find main area of image

Hello,

I am trying to lear how I can remove the small objects to find the main area of the image (the black section between the long lines (edges)? I want ot get rid of external and internal objects.

Thanks for the helps

Answer 1

[KALYAN ACHARJYA]

#Approximate Here

se = strel('line',3,3);
BW1=imdilate(binaryImage,se);
BW2=bwareafilt(~BW1,1);
se = strel('line',3,3);
BW3=imerode(BW2,se);
se=strel('disk',7);
result=imclose(BW3,se);
imshow(result);

[Very Determined]

Thanks KALYAN. I have binarized the image using imbinarize(I) and continued with your pirce of code. However I am getting an error message. If I dont binarize it I am getting pretty similar error which is related to the second strel command. Am I missing something?

[KALYAN ACHARJYA]

Which Matlab version you are using?

[Very Determined]

I am using R2019a.

[KALYAN ACHARJYA]

I didnot get any error, when I tried with the attach image. Also shown the result.

[KALYAN ACHARJYA]

Have you confirmed that the passing binaryImage data is binary image?

[Very Determined]

This is the whole code returns error the second time strel is used.

My understanding is that the imbinarize gives me a binary image, by the way. Right? Looking to numbers I see all have bianary values.

BW = imread('Capture.png');
imshow(BW)
BW = imbinarize(BW);
se = strel('line',3,3);
BW1=imdilate(BW,se);
BW2=bwareafilt(~BW1,1);
se = strel('line',3,3);
BW3=imerode(BW2,se);
se=strel('disk',7);
result=imclose(BW3,se);
imshow(result);

[KALYAN ACHARJYA]

#Try

BW = rgb2gray(imread('Capture.png'));
figure,imshow(BW)
BW = imbinarize(BW);
se = strel('line',3,3);
BW1=imdilate(BW,se);
BW2=bwareafilt(~BW1,1);
se = strel('line',3,3);
BW3=imerode(BW2,se);
se=strel('disk',7);
result=imclose(BW3,se);
figure,imshow(result);

[Very Determined]

Great. It works. So I have to convert it from RGB first.

Thak you again.

Answer 2

[Image Analyst]

Looks like you're doing an edge detection for some reason. Can I see the original image? Because usually edge detection is NOT what you want to do, for exactly the reasons you are seeing. It would probably be better to use a different segmentation method, like thresholding.

[Image Analyst]

Exactly as I thought. You don't want to do edge detection. You should do color segmentation. If all your subjects are on green grass, you should use the Color Thresholder app on the Apps tab of the tool ribbon to find all green stuff. Anything that is not green is your subject. Let me know if you can't figure it out.

Or you might try training a deep learning model to identify the subjects from overhead images.

[Very Determined]

Hello, I will try this then. Thank you for the reponse.

[EDIT] Flag moved here to a comment.

Could you remove the images and the key word (cattle) from your response. I am not allowed to post the images/topic in public as a part of my research agreement.

[Image Analyst]

I changed the keyword to subject. It looks like the images are no longer there for some reason, perhaps because you took them down from your site.

[Very Determined]

Thank you.