How to crop image using imfreehand and save the output without connecting the boundaries

I have an input Image, I want to crop the image using imfreehand with closed boundary option. Unfortunately when I create mask and save the output:

h = imfreehand('Closed',false);
output_mask = h.createMask();

the end points of the cropped region becomes a closed polygon.

This is input file

This is the required output.

I am getting this output

Thanks in advance

Answer 1

[Matt J]

I think you simply neglected to trace the full boundary of the shape. Though I traced very quickly and unneatly, the result I get below shows what you should be getting when you trace the full boundary.

[Praveen GB]

Thanks for the answer. After finding the bwboundaries, what step should be performed.

 [B,L] = bwboundaries(mask);

[Matt J]

But you said you were using imfreehand. What does bwboundaries have to do with your posted question?

[Praveen GB]

Yes I used imfreehand and upon your suggestion of tracing the boundary, i tried to find the bwboundaries of the mask. Do i need to use B = bwtraceboundary(BW,P,fstep) for tracing the boundaries? How can i automatically feed the coordinates of the end points to be removed.

The below code, i use to crop the image with closed option and the create a mask of the cropped region.

 h = imfreehand('Closed',false);
 mask = h.createMask();

[Matt J]

If the point all along was just to get the boundaries, I think you could just use edge,

>> A=imread('image.png');
>> B=rgb2gray(A);
>> imshow(edge(B))

[Praveen GB]

Hi Matt, Thanks for the reply. I have so many such structures in a image, I want to manually crop these structures and find the edges as shown in the depicted output.

Thanks

[Image Analyst]

Your depicted desired output does not show the edges - it shows the entire, filled blob.

Answer 2

[Image Analyst]

If you want that output, why not simply binarize the image:

grayImage = imread('image.png');
if ndims(grayImage) == 3
	grayImage = rgb2gray(grayImage); % Convert RGB image into gray scale.
end
subplot(1, 2, 1);
imshow(grayImage);
binaryImage = ~imbinarize(grayImage);
subplot(1, 2, 2);
imshow(binaryImage);

It looks like the above image is what you said you want, so why do freehand masking?

[Praveen GB]

Hi Image Analyst, Thanks for the reply. I have so many such structures in a image, I want to manually crop these structures and find the edges as shown in the depicted output.

Thanks

[Image Analyst]

But why? If you need to crop each (which may not even be necessary) and can do it automatically by using regionprops() to get the bounding box and imcrop to get a subimage, why would you want to do it manually.

See my image segmentation tutorial in my File Exchange to see how to crop automatically.

See my attached imfreehand() demo to do it manually.

And please explain why you want the edges. Because many, many times novices think the first step in any image analysis algorithm is edge detection when it rarely is, and can be analyzed better without doing edge detection.

[Praveen GB]

This is my input and I want to crop the input and get output like .

Since so many structures are present in input image, I have to crop it using free hand.

Thanks

[Image Analyst]

OK, no explanation for why you need edges - I guess it's proprietary or something.

However my masking demo will allow you to mask freehand-drawn regions with optional cropping into a subimage if you want. Does it not do what you want? You can still detect edges if you feel that is definitely the best approach.