Estimating pixel variances in the scenes of staring sensors

1. A computer implemented method, comprising:

acquiring a reference image frame and a current image frame of a scene with a camera;

generating a raw difference frame based upon differences between the reference image frame and the current image frame;

generating pixel error estimates for each pixel in the raw difference frame based at least in part upon spatial error estimates related to spatial intensity gradients in the scene; and

using the pixel error estimates to mitigate effects of camera jitter between the reference image frame and the current image frame, wherein using the pixel error estimates to mitigate the effects of the camera jitter comprises scaling pixel intensity values in the raw difference frame with the pixel error estimates.

2. The computer implemented method of claim 1 , wherein scaling the pixel intensity values in the raw difference frame with the pixel error estimates comprises:

dividing each of the pixel intensity values within the raw difference frame by its corresponding pixel error estimate to generate normalized pixel values; and

generating a normalized difference frame including the normalized pixel values.

3. The computer implemented method of claim 2 , further comprising detecting changes in the scene between the current image frame and the reference image frame based upon the normalized difference frame.

4. The computer implemented method of claim 1 , wherein the pixel error estimates are computed by interpolating pixel intensities between adjacent pixels in a single image frame and assuming a jitter distribution, which is manifested as rigid translations over an entire image frame.

5. The computer implemented method of claim 1 , wherein the pixel error estimates comprise hybrid error estimates computed based upon a combination of the spatial error estimates that are related to intensity gradients in the scene and temporal error estimates based upon a history of past image frames of the scene.

6. The computer implemented method of claim 5 , wherein the combination of the spatial error estimates and the temporal error estimates comprises using a larger value between the spatial error estimates and the temporal error estimates for each pixel.

7. A computer-accessible non-transitory storage medium that provides instructions that, if executed by a machine, will cause the machine to perform operations comprising:

acquiring a reference image frame of a scene with a staring sensor;

acquiring a current image frame of the scene with the staring sensor;

generating a raw difference frame based upon differences between the reference image frame and the current image frame;

generating pixel error estimates for each pixel in the raw difference frame based at least in part upon spatial error estimates related to spatial intensity gradients in the scene, wherein the pixel error estimates are computed by interpolating pixel intensities between adjacent pixels; and

using the pixel error estimates to mitigate effects of camera jitter of the staring sensor between the reference image frame and the current image frame, wherein using the pixel error estimates to mitigate the effects of the camera jitter comprises scaling pixel intensity values in the raw difference frame with the pixel error estimates.

8. The computer-accessible storage medium of claim 7 , wherein scaling the pixel intensity values in the raw difference frame with the pixel error estimates comprises:

dividing each of the pixel intensity values within the raw difference frame by its corresponding pixel error estimate to generate normalized pixel values; and

generating a normalized difference frame including the normalized pixel values.

9. The computer-accessible storage medium of claim 7 , wherein using the pixel error estimates to mitigate the effects of the camera jitter comprises:

comparing pixel intensity values in the raw difference frame to the pixel error estimates to determine if the pixel intensity values fall outside their corresponding pixel error estimates.

10. The computer-accessible storage medium of claim 7 , wherein the pixel error estimates are computed by interpolating pixel intensities between adjacent pixels of a single image frame and assuming a known jitter distribution, which is manifested as rigid translations over an entire image frame.

11. The computer-accessible storage medium of claim 7 , wherein the pixel error estimates comprise hybrid error estimates computed based upon a combination of the spatial error estimates that are related to intensity gradients in the scene and temporal error estimates based upon a history of past image frames of the scene.

12. The computer-accessible storage medium of claim 11 , wherein the combination of the spatial error estimates and the temporal error estimates comprises using a larger value between the spatial error estimates and the temporal error estimates for each pixel.

13. A computer implemented method for detecting changes in a scene perceived by a staring sensor, the computer implemented method comprising:

acquiring a reference image frame of the scene;

acquiring a current image frame of the scene;

generating a raw difference frame based upon the reference image frame and the current image frame; and

scaling pixel values within the raw difference frame with pixel error estimates for the pixel values to generate scaled pixel values, wherein the pixel error estimates are computed based at least in part upon spatial error estimates determined from a single image frame and which are related to spatial intensity gradients in the scene.

14. The computer implemented method 13 , further comprising

thresholding the scaled pixel values;

determining if a change detection event has been perceived in the scene based upon the thresholding; and

reporting the change detection event if the change detection event is determined to have occurred.

15. The computer implemented method of claim 13 , wherein scaling the pixel values within the raw difference frame with the pixel error estimates for the pixel values comprises:

dividing each of the pixel values within the raw difference frame by its corresponding pixel error estimate to generate normalized pixel values; and

generating a normalized difference frame including the normalized pixel values.

16. The computer implemented method of claim 15 , further comprising scaling all of the normalized pixel values by a single constant to adjust for camera jitter of the staring sensor.

17. The computer implemented method of claim 13 , wherein the pixel error estimates comprises standard deviation estimates for the pixel values within the raw difference frame and wherein the spatial error estimates comprise spatial standard deviation estimates.

18. The computer implemented method of claim 17 , wherein the spatial standard deviation estimates are computed by interpolating pixel intensities between adjacent pixels and assuming that camera jitter displacements are Gaussian and manifested as rigid translations over an entire image frame.

19. The computer implemented method of claim 13 , wherein the pixel error estimates comprise hybrid error estimates computed based upon a combination of the spatial error estimates that are related to intensity gradients in the scene and temporal error estimates based upon a history of past image frames of the scene.

20. The computer implemented method of claim 19 , wherein the combination of the spatial error estimates and the temporal error estimates comprises using a larger value between the spatial error estimates and the temporal error estimates for each pixel.

21. The computer implemented method of claim 13 , wherein the staring sensor is selected from the group consisting of a webcam, a traffic camera, a surveillance camera, a security camera, staring space-based sensor, and a radiometric sensor.

22. The computer implemented method of claim 13 , wherein the pixel error estimates are computed at full frame rates of the staring sensor.