Extraction of rectangular blob from binary image

i have a binary image having many blobs of different shapes. the one blob is full rectangle. now i want to extract the rectangular blob.

Answer 1

[Akira Agata]

How about comparing length around the blob and length around the bounding box of the blob, like:

% sample blogs
I = imread('pillsetc.png');
Iblob = rgb2gray(I) > 200;

% measure perimeter and bounding box for each blob
stats = regionprops(Iblob,{'BoundingBox','perimeter'});
stats = struct2table(stats);

% extract the blob whose perimeter is close to round length of its bounding box
stats.Ratio = 2*sum(stats.BoundingBox(:,3:4),2)./stats.Perimeter;
idx = abs(1 - stats.Ratio) < 0.1;
stats(~idx,:) = [];

% show the result
imshow(Iblob);
hold on
for kk = 1:height(stats)
  rectangle('Position', stats.BoundingBox(kk,:),...
      'LineWidth',    3,...
      'EdgeColor',    'g',...
      'LineStyle',    ':');
end

[Ayaz Wazir]

Dear Akira Agata, i want to extract the blob which is very very closed to rectangular shape as in the image with red circle. the image is given below.

[Akira Agata]

OK. How about the following coce? In this code, I have added the 2nd metric to identify the rectangular shape.

% Reading Rect.jpg and make it's blob image
I = imread('Rect.jpg');
I = rgb2gray(I);
Iblob = I > 127;
Iblob = imclearborder(Iblob,4);

% measure perimeter and bounding box for each blob
stats = regionprops(Iblob,{'Area','BoundingBox','perimeter'});
stats = struct2table(stats);

% 1st metric: ratio between perimeter and round length of its bounding box
stats.Metric1 = 2*sum(stats.BoundingBox(:,3:4),2)./stats.Perimeter;
idx1 = abs(1 - stats.Metric1) < 0.1;

% 2nd metric: ratio between blob area and it's bounding box's area
stats.Metric2 = stats.Area./(stats.BoundingBox(:,3).*stats.BoundingBox(:,4));
idx2 = stats.Metric2 > 0.8;

idx = idx1 & idx2;

stats(~idx,:) = [];

% show the result
imshow(Iblob);
hold on
for kk = 1:height(stats)
rectangle('Position', stats.BoundingBox(kk,:),...
    'LineWidth',    3,...
    'EdgeColor',    'g',...
    'LineStyle',    ':');
end

Answer 2

[Image Analyst]

See my Image Processing Tutorial http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862&sort=downloads_desc where I do exactly that.