How to detect straight lines on grayscale images？

Question

I have some grayscale image，one of which is shown following：

This image is like a “disk” which contains a lot of noise in the non-black area. I want to detect the line in it. The line is easy to be seen by eyes but quite difficult to be seen by computer. I handed out this line for a better demonstration.

Now I've tried many methods，such as Hough Line transform, Radon transform and so on. However, it's hard to tell whether I used them in a wrong way or those methods do not apply to this task.

I am a beginner in image processing. Could anyone here provide some ideas? Thanks a lot!

Answer 1

You can try using Gabor filters. Here is a start:

img = imread('https://i.sstatic.net/sHA7w.png');
% gray out empty spaces
img(img == 0) = 123;
% create gabor filter bank
wavelength = 6;
orientation = 5:5:180;
g = gabor(wavelength,orientation);
outMag = imgaborfilt(img,g);
% plot the magnitude at peak orientation
ii = 23;
figure
imagesc(outMag(:,:,ii))
title(['Magnitude at ',num2str(orientation(ii)),'deg'])
% create a mask
mask = outMag(:,:,ii) > 250;
figure;
imshow(mask)

You still have to detect in which layer you got high magnitude, and filter the mask to hide the edges and small patches.