extract text from image

this code extract object random i want to extract letter line by line sequential

%%Image segmentation and extraction
%%Read Image
imagen=imread('image_a.jpg');
%%Show image
figure(1)
imshow(imagen);
title('INPUT IMAGE WITH NOISE')
%%Convert to gray scale
if size(imagen,3)==3 % RGB image
  imagen=rgb2gray(imagen);
end
%%Convert to binary image
threshold = graythresh(imagen);
imagen =~im2bw(imagen,threshold);
%%Remove all object containing fewer than 30 pixels
imagen = bwareaopen(imagen,30);
pause(1)
%%Show image binary image
figure(2)
imshow(~imagen);
title('INPUT IMAGE WITHOUT NOISE')
%%Label connected components
[L Ne]=bwlabel(imagen);
%%Measure properties of image regions
propied=regionprops(L,'BoundingBox');
hold on
%%Plot Bounding Box
for n=1:size(propied,1)
  rectangle('Position',propied(n).BoundingBox,'EdgeColor','g','LineWidth',2)
end
hold off
pause (1)
%%Objects extraction
figure
for n=1:Ne
  [r,c] = find(L==n);
  n1=imagen(min(r):max(r),min(c):max(c));
  imshow(~n1);
  pause(0.5)
end

[Image Analyst]

ahmed hasn't been here in 3 years. You'll have to do it without him. Why not use text() or line()?

Answer 1

[Image Analyst]

It's not random, though you don't know the algorithm it uses (it's top left to bottom right, column by column).

You will need to find each line and extract it before you label. See attached demo, which will produce this:

There is not OCR package built in to MATLAB. You'll have to find one or do it yourself.

[Image Analyst]

By the way, you should probably start officially Accepting some of your old questions so we know you're done with them. You haven't accepted a single question yet.

[Waheed Ullah]

Its a good code for text extraction but respected sir I want to extract faces from an image, so please share that code. Thanks.

[Image Analyst]

See this link: http://www.mathworks.com/help/vision/examples/face-detection-and-tracking-using-camshift.html?prodcode=VP&language=en

[malik nouman]

respected sir i want to extract text from natural images will this code work? plz suggest any technique for text extraction.

[Image Analyst]

No, you'd need the code in this example: http://www.mathworks.com/help/vision/examples/automatically-detect-and-recognize-text-in-natural-images.html?prodcode=VP&language=en

[Srika]

respected sir.i need to save this segmented text line in seperate folder...while i tried with my code.I got this error

Error using horzcat Dimensions of matrices being concatenated are not consistent.

Error in textext (line 70) imwrite(thisLine,['Datasets/',num2str(thisLine),'.jpg']);

[Image Analyst]

Try this:

whos thisLine
fprintf('thisLine = %s\n', thisLine);
fileName = sprintf('Datasets/%s.jpg', thisLine)
whos fileName
fprintf('fileName = %s\n', fileName);
imwrite(thisLine, fileName);

What do you see in the command window?

[Srika]

Sir can i get any other images for this code?

[Image Analyst]

Try a web search for images that have text in them.

[Srika]

i had searched for images.Where text is not segmented in all images by this code.

[SATISH KUMAR]

this code is not working with my 1024x1024 block of document image. this is my document image. is there any changes to be made to above code so that i can extract the words from my document image.the characters are extracting but i need words and text line also.

[Image Analyst]

OK Satish - fess up. You didn't even try to adjust the thresholds for your image, did you? When you do that, it works fine. Adjusted code is attached.

[SATISH KUMAR]

thank you Sir,

[Walter Roberson]

oza san comments to Image Analyst:

Dear sir here i am again to ask you some questions having the code in the link how i can make it for multilingual and multi font texts can i use it as it is??and can i use it for also artificial texts.i know i can but i have a little doubt.

[Image Analyst]

If it doesn't work for your language, then you'll just have to tweak it. There's no guarantee that it works right out of the box for every language.

[oza san]

it works fine but it looks like size dependent...i have to adjust every time i gave images.am i right or there is another way to make it size independent!!!

[Walter Roberson]

oza san comments to me:

Dear Walter R. here you post my questions due to i think it doesn't make sense.here is what i want to ask i want to detect segment and identify scripts from video images using the code in the link http://www.mathworks.com/help/vision/examples/automatically-detect-and-recognize-text-in-natural-images.html?prodcode=VP&language=en..so can it handle what i state in the original question.. mind you if you can help me!!!

[Walter Roberson]

oza san: you should only "flag" something if you need to bring it to the attention of the moderators, such as if it is spam or contains inappropriate language. Otherwise you should just add a comment. I moved your remarks out of "flags" into comments.

[Walter Roberson]

Can it handle what you asked before? No.

[oza san]

ok.sorry for the mistake and thank you for your constructive comments but i test it with some images with multilingual texts and works fine with some errors.what i face is when the fonts are different it is too much poor.is there anyway that helps me to handle it?

Answer 2

[deeksha h r]

respected sir,the code used is displaying segmented letter but we need segmented line to be displayed ie each text line from the image is to be displayed.can you please help us with this code

Answer 3

[Arvinder singh]

https://www.youtube.com/watch?v=qzjE-4yRMCA

Answer 4

[Sharad Sirsat]

You dont need to used "imagen = bwareaopen(imagen,30);" over here, Simply after converting image into binary,Use " bwlabel" to count number of character/objects in an image,After that in an image you can find the centroid of each character and then using Regionprops find their Length, Width to create a bounding box around character.Once you have done bounding box you can crop character individually by giving original image input.

Note: For centroid, bounding box(regionprops) and for imagecrop apply forloop till the counted no of objects/characters.

[Image Analyst]

But if you don't use bwareaopen(), you'll get a lot of small noise specks that are not letters. Who wants to deal with those????

Plus, if you don't do it like I did, line-by-line, then you'll have labels that are randomly (almost) chosen from each line. To avoid that you'd have to go through some special code, like get the y coordinates of each blob, and use something like kmeans() to find out which blobs are in the same line of text.