Method for computing optical flow under the epipolar constraint
View Patent ↗Point matches between images within an image sequence are identified by sparse optical flow computation and employed to compute a fundamental matrix for the epipolar geometry, which in turn is employed to derive an epipolar geometry constraint for computing dense optical flow for the image sequence. The epipolar geometry constraint may further be combined with local, heuristic constraints or robust statistical methods. Improvements in both accuracy and performance in computing optical flow are achieved utilizing the epipolar geometry constraint.
1. A system for computing optical flow between images within an image sequence comprising:
an image processor processing the image sequence, wherein the image processor: derives epipolar geometry for the images from point matches between the images; and
computes optical flow for each pixel within at least one of the images under a constraint derived from the epipolar geometry by computing a component of the optical flow using a median.
2. The system according to claim 1 , wherein the image processor, in deriving the epipolar geometry for the images, computes sparse optical flow between the images.
3. The system according to claim 1 , wherein the image processor, in computing optical flow for each pixel within at least one of the images, employs a constraint derived from a fundamental matrix between the images.
4. The system according to claim 1 , wherein the image processor utilizes the constraint derived from the epipolar geometry in combination with least squares minimization to compute optical flow for each pixel within at least one of the images.
5. The system according to claim 1 , wherein the image processor utilizes the constraint derived from the epipolar geometry in combination with robust statistical methods to compute optical flow for each pixel within at least one of the images.
6. The system according to claim 1 , wherein the image processor computes optical flow u,v for each pixel within at least one of the images from I x u+I y v+I t =0, where I x , I y , and I t are known spatio-temporal derivatives of image intensity at each pixel within the at least one image, and a x,y u+b x,y v+c x,y =0, where a x,y , h x,y and c x,y are derived from a fundamental matrix F between the images.
7. The system according to claim 1 , wherein the image processor computes dense optical flow between the images.
8. A system for computing optical flow between images within an image sequence comprising:
a video receiver including an input for receiving the image sequence;
an image processor within the video system processing the image sequence, wherein the image processor:
derives epipolar geometry forte images from point matches between the images; and
computes optical flow for each pixel within at least one of the images under a constraint derived from the epipolar geometry by computing a component of the optical disc flow using a median.
9. The system according to claim 8 , wherein the image processor, in deriving the epipolar geometry for the images, computes sparse optical flow between the images.
10. The system according to claim 8 , wherein the image processor, in computing optical flow for each pixel within at least one of the images, employs a constraint derived from a fundamental matrix between the images.
11. The system according to claim 8 , wherein the image processor utilizes the constraint derived from the epipolar geometry in combination with least squares minimization to compute optical flow for each pixel within at least one of the images.
12. The system according to claim 8 , wherein the image processor utilizes the constraint derived from the epipolar geometry in combination with robust statistical methods to compute optical flow for each pixel within at least one of the images.
13. The system according to claim 8 , wherein the image processor computes optical flow u,v for each pixel within at least one of the images from I x u+I y v+I t =0, where I x , I y , and I t are known spatio-temporal derivatives of image intensity at each pixel within the at least one image, and a x,y u+b x,y v+c x,y =0, where a x,y , b x,y and c x,y are derived from a fundamental matrix F between the images.
14. The system according to claim 8 , wherein the image processor computes dense optical flow between the images.
15. A method for computing optical flow between images within an image sequence comprising:
deriving epipolar geometry for the images from point matches between the images; and
computing optical flow for each pixel within at least one of the images under a constraint derived from the epipolar geometry by computing a component of the optical flow using a median.
16. The method according to claim 15 , wherein the step of deriving the epipolar geometry for the images from point matches between the images further comprises: computing sparse optical flow between the images.
17. The method according to claim 15 , wherein the step of computing optical flow for each pixel within at least one of the images under a constraint derived from the epipolar geometry further comprises: computing optical flow employing a constraint derived from a fundamental matrix between the images.
18. The method according to claim 15 , wherein the step of computing optical how for each pixel within at least one of the images under a constraint derived from the epipolar geometry further comprises: utilizing the constraint derived from the epipolar geometry in combination with least squares minimization to compute optical flow for each pixel within at least one of the images.
19. The method according to claim 15 , wherein the stop of computing optical flow for each pixel within at least one of the images under a constraint derived from the epipolar geometry further comprises: utilizing the constraint derived from the epipolar geometry in combination with robust statistical methods to compute optical flow far each pixel within at least one of the images.
20. The method according to claim 1 , wherein the step of computing optical flow for each pixel within at least one of the images under a constraint derived from the epipolar geometry further comprises:
computing optical flow u,v for each pixel within at least one of the images from I x u+I y v+I t =0, where I x , I y , and I t are known spatio-temporal derivatives of image intensity at each pixel within the at least one image, and a x,y u+b x,y v+c x,y =0, where a x,y , b x,y and c x,y are derived from a fundamental matrix F between the images.
21. The system of claim 1 , wherein the computation of the optical flow comprises of:
u=median (u i )
i=1, . . . ,n
where u is a component of the optical flow vector and n is the number of optical flow constraint lines.
22. The system of claim 8 , wherein the computation of the optical flow comprises of:
u=median (u i )
i=1, . . . ,n
where u is a component of the optical flow vector and n is the number of optical flow constraint lines.
23. The system of claim 15 , wherein the computation of the optical flow comprises of:
u=median (u i )
i=1, . . . ,n
where u is a component of the optical flow vector and n is the number of optical flow constraint lines.