How to remove the black regions in the holes of the letters?

I have detected and cropped the license plate.But the image looks like this.I have to remove the black colur area in the holes in letter"B" and numbers "6" and"9".How can I do so?Anyone please help me.

Answer 1

[Image Analyst]

Looks like you need to improve your segmentation algorithm. Were the white spots ever there in a binary image earlier in the process? Then figure out which step got rid of them. Perhaps it was a bwareaopen() or bwareafilt().

[ezhil K]

figure,subplot(2, 1, 1);
imshow(c);
if ndims(c) > 1
	c = c(:,:,1); % Take red plane
end
%binaryImage = imbinarize(c);
% Get rid of anything touching the edge of the image
% because this will definitely not be a license plate.
binaryImage = imclearborder(c);
% Extract the biggest 4 blobs
biggest4 = bwareafilt(binaryImage, 1, 'largest');
subplot(2, 1, 2);
imshow(biggest4);

This is what I have done as a pervious step and in this step itself the letters got changed as above.So how should I change my code to get back the original letters?Please help me

[ezhil K]

This is the input image to the above line of code

[Image Analyst]

You need to use the code I gave you in the other question.

Then fill the blobs before taking the largest ones. Use that as a mask against the original mask

mask = imclearborder(mask);
mask2 = imfill(mask, 'holes');
mask2 = bwareafilt(mask2, 4); % Take 4 largest.
% Use it to mask original
mask(~mask2) = false;

[ezhil K]

What is mask here?

[ezhil K]

actually where should I use these lines u provided?

[Image Analyst]

You need to use those right after you've thresholded/binarized the image to create the initial mask. Of course it's no good so we need to do some spatial things to clean it up, like getting rid of things that touch the border, things outside the license plate, etc.

By the way, why aren't you using published algorithms for this? They're out there - no need to invent your own.

[ezhil K]

What is mask here?Does it mean the binary image?

[ezhil K]

Yeah,I got the output.Thanks a lot.