Issues with bwskel skeleton

Hello,

I wanted to use bwskel to get a line down this image of a printed line. However, I am getting small line chunks of the outline instead. What am I doing wrong? The code I used is below and I attached the image.

img= imread('C:\Users\aluce\Documents\random lines\25g 1 10 PSI 5 MMS.jpg'); %put in image directly from FLIR camer
imgray = rgb2gray(img); % turn color image to greyscale 
[imgray, rect] = imcrop(imgray); %opens up image, drag crop box,  double click to select, rect coorinates use for crop
%%
BW = im2bw(imgray,0.8); % convert to binary image 
%imshow(BW)
out = bwskel(BW);
imshow(out)

Answer 1

[Image Analyst]

Andrew:

After you binarized the image, there were some small holes in it that gave loops in the skeleton. I took the largest blob (to get rid of a noise speck) and then filled the holes in the remaining blob. Corrected code:

img= imread('25g 1 10 PSI 5 MMS.jpg'); %put in image directly from FLIR camer
imgray = rgb2gray(img); % turn color image to greyscale 
subplot(2, 2, 1);
imshow(imgray, []);
% [imgray, rect] = imcrop(imgray); %opens up image, drag crop box,  double click to select, rect coorinates use for crop
BW = im2bw(imgray, 0.8); % convert to binary image 
% Take largest blob only.
BW = bwareafilt(BW, 1);
% Fill holes.
BW = imfill(BW, 'holes');
subplot(2, 2, 2);
imshow(BW, []);
%imshow(BW)
minBranchLength = round(sum(BW(:)) / 2)
skelImage = bwskel(BW, 'MinBranchLength', minBranchLength);
subplot(2, 2, 3);
imshow(skelImage)
g = gcf;
g.WindowState = 'maximized'

[Andrew Luce]

Thank you, however, when I run this code, the skeleton is like 7 specks. I'm looking to have a line that traces down over the shape like in the bwskel examples.

[Image Analyst]

Zoom in. You'll see it's continuous. It's just that when it subsamples that huge image to display it in a tiny viewport, some of the pixels are subsampled away. But you can see if you zoom in, or look at the matrix in the Workspace, that there actually is a continuous line. It's not just 7 specks.

[Andrew Luce]

oh man, I'm a dingus, thank you very much.

Andrew

[Andrew Luce]

So I use Hu moments to compare the binary image to an image of the original line ( which has nice sharp edges). Is there a way to flatten out the curve edges to give a better match?

Thnak you

Andrew

[Image Analyst]

Well you could take a single color channel and threshold it. That would give sharp edges. But of course that was the first step in getting the skeleton image in the first place. So I guess I don't know what you think using Hu's moments would get you.