How do I get only the width and height?

I have a set a minimum width = 300 and height = 300 for the bounding box. I need to compare the width and height that is above 300. But right now, I am unable to compare them as I could not return the width or height values in each row. May I know how can I achieve that? Thanks.

bbox = 
1174  557  63  66
1692  1517  74  78
1157  467  69  73
1123  1447  72  77
887  1572  77  82
1586  1275  93  100
1589  1517  89  95
706  1300  115  122
540  1309  122  130
496  1718  1262  1342

Answer 1

[Image Analyst]

To get the overall bounding box:

bbox = [
  1174  557  63  66
  1692  1517  74  78
  1157  467  69  73
  1123  1447  72  77
  887  1572  77  82
  1586  1275  93  100
  1589  1517  89  95
  706  1300  115  122
  540  1309  122  130
  496  1718  1262  1342]
% Find the xRight columns
xRight = bbox(:, 1) + bbox(:, 3)
% Find the yBottom rows
yBottom = bbox(:, 2) + bbox(:, 4)
% Find the overall left-most x value
xLeft = min(bbox(:, 1))
% Find the overall right-most x value
xRight = max(xRight)
% Find the overall top-most y value
yTop = min(bbox(:, 2))
% Find the overall right-most x value
yBottom = max(yBottom)
% Get the overall bounding box
overallBox = [xLeft, yTop, xRight-xLeft, yBottom-yTop]

% Plot the individual boxes
for row = 1 : size(bbox, 1)
  rectangle('Position', bbox(row, :), 'EdgeColor', 'b', 'LineWidth', 2);
  hold on;
end
% Plot the overall bounding box in red.
rectangle('Position', overallBox, 'EdgeColor', 'r', 'LineStyle', '--', 'LineWidth', 2);
axis equal
xlim([0,2000]);
ylim([0,3500]);

[Leonard Yeo]

How do I compare and remove bbox that are smaller than for example 300 by 300? and keep those that are above the size.

[Image Analyst]

If you just want to remove them from the bbox array, you can do

smallBoxIndexes = bbox(:, 3) < 300; % Find rows.
bbox(smallBoxIndexes, :) = []; % Remove rows.

It's a whole different answer if you're talking about imaging and a binary image and labeled image. You can use ismember() to get a new labeled image, then remeasure the blobs that are left:

bigBoxIndexes = find(bbox(:, 3) > 300); % Find rows.
labeledImage = ismember(labeledImage, bigBoxIndexes);
measurements = regionprops(labeledImage, 'BoundingBox'); % Re-measure what's left.

Or, you might want to use bwareaopen() before labeling to get rid of blobs based on areas.

[Leonard Yeo]

This is what I need! Thank you very much!

Answer 2

[Andrei Bobrov]

bbox(3);

[Leonard Yeo]

Hi, I've tried bbox(3) but it returns the value of the x coordinate.

bbox(3) = 1157

[Andrei Bobrov]

[Leonard Yeo]

I have a set a minimum width = 300 and height = 300 for the bounding box. I need to compare the width and height that is above 300. But right now, I am unable to compare them as I could not return the width or height values in each row.

bbox = 
1174  557  63  66
1692  1517  74  78
1157  467  69  73
1123  1447  72  77
887  1572  77  82
1586  1275  93  100
1589  1517  89  95
706  1300  115  122
540  1309  122  130
496  1718  1262  1342

So if I were to put bbox(3), it only returns 1157 instead of 69.

>> bbox(3)

    ans =

            1157

Answer 3

[Image Analyst]

Try this:

bbox = [
  1174  557  63  66
  1692  1517  74  78
  1157  467  69  73
  1123  1447  72  77
  887  1572  77  82
  1586  1275  93  100
  1589  1517  89  95
  706  1300  115  122
  540  1309  122  130
  496  1718  1262  1342]
% Set min width in column 3 to 300
bbox(:, 3) = max(300, bbox(:, 3))
% Set min height in column 4 to 300
bbox(:, 4) = max(300, bbox(:, 4))

The final result:

bbox =
        1174         557         300         300
        1692        1517         300         300
        1157         467         300         300
        1123        1447         300         300
         887        1572         300         300
        1586        1275         300         300
        1589        1517         300         300
         706        1300         300         300
         540        1309         300         300
         496        1718        1262        1342

[Leonard Yeo]

Hi, sorry but there will be setting the bbox width and height to 300. But, there is not the objective of it. My goal is to filter out bounding box size which are too small to be classified as an object and only put a bounding box which are of certain sizes.