Binary Image similarity classification

I have number of binary images and I want to classify these images into groups based on its similarities. So the result is number of groups such as group1 contains img1 and img6, group2 contains img2 and img5, and group3 includes img3 and img4. The following are some of the images that I work on. These images represent the vertical projection of some letters and I want to classify similar letters into a group. * * _ So is there any tool or program can do the above functionality._**

[Image Analyst]

I think you'd be better off not using the images you posted and instead use the 1-D arrays themselves - what you got from getting the profiles.

[Mohammad]

Thanks for your reply, You mean that I should use the vector that I depend on to draw these figure? So is there any idea how to compare these vectors? Or is there any tool or program to do that?

[Image Analyst]

Yes, you don't want to compare screenshots of bar charts as if they were images. You want to consider the actual data that was plotted, not the plots themselves. Like I said in my answer below, you might want to consider RNN.

Answer 1

[Image Analyst]

Try recurrent neural networks: RNN