Azzera filtri
Azzera filtri

How to centralize alphabets in an image?

2 visualizzazioni (ultimi 30 giorni)
Hello,
I have an 80x80 image that contains an Arabic alphabet (black pixels) against white background. The alphabet is dislocated from the center of the image. The idea is to centralize the alphabet in the image. As I have many alphabets, the algorithm should do the same for other alphabets.
Demo image is here,
Your help is highly appreciated.
Thanks.

Risposta accettata

Walter Roberson
Walter Roberson il 15 Apr 2018
regionprops() of imcomplement() of the image, requesting 'SubarrayIdx', and use those to access the image.
Note: you should also consider regionprops() of imcomplement() of the image, requesting 'Image' . You could use this if you are not interested in analyzing grayscale levels, just on/off
Note: if your image is binary, and your characters can have disconnected parts, then
regionprops(double(~BinaryImage), ...)
This prevents regionprops from considering each disconnected component separately.
Note: I am not familiar with the rules for Arabic, but in a number of other alphabets, the position of a character relative to the base-line can make a difference in the meaning of the character. For example in some languages a dot near the base-line is a period (punctuation) but a raised dot can indicate an accent or a vowel mark. Because of that, it is sometimes important to either not center the object, or else to also retain information about the object position for later testing against the base-line coordinates.
  1 Commento
Ihtisham Khan
Ihtisham Khan il 15 Apr 2018
Thanks a lot Walter Roberson , your answer cleared my concept.

Accedi per commentare.

Più risposte (1)

Image Analyst
Image Analyst il 15 Apr 2018
I'd threshold and use regionprops() to find the centroid of the character.
binaryImage = grayImage < 128;
binaryImage = bwareafilt(binaryImage, 1);
props = regionprops(binaryImage, 'Centroid');
xCentroid = props.Centroid(1);
yCentroid = props.Centroid(2);
Then I'd use circshift() or imtranslate() to move the character centroid to the middle of the image.
[rows, columns] = size(binaryImage);
middleRow = rows/2;
middleColumn = columns/2;
etc.
  2 Commenti
Ihtisham Khan
Ihtisham Khan il 15 Apr 2018
Thats OK, but for alphabets with disconnected parts regionprops calculates more than one center. How to fix that.
Walter Roberson
Walter Roberson il 15 Apr 2018
props = regionprops(double(binaryImage), 'Centroid');

Accedi per commentare.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by