Comparing two binary images: if a pixel is 1 in both images, write a 1 into a new image, otherwise, write a 0.

Hello everyone,

I have a somewhat noisy image and I would like to reduce or clear the noise.

To do so, I performed and opening. Since I don't want to lose any information of the objects' surface, I think it is easiest, to "compare" the two pictures, the original and the opened one. Meaning, when I have a 1 in the original image and a 1 in the opened image, I want to write a 1 into a third, new image. When I have a 1 in the original picture and a 0 in the opened picture I want to write a 0 in the new, third image. And when there are two 0s, I of couse want a 0 in the new image.

If you click on the link, you see the original image on the left and the opened image in the middle. It should look like a mixture of the first to images with all the surface information...

I tried to multipy the images (result=ImgOpen*ImgOrg), which in my opinion should lead to the desired result (1*1=1; 1*0=0), but the result looks like the image on the right...?!

http://s14.directupload.net/file/d/3051/29ym8bbc_jpg.htm

Do you understand what I mean? It's a little hard to describe that in another language. Sorry!

Thank you! Thymes

Answer 1

[Walter Roberson]

result = ImgOpen .* ImgOrg;

The * operator is mathematical matrix multiplication, not element-by-element multiplication.

Note: the answer given for your other question is still correct,

result = ImgOpen & ImgOrg;

[Falak Memon]

i want to compare two images which are segmented already

[Walter Roberson]

@Falak Memon

You should open a new question about that; it is not related to the current topic.

Note: "compare" of two images is not well-defined unless you define it precisely. "FInd 12 mistakes in the second image" is a quite different technique than "Decide which kind of brain tumour this patient has" or "Figure out where in the world this picture was taken".

Answer 2

[Julian]

Hello and thank you!

I tried it with the element-by-element multiplication with .* and the result is again on the right.

http://s7.directupload.net/file/d/3052/y94k2vch_jpg.htm

Do I think wrong, or shouldn't the surface of the main structure be much more like the one of the original picture on the right? I thought that this might be due to the size and shape of the sructuring element, but when I increased it's size it looked even worse...

Any ideas?

Answer 3

[Julian]

Ok, I found the mistake in my thoughts. imopen performs dilation and erosion with the same structuring element. I chose the wrong size of it, deleting to many informations in the first place. I added a second structuring element with a bigger size and perform the dilation with it.

That works :)