Error propogation and variable-bandwidth mean shift for feature space analysis
View Patent ↗The present invention comprises using error propagation for building feature spaces with variable uncertainty and using variable-bandwidth mean shift for the analysis of such spaces, to provide peak detection and space partitioning. The invention applies these techniques to construct and analyze Hough spaces for line and geometrical shape detection, as well as to detect objects that are represented by peaks in the Hough space. This invention can be further used for background modeling by taking into account the uncertainty of the transformed image color and uncertainty of the motion flow. Furthermore, the invention can be used to segment video data in invariant spaces, by propagating the uncertainty from the original space and using the variable-bandwidth mean shift to detect peaks. The invention can be used in a variety of applications such as medical, surveillance, monitoring, automotive, augmented reality, and inspection.
1. A method for modeling a background of a video image, the method comprising:
generating, by a processor, a combined covariance matrix from covariances for colors of the video image and an optical flow of the input image;
estimating, by the processor, a probability density from performing a kernel density estimation on the combined covariance matrix; and
updating, by the processor, the background model using the estimated probability density.
2. The method of claim 1 , further comprising adding a vector of features of a video frame to said background model.
3. The method of claim 1 , further comprising detecting a change between a video frame and the video image by evaluating a vector of features of the video frame and said background model.
4. The method of claim 3 , wherein detecting the change includes determining if said vector of features exceeds a threshold and adjusting said background model accordingly.
5. The method of claim 3 , further comprising applying morphological operations to said detections.
6. A non-transitory tangible computer readable storage medium tangibly embodying a program of instructions executable by the machine to perform method steps for modeling a background of an video image the method steps comprising:
generating a combined covariance matrix from a covariance for colors of the video image and a covariance of an optical flow of the video image;
estimating a probability density from performing a kernel density estimation on the combined covariance matrix; and
updating the background model using the estimated probability density.
7. The computer readable storage medium of claim 6 , wherein the method further comprises adding a vector of features of a video frame to said background model.
8. The computer readable storage medium of claim 6 , wherein the method further comprises detecting a change between a video frame and the video image by evaluating a vector of features of the video frame and said background model.
9. The computer readable storage medium of claim 8 , wherein detecting the change includes determining if said vector of features exceeds a threshold and adjusting said background model accordingly.
10. The program computer readable storage medium of claim 8 , wherein the method further comprises applying morphological operations to said detections.
11. The method of claim 1 , wherein generating the combined covariance matrix comprises:
performing an illumination invariant transformation on the colors: and
generating a combined covariance matrix from a covariance of the transformation and the covariance of the optical flow.
12. The method of claim 1 , wherein estimating the probability density comprises performing a variable bandwidth mean shift on the video image using the combined covariance matrix.
13. The method of claim 12 , wherein the variable bandwidth mean shift uses a product kernel with a variable bandwidth for coordinates of the colors.
14. The computer readable storage medium of claim 6 , wherein generating the combined covariance matrix comprises:
performing an illumination invariant transformation on the colors; and
generating a combined covariance matrix from a covariance of the transformation and the covariance of the optical flow.
15. The computer readable storage medium of claim 6 , wherein estimating the probability density comprises performing a variable bandwidth mean shift on the video image using the combined covariance matrix.
16. The computer readable storage medium of claim 15 , wherein the variable bandwidth mean shift uses a product kernel with a variable bandwidth for coordinates of the colors.
17. A method for modeling a background of a video image, the method comprising:
generating, by a processor, a combined covariance matrix from covariances of a joint motion-color feature space of the video image;
estimating, by the processor, a probability density from performing a kernel dens estimation on the combined covariance matrix; and
updating, by the processor, the background model using the estimated probability density.
18. The method of claim 17 , wherein generating the combined covariance matrix comprises:
performing an illumination invariant transformation on colors of the feature space; and
generating a combined covariance matrix from a covariance of the transformation and a covariance of motions of the feature space.
19. The method of claim 18 , wherein estimating the probability density comprises performing a variable bandwidth mean shift on the video image using the combined covariance matrix.
20. The method of claim 19 , wherein the variable bandwidth mean shift uses a product kernel with a variable bandwidth for coordinates of colors of the feature space.