I want to make a 3x2 matrix on an image

I want to make a 2x3 matrix on each component of an image. Attached is the picture.

Actually, I want to make a matrix, so that each dot in the picture lies in different index of the matrix. I need code. Kindly help please!

[Guillaume]

Your patterns appear to be 2x2 or 3x2 (rows by columns is the matlab convention), not 2x3.

An example of the desired output for the above input would clarify what you need.

[Zaheer Abbas]

You guessed it right. My mistake. I want 3x2 matrix. This is my actual image. These are braille dots. In Braille, Each letter is composed with the combination of 6 cells. So I have cropped each letter of Braille and made it a single component/image.

This is the cropped image.

Now I want to make a 3x2 matrix on each. so that each dot lies in each index of a matrix. Hope it is clear now! If not then you can ask more. I Need help!

Answer 1

[Image Analyst]

Get a library of all possible dot patterns then just paste onto the desired location in your image. See my copy and paste code, attached.

[Zaheer Abbas]

I don't get this. This code is copying and pasting the selected area on the same image. How it can be useful in my work? And also from where i can find library of all possible dot patterns?

[Image Analyst]

You already have a bunch of examples of dots, or complete dot patterns. You said you "want to make a 2x3 matrix on each component of an image", so I assume you have a blank image and you want to create dots on that image, right? So why is that not a pasting operation? I don't know what "component" means to you but I assume it means a color channel, so just paste onto the red, green, and blue color channels.

[Guillaume]

My understanding is that he wants a 3x2 (not 2x3 as stated) binary matrix for each dot pattern (component?) indicating the location of the braille dots. e.g, for the first pattern

 [1 1
  0 1
  0 0]

That's just a guess. The question is very unclear indeed. It's not even clear if the patterns are in the same image (and thus need to be located) or in different images.

[Zaheer Abbas]

You guessed it right. My mistake. I want 3x2 matrix. This is my actual image. These are braille dots. In Braille, Each letter is composed with the combination of 6 cells. So I have cropped each letter of Braille and made it a single component/image.

This is the cropped image.

Now I want to make a 3x2 matrix on each. so that each dot lies in each index of a matrix. Hope it is clear now! If not then you can ask more. I Need help!

[Image Analyst]

It's still not clear, since I don't know what it means to make a matrix ON an image. Now, another possibility of "make" (instead of synthesizing a new image which I first thought it meant) is to "analyze" the cropped images of the patterns and decide which location has a dot or no dot and then determine which letter/character the pattern represents. So please clarify - do you want to make/synthesize/create a new image, or do you want to analyze the existing image, and cropped portions of it, to determine which pattern is there?

By the way, there are a number of Braille image analysis publications cited here: http://www.visionbib.com/bibliography/char1030.html#KK10300

[Zaheer Abbas]

"analyze the cropped images of the patterns and decide which location has a dot or no dot and then determine which letter/character the pattern represents."

Yes, exactly, this is the case ! After analyzing the pattern, I will compare the pattern with dictionary i have made, in which 1 represents dot and 0 represents no dot.

[Image Analyst]

Use stdfilt() or imbinarize() on the cropped image to get high signal where there are spots. Then divide the cropped image up into zones and inspect each of the 6 zones to see if the signal is high enough to indicate a spot or not. If you can't figure it out, then upload an image of just one little cropped portion and maybe tonight I might get time to look at it.

[Zaheer Abbas]

Thanks a lot :) Here is the cropped image.

I am also trying.

Ps. I am a newbie

[Zaheer Abbas]

This is how I crop the image. I joined the nearest dots and made it a single component by using imdilate with disk strel of 30 size on binary image. Then crop each component into a single image using their Bounding Boxes.

[Image Analyst]

You'll need to find whole rows, because otherwise you if you have, say, a pair of dots, you won't know if it's in the first row, second row, or third row. You won't know that with your method of dilating and taking the bounding box. So you're going to have to find tops and bottoms of the whole line of characters first. Then you can do what you said to get the right and left sides. THEN you can crop the bounding boxes, and then divide into 6 zones to determine what zones have a spot in them.

[afif fadilah]

sir zaheer abbas, would do you like to give your code ?