how to clean up and refine edges of this image

Hi, In the image shown above, I want to remove the noise from the main image shown in the red circles. Also i want to smooth out the edges of the image. I tried curve fitting of the end points, but I am not getting the best results. Any suggestions?

[KSSV]

Upload the original image..

[Hemant Agiwal]

by original you mean without the red circles? Because this is the starting image...

[KALYAN ACHARJYA]

You can do that very easily, upload/attach the original image(without red circles)

Answer 1

[Matt J]

BW= bwareafilt(~yourImage,1);

newImage=~(imclose(BW,strel('disk', something) )  );

Answer 2

[Image Analyst]

One way is to blur it and re-threshold it.

See attached testGrayImage.m

Another way is to get the x and y coordinates with bwboundaries() and smooth them with a Savitzky-golay Filter (a moving polynomial filter). See attached demo.

[Hemant Agiwal]

hey, How did you get the plot of smoothened image after applying sgolay filter. The filter provides new coordinates for the boundary, what did you do after that... Couldn't really understand from the code

[Image Analyst]

I used this code:

smoothX = sgolayfilt(x, polynomialOrder, windowWidth);
smoothY = sgolayfilt(y, polynomialOrder, windowWidth);
% First, display the original gray scale image.
subplot(2, 3, 5);
imshow(grayImage, []);
axis image; % Make sure image is not artificially stretched because of screen's aspect ratio.
hold on;
plot(smoothX, smoothY, 'r-', 'LineWidth', 2);

Are you asking how I did I "get the plot"? I just used the plot() function.

To use this code, you'd first call bwboundaries() to get the actual/noisy x and y coordinates, then call sgolayfilt() on them.

[Hemant Agiwal]

hi, Thanks for all your help.

Can you please explain

 "% Threshold again.
binaryImage = blurredImage > 0.5;"

How did you come up with the number 0.5 and what is its significance. If you could send me some link that explains the theory behind this, it would be really helpful.

Thanks again

[Image Analyst]

That's the midpoint of the blur. If you pick that, then the area of the second binary image should be really close to the area of the original binary image.