For font recognition thinning will be an appropriate step to extract feature?

Hi, I am doing a project on Font recognition using HMM model. I have done pre-processing. Converted RGB to Gray,Then Gray to binary, and then binary to thinning. Now I want to extract the feature . My data is a sentence written in Times New Roman .I have written code and each word is having bounding box. I want to extract features of each word to get a feature vector of all words in form of row vector. Thanks in advance

Answer 1

[harjeet singh]

dear rutika, yes thinning is an appropriate step for extract features unless until you only need geometric features for making a feature vector table.

[Rutika Titre]

Hello Sir, Thanks for your reply. Sir,Geometric features as in???plz help...

[Rutika Titre]

Sir I have attached my code's output named "untitled.fig" I want to extract the features of each word inside the bounding. Will Area , Eccentricity, Orientation give me features of bounding box? Can you please help me to extarct feature vector of the image I have attached. Thanking you!

[Image Analyst]

Attach an image. Use the green and brown frame icon.

Answer 2

[harjeet singh]

try to use this code and do include more features for word, i used the number of alphabets a word has and area of the word and center of gravity for written word.

clear all
close all
clc

image=imread('untitled.tif');
figure(1)
imshow(image)
drawnow

img=image(:,:,1)==255 & image(:,:,2)==0 & image(:,:,3)==0;
img=bwareaopen(img,20);
[lab,num]=bwlabel(img);

sub=ceil(sqrt(num));
for i=1:num
    [r,c]=find(lab==i);
    img_1=image(min(r)+1:max(r)-1,min(c)+1:max(c)-1,:);
    figure(2)
    subplot(sub,sub,i)
    imshow(img_1);
    drawnow

      img_2=logical(img_1);
      [lab1,num1]=bwlabel(img_2(:,:,1));

      alphabets=num1;
      area=length(r);
      cog_r=mean(r);
      cog_c=mean(c);

      fvt(i,:)=[alphabets area cog_r cog_c];
  end

[Rutika Titre]

Sir, The code is not working num=0, .Sir there is a technique using sliding window, which mover from left to right and at the same time extracts the features of every word. As, m doing Font recognition I read papers on it. And this sliding window is used. to extract features. Can you help me regarding this technique. Thanking You!

Answer 3

[harjeet singh]

do upload the code you are using for features, and i used the clipped snap which you uploaded as figure, do use this pic for the attache code

clear all
close all
clc

image=imread('untitled.bmp');
figure(1)
imshow(image)
drawnow

img=image(:,:,1)==255 & image(:,:,2)==0 & image(:,:,3)==0;
img=bwareaopen(img,20);
[lab,num]=bwlabel(img);

sub=ceil(sqrt(num));
for i=1:num
    [r,c]=find(lab==i);
    img_1=image(min(r)+1:max(r)-1,min(c)+1:max(c)-1,:);
    figure(2)
    subplot(sub,sub,i)
    imshow(img_1);
    drawnow

      img_2=logical(img_1);
      [lab1,num1]=bwlabel(img_2(:,:,1));

      alphabets=num1;
      area=length(r);
      cog_r=mean(r);
      cog_c=mean(c);

      fvt(i,:)=[alphabets area cog_r cog_c];
  end

[Rutika Titre]

Thank you sir, Actually after I tried the code and my output had come...sorry did not inform you. I think the sliding question was not correct question to ask.Thank-you sir for kind consideration.

[Ihtisham Khan]

Rutika Titre How do you apply these features to the hmm model??? Please help,,,