How to improve the filtering of a bubble image

I need to better filter my image (a bubble) in order to obtain a better filtering than the one I already have (code), in order to get a better approximation of the contour of the bubble.

Would you know of any other filters or what I could change in my code to improve the borders I drew in red and to make the filtered image of the bubble (image 2) better look like the real bubble outline? (it should look like a semi-ellipse)

Thanks in advance

clc
clear
close all

% 1 - Reading the image
image = imread('bubble.png');
figure, imshow(image);

% 2 - Cropping the image
cropped = imcrop(image,[552.9 0 639.2 1080]);
figure, imshow(cropped);

% 3 - Gray scale
gray = rgb2gray(cropped);
figure, imshow(gray);

% 4 - Threshold filter
thresh = im2bw(gray, 0.5);
figure, imshow(thresh);

% 5 - Small pixel cluster removal filter
remove = bwareaopen(thresh, 13200);
figure, imshow(remove);

% 6 - Close connections filter
se = strel('line',300,0);
closing = imclose(remove,se);
figure, imshow(closing);

% 7 - Fill center of objects filter
OriginalLine = closing(1 , :);
OriginalLine2 = closing(end , :);
closing(1, :) = true; 
closing(end, :) = true;
fill = imfill(closing, 'holes');
closing(end,:) = false;
closing(1,:) = false;
fill(1, :) = OriginalLine;
fill(end, :) = OriginalLine2;
figure, imshow(fill);

[Rik]

(Retrieved from Google cache)

Question title:

How to improve the filtering of a bubble image

Question body:

I need to better filter my image (a bubble) in order to obtain a better filtering than the one I already have (code), in order to get a better approximation of the contour of the bubble.

Would you know of any other filters or what I could change in my code to improve the borders I drew in red and to make the filtered image of the bubble (image 2) better look like the real bubble outline? (it should look like a semi-ellipse)

Thanks in advance

clc
clear
close all

% 1 - Reading the image
image = imread('bubble.png');
figure, imshow(image);

% 2 - Cropping the image
cropped = imcrop(image,[552.9 0 639.2 1080]);
figure, imshow(cropped);

% 3 - Gray scale
gray = rgb2gray(cropped);
figure, imshow(gray);

% 4 - Threshold filter
thresh = im2bw(gray, 0.5);
figure, imshow(thresh);

% 5 - Small pixel cluster removal filter
remove = bwareaopen(thresh, 13200);
figure, imshow(remove);

% 6 - Close connections filter
se = strel('line',300,0);
closing = imclose(remove,se);
figure, imshow(closing);

% 7 - Fill center of objects filter
OriginalLine = closing(1 , :);
OriginalLine2 = closing(end , :);
closing(1, :) = true; 
closing(end, :) = true;
fill = imfill(closing, 'holes');
closing(end,:) = false;
closing(1,:) = false;
fill(1, :) = OriginalLine;
fill(end, :) = OriginalLine2;
figure, imshow(fill);

[Rena Berman]

(Answers Dev) Restored edit

Answer 1

[Image Analyst]

Can you assume the ellipse axes will be vertical and horizontal? If not, see the FAQ: Click here

Otherwise you can use bwferet() to get the caliper dimension in one direction. Call it 2*a. Use sum() or regionprops() to get the area and centroid,

binaryImage = bwareafilt(binaryImage, 1); % Take largest blob only.
props = regionprops(binaryImage, 'Area', 'Centroid');
area = props.Area;
xCenter = props.Centroid(1);
yCenter = props.Centroid(2);
hold on;
plot(xCenter, yCenter, 'r+', 'MarkerSize', 60, 'LineWidth', 2); % Plot crosshairs at the centroid.

then, since the area of an ellipse is pi*a*b,

b = area / (pi * a);

Create an ellipse from the code in the FAQ:

xCenter = 12.5;
yCenter = 10;
xRadius = 2.5;
yRadius = 8;
theta = 0 : 0.01 : 2*pi;
x = xRadius * cos(theta) + xCenter;
y = yRadius * sin(theta) + yCenter;
plot(x, y, 'LineWidth', 3);
grid on;

where xRadius = b and yRadius = a (if vertical is the feret diameter direction you picked).

Or you might try activecontour(). See attached demo.

[Image Analyst]

You have an old version then. Try using regionprops() and getting the MajorAxisLength and MinorAxisLength, which are the axes of an ellipse fit to your blob.

[Image Analyst]

See step by step instructions in Steve's Image Processing Blog.