How to select boundary box?

How to select boundary box of 221.jpg and crop it like 222.jpg ?

221.jpg

222.jpg

Answer 1

[Indunie Florence]

Thanks for everyone who tried to help me. Finally I found the simple way to do it. this is a really awesome method. bwconvhull(Iout)

This is the code segment;

if true
  LB = 100;
  UB = 100000;
  IL = bwlabel(Ifill);
  R = regionprops(Ifill,'Area');
  ind = find([R.Area] >= LB & [R.Area] <= UB);
  Iout = ismember(IL,ind);
  %Ibw1=imcomplement(Iout);
  CH = bwconvhull(Iout);

    %Ibw3=imcomplement(CH);
    [Ilabel num] = bwlabel(CH);
    Iprops = regionprops(Ilabel);
    Ibox = [Iprops.BoundingBox];      
    Icrop = imcrop(I,Ibox);

    resized=imresize(Icrop,[100 100]);

end

Answer 2

[Image Analyst]

This can also be solved using the algorithm I laid out in your previous question http://www.mathworks.com/matlabcentral/answers/132000#answer_139060

[Indunie Florence]

No it doesn't work for this input. This is a binary image. It gave an error,

rows = 3264

columns = 2448

numberOfColorBands = 1

Index exceeds matrix dimensions.

Error in test (line 38)

greenChannel = rgbImage(:, :, 2);

>>

[Image Analyst]

Do you know what the line

greenChannel = rgbImage(:, :, 2);

does? It takes the second plane of the 3D image variable. So the image must be 3D. If it's 2D, like yours which has only 1 color channel in a single 2D image, you can't take the second plane and must just use the image itself. So you have to check for that.

if numberOfColorBands > 1
  % It's color, take just one color channel
  greenChannel = rgbImage(:, :, 2);
else
  % It's monochrome, take the original image.
  greenChannel = rgbImage; % It's not really RGB
end

[Indunie Florence]

Thank you very much for helping me.I didn't know it.I have done all other things and I am stuck in cropping. It seems very simple. But I couldn't do yet.

I changed it as you said for 3 channels. Then there are no errors. But cropping is still not happened.Output is below,

[Image Analyst]

You need to invert your binary image since it's darker than the background. You can either flip the > sign when doing thresholding (which is the easiest) or you can do it in a separate steps (more time consuming) binaryImage = ~binaryImage.

Answer 3

[dhiaa almalki]

if true
  % code
end

                        if true
                          % code
                        end

[Image Analyst]

What is this for? Why don't you start your own new question?