A10 – Preprocessing Handwritten Text

In this activity, we will try to extract handwritten text from an imaged document with lines. In handwriting recognition, individual examples of letters must be extracted.

Procedure.
1. Download one of the images DSC00469.JPG or Untitled_0001.jpg.
2. Crop a portion of the image with handwritten text within lines in the form.

image 1

I cropped image 1 to

image 2 - cropped image

3. Remove the lines using image processing techniques.
To remove the horizontal lines we will use the filter below.

image 3 - filter

This filter is based on the fft of the cropped image which is shown in figure 4.

image 4 - fft of cropped image

Then we will filter the image by convolution,

image 5 - filtered image

4.Binarize the handwriting, clean the image and process such that each handwritten letter is one pixel thick.

image 6 - binarized image

image 7 - image after using opening operator

5. Use bwlabel to label each letter. Comment on the quality of your preprocessing.

image 8 - labeled image

code used:

im=imread('image.jpg');
img=imread('filter.jpg');
im = im2gray(im);
imf = im2gray(img);
ff1=fft2(im);
ff2=fftshift(imf);
imft=ff1.*ff2;
imshow(ff1);


//filtering the image
a=real(fft2(imft));
scf(1)
imshow(a,[]);

//binarize the image
b=im2bw(a, 127/255);//threshold from GIMP
c=1-b;
scf(2);
imshow(c,[]);

// structuring element
SE = ones(2,2)

//opening operation
d=erode(c,SE);
e=dilate(d,SE)
scf(3);
imshow(e,[]);

//label the image
f=bwlabel(e);
scf(4);
imshow(f,[]);

From the labeled image, the processed image is not that clear. :(

Acknowledgements:
Ed
- for the help

Grade:
10/10
- because I did my best and I was able to use what I learned in the activities before.