-4

I was a student, has just begun the matlab Matlab would like to ask, I hope someone can answer

Now there are three images are A, B, C b is a diagram capture down on a small area, it may be a solid color block, c is high similarity with a graph picture may be some position moves slightly changed,

How to use matlab to compare and identify the relative position of the b Figure c Figure :( ? Such as coordinates or marked.

The problem probably like this picture

image

Amro
  • 123,847
  • 25
  • 243
  • 454
user2359356
  • 3
  • 1
  • 5
  • 1
    it is not very clear what you are asking, is this related to [image registration](http://en.wikipedia.org/wiki/Image_registration)? Please improve your question and use relevant tags.. – Amro May 07 '13 at 17:41

1 Answers1

0

It sounds like you're looking for template matching (find a smaller template image in a larger image).

Check out this previous answer of mine that shows how to use Normalized Cross-Correlation to recover the translation offset.

demo

Community
  • 1
  • 1
Amro
  • 123,847
  • 25
  • 243
  • 454
  • It should be said that, I want to let the user picture A a marquee, The selected block is the picture B, Pictures B to compare the follow-up of all the pictures and find picture B "in the other pictures, such as a tag or value of the output coordinates – user2359356 May 08 '13 at 06:44
  • um ...... First of all, the picture C inside six color blocks, shoulders, elbows, hands, waist, knees, ankles. At present, I have a lot of pictures with the picture C similar picture,the difference may be a slight change of action figures Suppose there are 3,000 pictures,now,I want to target is when I circle the block you want to mark, whether it is the red block of the shoulder or elbow purple block, Matlab can automatically mark or identify the remaining 2900 picture, red block of the shoulder or elbow purple block, EX:mark or output x, y values – user2359356 May 08 '13 at 08:08
  • alright, have you tried the `normxcorr2` I mentioned above? It will take the template image, convolve it with the reference image. You then take find the point of highest correlation, corresponding to where it matches the image best... You repeat this process for all 3000 images. – Amro May 08 '13 at 08:16
  • Earlier research did not understand :)) I will carefully study, Thank you for your answers – user2359356 May 08 '13 at 14:21