MATLAB Image processing filtering

 One sample image is http://goo.gl/0IOkBZy  .Download the file and name it as

My code is

Now here , I am getting two regions, one the small area (1) and the other as the big one (2)

Answer 1

[Walter Roberson]

bwareafilt() if you have a newer MATLAB. If not then regionprops() 'Area' and select only the smaller of the two regions.

[Rusty]

Thank you for answering. There was a slight error in the question.I have corrected it. The problem is that I may not get only 2 regions always. The bigger region may get split in to 2 or three parts because of value of threshold chosen. please help. Thanks :)

[Walter Roberson]

Look at the eccentricity of the region. The split regions are probably not going to be as circular as the region you want.

Also, image dilation would have a good chance of joining back together the split large region long before the dilation reached the region of interest.

[Rusty]

Actually I am new to image processing. I also tried image dilation and image erosion but I failed. Could you please help me more?

Answer 2

[Image Analyst]

For that particular image, you can use

labeledImage = bwlabel(binaryImage);
smallBlob = ismember(labeledImage, 2) > 0;

In general somehow you have to identify the blob you want, then use ismember() to extract it. For this one, I know that you have only 2 blobs and that the small inside one will be blob #2 because of the way it does labeling. But in general (like your second image), you may have to use regionprops() to get other measurements to identify the particular blob(s) you want, and then use ismember(). See my Image Segmentation Tutorial http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862 for examples of that where I pick out different coins in the image.

You may also want to learn about bwareafilt() or bwselect().

[Rusty]

Sir, Can you please help me in doing this using watershed?

[Image Analyst]

That would not apply for the set of input and output binary images you showed. Who told you that it would?

[Rusty]

Sir, I thought that we can use watershed on the input image, which is not binary but rgb. (( after converting it into grayscale image). Please clear my doubt about this. I applied the algorithms given here I was getting something.

[Image Analyst]

Watershed is used for splitting apart blobs that are connected together. In your examples, you don't show any like that. All your blobs are separate - not touching - so you don't need it.

[Rusty]

But in some cases the blob is near the bigger wall. So when I am using this algorithm I have formed, the small blob merges with the bigger one. So , please guide me for that. I have used your segamentation tutorial and have formed an algorithm, you can see the video of my simulations here : https://goo.gl/kWd88X.

But please guide me for watershed algorithm . Thanks

[Image Analyst]

But you never posted any image where they were connected. Anyway, if you do have images like that, you can split them apart with code Steve put on his blog: http://blogs.mathworks.com/steve/2013/11/19/watershed-transform-question-from-tech-support/ Please try Steve's code on your image and if you can't tweak it enough to get it working, then post that image and your script back here.

[Rusty]

Sir, I have posted the new image ( in which I am facing problem). Please have a look at it. Thanks

Answer 3

[harjeet singh]

try to use this code for these type of image, but not for every slice

clear all
close all
clc

img=imread('ok2.jpg');
img=rgb2gray(img);
img1=img>mean(img(:));
img2=imfill(img1,'holes');
se=strel('disk',2);
img2=imdilate(img2,se);
img3=img2 & ~img1;
img4=imfill(img3,'holes');
img5=img4 & ~img3;
img6=bwareaopen(img5,100);
figure(7)
imshow(img6)

[Rusty]

Thank you

[harjeet singh]

is this works for you?

[Rusty]

Actually I did something else. YOu can see a video of my simulation: at https://goo.gl/kWd88X. I use the methods told by imageAnalyst in his tutorials. Could you please help me solving this algorithm using watershed?