Method and apparatus for tracking objects across images

1. A method for tracking an object across images using a processor, the method comprising:

retrieving an object location in a current frame;

determining an appearance signature of the object and motion signature of the object in the current frame;

predicting the new location of the object based on object dynamics;

searching for a location with similar appearance and motion signatures in a next frame utilizing Nedler-Mead Simplex; and

utilizing the location with similar appearance and motion signatures to determine the final location of the object in the next frame.

2. The method of claim 1 further comprising smoothing out the location of the object based on object dynamics.

3. The method of claim 2 , wherein the smoothing out step utilizes a Kalman Filter correction.

4. The method of claim 1 , wherein the step for predicting the new location utilizes a Kalman Filter prediction.

5. The method of claim 1 , wherein the step for searching for the location with similar appearance signature utilizes Simplex search method.

6. The method of claim 5 , wherein the Simplex search method comprises:

computing original appearance and motion signatures from the known location and size of the object in the current frame;

determining candidate locations and sizes in the next frame

computing appearance and motion signatures from these candidate image locations and sizes determined by the Simplex search; and

calculating a distance measure between the original appearance and motion signatures and each new appearance and motion signature computed during the Simplex search.

7. The method of claim 6 , wherein computing the appearance comprises:

computing spatially separated histograms of at least one of the intensity and color values of the pixels corresponding to the image region occupied by the object.

8. The method of claim 6 , wherein the motion signatures comprises:

computing spatially separated histograms of the intensity difference values of the pixels corresponding to the image region occupied by the object.

9. The method of claim 6 , wherein the distance measure comprises:

computing the a Bhattacharya distance between the original appearance histograms and the appearance histograms from image locations determined by the simplex search;

computing the Bhattacharya distance between the original motion histograms and the motion histograms from image locations determined by the simplex search; and

computing a weighted average of the two Bhattacharya distances to arrive at a single distance measure.

10. A non-transitory computer readable medium with an executable program stored thereon, wherein the program instructs a micro processor to perform a method for tracking an object across images, the method comprising:

retrieving object location in a current frame;

determining the appearance signature of the object and motion signature of the object in the current frame;

predicting the new location of the object based on object dynamics;

searching for a location with similar appearance and motion signatures in a next frame utilizing Nedler-Mead Simplex; and

utilizing the location with similar appearance and motion signatures to determine the final location of the object in the next frame.

11. The non-transitory computer readable medium of claim 10 further comprising smoothing out the position of the object based on object dynamics.

12. The non-transitory computer readable medium of claim 11 , wherein the smoothing out step utilizes a Kalman Filter correction.

13. The non-transitory computer readable medium of claim 10 , wherein the step for predicting the new location utilizes a Kalman Filter prediction.

14. The non-transitory computer readable medium of claim 10 , wherein the step for searching for the location with similar appearance signature utilizes Simplex search method.

15. The non-transitory computer readable medium of claim 14 , wherein the Simplex method comprises:

computing original appearance and motion signatures from the known location and size of the object in the current frame;

determining candidate locations and sizes in the next frame

computing appearance and motion signatures from these candidate image locations and sizes determined by the Nedler-Mead Simplex search; and

calculating a distance measure between the original appearance and motion signatures and each new appearance and motion signature computed during the Simplex search.