I want to smooth the boarder of all white areas in this Binary Image (even the tiny ones). Consider it like smoothing with your hand.

[Walter Roberson]

https://www.mathworks.com/help/images/ref/imdilate.html

or possibly

https://www.mathworks.com/help/images/ref/imclose.html

Answer 1

[KALYAN ACHARJYA]

In addition of the above links provided by @Walter morphological operations, imdilate and imclose, you may try with the following code also (@Image Analyst related answer, I cant find the link right now).

*Here image_binary is the input binary image, please note you have to manage the tradeoff between original shape changes vs. smoothness.

windowSize=12;  % Decide as per your requirements
kernel=ones(windowSize)/windowSize^2;
result=conv2(single(image_binary),kernel,'same');
result=result>0.5;
image_binary(~result)=0; 
figure, imshow(image_binary);

Hope it helps!

[Muhammad Saafan]

A lot of details will be lost.

[Walter Roberson]

... Yes?

Remember that we have not been shown an example of what you want the output to look like.

[Muhammad Saafan]

As you make it by your hand in a paper. May be extracting each white area and fitting its boundaries

[Walter Roberson]

Is your concern that some of the very small regions disappear with KALYAN's proposed code, or are you concerned about the larger areas as well?

[KALYAN ACHARJYA]

Hi Saafan, as per my undestanding, all smothing operation, you bound to loose data, you have to trade off between how much smooth data vs. loss of original information.

For better smoth operartion, the resolutions / samples size should be more, so that proposed approach can follow the trend (fitting).

[Muhammad Saafan]

Thanks for your help.

Answer 2

[Walter Roberson]

img = im2bw(imread('image.jpeg'));
subplot(1,2,1)
imshow(img);
subplot(1,2,2)
newimg = imdilate(img, ones(3));
imshow(newimg)

[Muhammad Saafan]

First, Thanks for your effort and help.

The final code enlarges the original data and I only want to smooth the edges of the white area.

[Muhammad Saafan]

Consider this edges. What I aiming at is smoothing this edges as the blue line.

May be I need to extract the white areas and smooth each one separately.

Thanks

[Walter Roberson]

You cannot do that without increasing the resolution. You are at the limits of drawing curves by approximating with square pixels.

[Bruno Luong]

He is not at the limit of the pixel, the boundary follows rectangles of of 3-6 pixels if you zoom in.

[Walter Roberson]

Some of the places I checked are single pixel boundary. For example on the left towards the center, there is a group of three somewhat small (but not tiny) areas sort of aligned vertically. The top of the middle of them is single pixel curvature.

[Bruno Luong]

Still the smoothness is not limit by the pixel, otherwise Kalyan's image would not be smoother.

[Walter Roberson]

A relevant question would be whether black and white is to be retained or if grayscale could be used, as grayscale could permit some antialiasing.

Answer 3

[Image Analyst]

Just blur the image with conv2() like KALYAN showed.

blurredImage = conv2(double(binaryImage), ones(3)/9, 'same');

It won't be a binary/logical image anymore, but it will be smoother. You didn't say it had to stay a binary image.

And, it it's not already obvious, you cannot smooth a boundary without losing details. I know you said blurring will lose details, but losing details is EXACTLY what you want when you smooth boundaries. If you don't lose details, it's not smoother!