Template Matching between two binary image.

Sir,

I have two binary images. One is the main image and another is for template image. Now i want to match them to see if they are equal or not. What will be the best approach to do that for this kind of problem? For your well understanding i attach two images

this is the main binary image

And this is the template image

Please help sir.

Answer 1

[Image Analyst]

Try using Hu's moments: http://www.youtube.com/watch?v=Nc06tlZAv_Q Hu's moments for the image will match if the image is rotated, translated, scaled, etc.

[Asif Hasan]

Thanks for reply sir. I have seen the video that you referred but i only want to match if (main image=template image) or not because its a binary matrix i can just match main image 1 = template image 1 if i am not wrong . @Image analist

[Image Analyst]

If you know for a fact that this, or some other, image will be an exact match pixel for pixel, then you can simply subtract them and check that all pixels are zero. If it's a pixel for pixel match but maybe just translated, then you can use normalized cross correlation, like in the attached demo. To me this looks like some biometric application that someone posted before about identifying individuals based on the blood vessel pattern in the sclera of their eye. If that's the case then neither of those two approaches would work well.

[Asif Hasan]

Your guess is right i am working on a thesis of identifying individual and i am getting this image of blood vessel of sclera but i can't understand how i will match these images with template image . I want to save these images in a database and next time i want to check using this image with the database if it's in my database or not. Can you help me with that , sir? i already used a template matching matlab code but there is a color or gray scale image as a input but my image is binary what approach should i take? @image Analist

[Image Analyst]

I already suggested Hu's moments. I don't know why you rejected that. I actually haven't used it myself but it looks promising. There are other methods. Look up methods in Vision Bib on pattern recognition. Do a search on CBIR which uses a large variety of methods to match features and objects. I can't help much more, especially since soon I will be traveling for most of today and tomorrow.