How to get multiple bounding boxes

I was trying to use regionprops to get bounding boxes around separated digits in an image, but the resulting bounding box ended up covering all of the digits like this:

Can this be fixed? Code can be found below, where im is the image of the digits

%Threshold image gray levels
moreThan25 = im > 25;
LEQThan25 = im <= 25;
im(moreThan25) = 255;
im(LEQThan25) = 0;

%Filter out noise
im = medfilt2(im);

boundingBoxes = regionprops(im, 'BoundingBox');
imshow(im)
hold on
for k = 1:length(boundingBoxes)
    curr = boundingBoxes(k).BoundingBox;
    rectangle('Position', [curr(1), curr(2), curr(3), curr(4)], 'EdgeColor', 'r');
end
pause;

Answer 1

[Matt J]

If that happens, it is likely that the background black pixels are not really zero. You could use imbinarize() to get a purely binary image.

The 5 digit is broken into multiple regions, however, so you will have to seal the breaks in order to get a single bounding box around that. Perhaps using imclose().

[Daniel Larsson]

imbinarize() made it work! Though I still need to close up the 5 digit as you mentioned, but I'll try that on my own for a bit (using imclose() sounds like a good start). Thanks for the help!

Answer 2

[Image Analyst]

Daniel:

Try this:

% Demo by Image Analyst
clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
clear;  % Erase all existing variables. Or clearvars if you want.
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 20;
markerSize = 40;
%--------------------------------------------------------------------------------------------------------
%    READ IN IMAGE
fileName = 'numbers.png';
grayImage = imread(fileName);
% Get the dimensions of the image.
% numberOfColorChannels should be = 1 for a gray scale image, and 3 for an RGB color image.
[rows, columns, numberOfColorChannels] = size(grayImage)
if numberOfColorChannels > 1
    % It's not really gray scale like we expected - it's color.
    % Extract the blue channel.
    grayImage = grayImage(:, :, 3);
end
%--------------------------------------------------------------------------------------------------------
% Display the image.
subplot(3, 3, 1);
imshow(grayImage);
impixelinfo;
axis('on', 'image');
title('Original Gray Scale Image', 'FontSize', fontSize, 'Interpreter', 'None');
hold on
drawnow;
% Maximize window.
g = gcf;
g.WindowState = 'maximized'
drawnow;
%--------------------------------------------------------------------------------------------------------
% Get a mask using locally adaptive thresholding.
mask = imbinarize(grayImage, 'adaptive');
% Get rid of blobs less than 300 pixels in area.
mask = bwareaopen(mask, 300);
% Display the image.
subplot(3, 3, 2);
imshow(mask, []);
impixelinfo;
axis('on', 'image');
title('Mask/Binary/Logical Image', 'FontSize', fontSize, 'Interpreter', 'None');
hold on
drawnow;
%--------------------------------------------------------------------------------------------------------
% Get the horizontal profile
horizontalProfile = any(mask, 1);
subplot(3, 3, 3);
plot(horizontalProfile, 'b-', 'Linewidth', 2);
grid on;
xlabel('Column', 'FontSize', fontSize)
% Get the bounding boxes of the profile
x1 = strfind(horizontalProfile, [0, 1])
x2 = strfind(horizontalProfile, [1, 0])
% Go through all the boxes and if they are within the profile, expand them to the profile.
numLetters = length(x1);
for k = 1 : numLetters
    thisLetter = mask(:, x1(k):x2(k));
    % Find overall bounding box.
    [r, c] = find(thisLetter);
    % Crop.
    r1 = min(r);
    r2 = max(r);
    c1 = min(c);
    c2 = max(c);
    thisLetter = thisLetter(r1:r2, c1:c2);
    subplot(3, 3, k+3)
    imshow(thisLetter);
    caption = sprintf('Letter #%d of %d', k, numLetters);
    title(caption, 'FontSize',fontSize)
end

[Daniel Larsson]

Oh wow, this looks great! Thanks a lot :)

[Image Analyst]

If you like an answer youi can click the Vote (thumbs up) icon to award the answerer "reputation points".

[Daniel Larsson]

Ah I see, sorry I'm new to posting on this forum, just gave you a thumbs up!

Answer 3

[Image Analyst]

I think this

%Threshold image gray levels
moreThan25 = im > 25;
LEQThan25 = im <= 25;
im(moreThan25) = 255;
im(LEQThan25) = 0;
%Filter out noise
im = medfilt2(im);
boundingBoxes = regionprops(im, 'BoundingBox');

should be this:

% Threshold image gray levels
moreThan25 = im > 25;
% Filter out noise
moreThan25 = bwmorph(moreThan25, 'majority'); % Or, better, use bwareafilt() or bwareaopen()
boundingBoxes = regionprops(moreThan25, 'BoundingBox');

[Daniel Larsson]

This solution also corrected the bounding boxes! Although the resulting image was not binarized (only in black and white), so this time I went with the other solution. Still though, thanks for the help!