling bwboundaries. If that doesn't work because the boundary is too faint, then try to seal it by using imdilate() before thresholding:

img = imdilate(grayImage, true(3));
binaryImage = img > 128; % Or whatever value works.
binaryImage = imfill(binaryImage, 'holes'); % Fill it so we can get the outer boundary only.
boundaries = bwboundaries(binaryImage);

[Rajesh]

Dilating the image make the interface less prominent even for a low value of the structuring element. So the subsequent operations become more difficult. Is there an alternate way I could perform dilation to join the pixels in the band of white pixels at the interface?

I = imread('frame (1).jpg');
figure(1), imshow(I);
J=im2double(I);
figure(2), imshow(J);
se = strel('disk',1);
K = imdilate(J,se);
figure(3), imshow(K);

[Image Analyst]

It should make the white ring larger. I don't know why it didn't.

You could try just thresholding the noisy image and then calling bwareafilt() or bwareafilt() to take only blobs of a certain size. Then call bwconvhull() to create the ellipses.