Incorporating prior information from pre aligned image pairs into EMST-based image registration

1. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for automatic image registration, the method steps comprising:

providing a previously aligned image pair;

extracting a first number samples of features from the previously aligned image pair;

extracting a second number of the samples of features from an observed image pair, each of the observed image pair being different from any of the previously aligned image pair;

determining a Euclidean minimum spanning tree of a union set of the samples of features from the previously aligned image pair and the observed image pair;

determining a similarity measure of the observed image pair based on the Euclidean minimum spanning tree;

estimating a gradient of the similarity measure, wherein a gradient estimate is used to update transformation parameters applied to register the observed image pair; and

outputting a registration of the observed image pair.

2. The method of claim 1 , wherein the previously aligned image pair is one of an image pair earlier in a sequence of registration problems and a training image pair.

3. The method of claim 1 , wherein the gradient of the similarity measure is estimated using a sum of gradients of power weighted edge lengths of the Euclidean minimum spanning tree.

4. The method of claim 1 , wherein the second number of samples of features is periodically extracted.

5. The computer-implemented method of claim 1 , wherein the registration of the observed image pair is determined as:

Q α ( I 1 ,I 2 )= J α,w ( p t ,p o )+λ* H α ( p o ),

where

J α,w (p t ,p o )

is an alpha-Jensen distance, and

λ= I o |/(| I t |+|I o )

is a weight that determines a relative influence of the term H α (p o ) and w =1−λ.

6. The method of claim 1 , wherein the first and second number of samples are one of a same and different number of samples.

7. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for automatic image registration, the method steps comprising:

providing a previously aligned image pair;

extracting a first number samples of features from the previously aligned image pair;

extracting a second number of the samples of features from an observed image pair, each of the observed image pair being different from any of the previously aligned image pair;

determining a Euclidean minimum spanning tree of a union set of the samples of features from the previously aligned image pair and the observed image pair;

determining a similarity measure of the observed image pair based on the Euclidean minimum spanning tree;

estimating a gradient of the similarity measure, wherein a gradient estimate updates a transformation parameter applied to register the observed image pair,

wherein the gradient of the similarity measure is estimated using a sum of gradients of power weighted edge lengths of the Euclidean minimum spanning tree of the union set, including the previously aligned image pair and the observed image pair; and

outputting a registration of the observed image pair determined by an updated transformation parameter.

8. The method of claim 1 , wherein the previously aligned image pair is one of an image pair earlier in a sequence of registration problems and a training image pair.

9. The method of claim 1 , wherein the second number of samples of features is periodically extracted.