Find Base Line of the words in a image

hai...my name is edward and i'm from indonesia...

i have read a paper Personal identification Based on handwriting by Sachin kumar Mangal...

here is the link : http://www.security.iitk.ac.in/contents/publications/more/Handwriting.pdf....

i really interested with that paper and i wanna make the code from that paper... but i have a trouble..i think to implementation this paper to a matlab code is very hard..i can understand how to find topline,upperline,base line,bottom line middle zone and lower zone from that words..

is everybode want to help me ?for example ihave a picture :

http://tinypic.com/view.php?pic=1nzslw&s=6

the formula is :

step 1. http://i49.tinypic.com/2nkph6c.png

step 2. http://i45.tinypic.com/2upckgk.png

Answer 1

[Image Analyst]

Sum the image horizontally using sum(yourImage, 2) and look for where the signal goes from dark to white using diff().

[edward]

can you give me an example image analyst ?

[Image Analyst]

verticalProfile = mean(grayImage, 2);
darkAreas = verticalProfile  < 128; % or whatever value works.
bottomEdges = diff(darkAreas);
bottomTextLineRows = find(bottomEdges == -1)

or something like that.

[edward]

i mean how i can found the value of topline,upperline,base line,bottom line middle zone and lower zone from that words ?

[Image Analyst]

Just plot verticalProfile and I think you can figure out what to do.

[edward]

if i change the image to BW is this code still works ?

because i use BW image and then dilate for remove the salt n pepper noise.. and then when i run this code i can't find the answer

[Image Analyst]

You don't need to worry about salt and pepper noise - it won't be a significant factor at all when you're taking the mean over the whole width of the image. Just take the mean of the grayscale image.

[edward]

i use this code.. clc

clear

read= imread('apodio3.jpg');

%%Grayscalling

readgray = rgb2gray(read);

%convert to binary

readbw = im2bw(readgray,graythresh(readgray));

%========================================

%========================================

Iskel = bwmorph(readbw,'dilate');

%========================================

%========================================

[row column] = size(Iskel);

right_boundary =[]

for column1=1:column

    for row1=1:row

        if Iskel(row1,column1)==0     

          right_boundary=column1;

            end

    end

end

for column2=column:-1:1

    for row2=1:row

    if Iskel(row2,column2)==0

        left_boundary=column2;

        end

    end

end

for row3=1:row

     for column3=1:column

          if Iskel(row3,column3)==0

              lowest_boundary=row3;

            end

        end

end

    for row4=row:-1:1

        for column4=1:column

            if Iskel(row4,column4)==0

                highest_boundary=row4;

            end

        end

    end

   height=((lowest_boundary-highest_boundary));

   width=((right_boundary-left_boundary));

A = imcrop(Iskel,[left_boundary highest_boundary width

height]);%%position vector, [xmin ymin width height].

Image = imresize (A, [100,1000]);

figure,imshow (Image), title ('Citra setelah proses Crop');

[row column] = size(Image)

[uu]=sum(Image,2)

verticalProfile = mean(Image,2);

darkAreas = verticalProfile < 128; % or whatever value works.

bottomEdges = diff(darkAreas);

bottomTextLineRows = find(bottomEdges == -1)

[edward]

i use this image

http://i45.tinypic.com/34npiqr.jpg

[Image Analyst]

Sorry - I'm leaving for the weekend in just a few minutes. Good luck with it. Look here for algorithms: http://iris.usc.edu/Vision-Notes/bibliography/contentschar.html#OCR,%20Document%20Analysis%20and%20Character%20Recognition%20Systems

[edward]

is anybody can help me ?