Adaptive image acquisition for multiframe reconstruction

1. A method for multiframe reconstruction of an object comprising:

acquiring a set of K images of the object at different acquisition settings, where K≧2;

for a (k+1)st of the K acquired images, where k≧1, after acquiring the first k images, determining the acquisition setting for the (k+1)st image based at least in part on selecting the acquisition setting to optimize a cost function, wherein one component of the cost function is based on an error between (a) an estimated image ŝ k+1 that is a multiframe reconstruction of the object from the (k+1) acquired images and (b) an ideal image; and

applying multiframe reconstruction to the set of K images, wherein the multiframe reconstruction estimates the image ŝ K from the K acquired images based on minimizing the error between the estimated image ŝ K and the ideal image.

2. The method of claim 1 wherein K≧3 and the step of determining an acquisition setting for the (k+1)st image is performed only for images after the first two initial images.

3. The method of claim 2 wherein, for every image acquired after the initial images, the step of determining the acquisition setting for said image does not depend on whether any additional images will be acquired after said image.

4. The method of claim 2 wherein, for at least one of the images acquired after the initial images, the step of determining the acquisition setting for said image is based at least in part on assuming that at least one additional image will be acquired after said image.

5. The method of claim 2 wherein, for every image acquired after the initial images, the step of determining the acquisition setting for said image is based at least in part on assuming a final set containing K images.

6. The method of claim 1 wherein the step of determining the acquisition setting for said (k+1)st image is based at least in part on increasing an effective signal-to-noise ratio.

7. The method of claim 1 wherein the step of determining the acquisition setting for said (k+1)st image comprises determining the acquisition setting based at least in part on increasing information captured by said (k+1)st image, compared to information captured by the previously acquired k images.

8. The method of claim 7 wherein said information is a measure of a predicted sharpness of the multiframe reconstructed image.

9. The method of claim 7 wherein said information is a measure of a change in local smoothness related to object depth.

10. The method of claim 1 wherein the step of determining the acquisition setting for the (k+1)st image further comprises:

determining the acquisition setting for said (k+1)st image based at least in part on reducing a total power consumption.

11. The method of claim 1 wherein the step of determining the acquisition setting for said (k+1)st image is based on modeling a relationship between the acquisition settings and an optical point spread function.

12. The method of claim 1 where the step of determining the acquisition setting for the (k+1)st image is based on maximizing a weighted sum of total sharpness over a collection of image regions in the previously acquired k images.

13. The method of claim 1 wherein the acquisition setting includes an aperture setting.

14. The method of claim 1 wherein the acquisition setting includes a focus setting.

15. The method of claim 1 wherein the acquisition setting includes an exposure setting.

16. The method of claim 1 wherein the object is modeled as a multi-depth object.

17. The method of claim 16 wherein the step of determining the acquisition setting for the (k+1)st image comprises:

estimating different object depths for different tiles of the multi-depth object, based at least in part on content of previously acquired images; and

determining the acquisition setting for the (k+1)st image based at least in part on the estimated object depths.

18. The method of claim 17 wherein the step of determining the acquisition setting for the (k+1)st image is based at least in part on a weighted average over the tiles of a merit function evaluated for each tile.

19. The method of claim 17 wherein the step of determining the acquisition setting for the (k+1)st image is based at least in part on a worst value for a merit function evaluated for each tile.

20. The method of claim 16 wherein the step of determining the acquisition setting for the (k+1)st image comprises:

estimating different object depths for different parts of the multi-depth object, based at least in part by applying a filter bank to the content of previously acquired images; and

determining the acquisition setting for the image based at least in part on the estimated object depths.

21. The method of claim 1 wherein the multiframe reconstruction step produces an estimated depth map.

22. A multiframe reconstruction system comprising:

an image capture device that acquires a set of K images at various acquisition settings, where K≧2;

an adaptive acquisition module coupled to the image capture device to receive the acquired images, the adaptive acquisition module (a) instructing the image capture device to acquire a set of at least two images of an object at different acquisition settings, and (b) for a (k+1)st of the K acquired images, where k≧1, after acquiring the first k images, determining the acquisition setting for the (k+1)st image based at least in part on selecting the acquisition setting to optimize a cost function, wherein one component of the cost function is based on an error between (a) an estimated image ŝ k+1 that is a multiframe reconstruction of the object from the (k+1) acquired images and (b) an ideal image; and

a multiframe reconstruction module coupled to the adaptive acquisition module, for applying multiframe reconstruction to the set of K images, wherein the multiframe reconstruction estimates the image ŝ K from the K acquired images based on minimizing the error between the estimated image ŝ K and the ideal image.

23. A non-transitory computer readable medium containing software instructions that, when executed on a computerized system, cause the computerized system to:

cause the acquisition of a set of K images of an object at different acquisition settings, where K≧2;

receive the acquired images; and

for a (k+1)st of the K acquired images, where k≧1, after acquiring the first k images, determine the acquisition setting for the (k+1)st image based at least in part on selecting the acquisition setting to optimize a cost function, wherein one component of the cost function is based on an error between (a) an estimated image ŝ k+1 that is a multiframe reconstruction of the object from the (k+1) acquired images and (b) an ideal image;

wherein an estimated image ŝ K from the K acquired images is estimated by applying a multiframe reconstruction that is based on minimizing the error between the estimated image ŝ K and the ideal image.

24. The method of claim 1 wherein the estimated image ŝ K is based on a multiframe Wiener solution.

25. A method for multiframe reconstruction of an object comprising:

acquiring a set of K images of the object at different acquisition settings, where K≧2;

for a (k+1)st of the K acquired images, where k≧1, after acquiring the first k images, determining the acquisition setting for the (k+1)st image based at least in part on selecting the acquisition setting to optimize a cost function, wherein one component of the cost function is a Cramer-Rao performance bound on the mean-square-error between (a) the estimated image ŝ K that is a multiframe reconstruction of the object and (b) an ideal image; and

applying multiframe reconstruction to the set of K images, wherein the multiframe reconstruction estimates the image ŝ K from the K acquired images based on a multiframe Wiener solution that minimizes the error between the estimated image ŝ K and the ideal image.