Presentation is loading. Please wait.

Presentation is loading. Please wait.

Mutual Information Narendhran Vijayakumar 03/14/2008.

Similar presentations


Presentation on theme: "Mutual Information Narendhran Vijayakumar 03/14/2008."— Presentation transcript:

1 Mutual Information Narendhran Vijayakumar 03/14/2008

2 Definition Mutual Information – I(A,B) Similarity measure – Amount of information B contains about A Registration metric – Maximum MI corresponds to Perfect Alignment – Amount of information B contains about A is maximum when the images are aligned perfectly 2

3 Mathematical Definition I(A,B) = H(A)+H(B)-H(A,B) – A is the reference image – B is the floating image – H(.) is entropy – I(.,.) is mutual information 3

4 Entropy H(A) is defined as – H(A) = -∑p i log 2 p i – P i – Probability of the intensities in an image Measure of dispersion of pdf Measure of intensity 4

5 Joint Probability 1234 4981 7627 0513 4567 713124 1110511 3846 Image 1 Image 2 5

6 Joint Probability-Registered image 012345678910111213 01 13 22 33 43 53 61 71 81 91 10 11 12 13 6

7 Misaligned image 567X 13124X 10511X 846X 7

8 Joint Probability-misregistered images 012345678910111213 01 111 211 31 41 51 61 71 81 91 10 11 12 13 8

9 Conclusion Dispersion Increases (for misaligned images) Joint Probability Value Increases I(A,B) reduces Registration is achieved by minimizing H(A,B) 9


Download ppt "Mutual Information Narendhran Vijayakumar 03/14/2008."

Similar presentations


Ads by Google