IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
US-0614841
(2000-07-12)
|
발명자
/ 주소 |
- Moura, Jose' M. F.
- Aguiar, Pedro M. Q.
|
출원인 / 주소 |
- Carnegie Mellon University
|
대리인 / 주소 |
Kirkpatrick & Lockhart LLP
|
인용정보 |
피인용 횟수 :
75 인용 특허 :
4 |
초록
▼
A system for generating a three-dimensional model of an object from a two-dimensional image sequence. According to one embodiment, the system includes an image sensor for capturing a sequence of two-dimensional images of a scene, the scene including the object, a two-dimensional motion filter module
A system for generating a three-dimensional model of an object from a two-dimensional image sequence. According to one embodiment, the system includes an image sensor for capturing a sequence of two-dimensional images of a scene, the scene including the object, a two-dimensional motion filter module in communication with the image sensor for determining from the sequence of images a plurality of two-dimensional motion parameters for the object, and a three-dimensional structure recovery module in communication with the two-dimensional motion filter module for estimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters from the set of two-dimensional motion parameters using a rank 1 factorization of a matrix.
대표청구항
▼
1. A system for generating a three-dimensional model of an object from a two-dimensional image sequence, comprising:an image sensor for capturing a sequence of two-dimensional images of a scene, the scene including the object;a two-dimensional motion filter module in communication with the image sen
1. A system for generating a three-dimensional model of an object from a two-dimensional image sequence, comprising:an image sensor for capturing a sequence of two-dimensional images of a scene, the scene including the object;a two-dimensional motion filter module in communication with the image sensor for determining from the sequence of images a plurality of two-dimensional motion parameters for the object; anda three-dimensional structure recovery module in communication with the two-dimensional motion filter module for estimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters from the set of two-dimensional motion parameters using a rank 1 factorization of a matrix. 2. The system of claim 1, wherein the image sensor includes a digital camera. 3. The system of claim 1, wherein the image sensor includes an analog camera, and further comprising a frame grabber connected between the image sensor and the two-dimensional motion filter module for digitizing images captured by the image sensor. 4. The system of claim 1, wherein:the two-dimensional motion filter module is for determining a plurality of two-dimensional motion parameters for each of a plurality of surface patches of the object; andthe three-dimensional structure recovery module is for estimating the set of three-dimensional shape parameters from the set of two-dimensional motion parameters for each of the plurality of surface patches of the object. 5. The system of claim 1, wherein:the two-dimensional motion filter module is further for assigning confidence weights to certain of the two-dimensional motion parameters; andthe three-dimensional structure recovery module is for estimating the set of three-dimensional shape parameters and the set of three-dimensional motion parameters from the set of two-dimensional motion parameters using a weighted rank 1 factorization of a matrix based on the confidence weights. 6. The system of claim 1, further comprising a texture recovery module in communication with the three-dimensional structure recovery module for recovering a set of texture parameters for the object based on the set of three-dimensional shape parameters and three-dimensional motion parameters. 7. The system of claim 1, further comprising a three-dimensional shape refinement module in communication with the three-dimensional structure recovery module for generating a refined set of three-dimensional shape parameters. 8. The system of claim 7, wherein the three-dimensional shape refinement module is for generating the refined set of three-dimensional shape parameters using an iterative multiresolution continuation method based on a maximum likelihood estimation technique. 9. The system of claim 7, further comprising a texture recovery module in communication with the three-dimensional structure recovery module and the three-dimensional shape refinement module for recovering a set of texture parameters for the object based on the refined set of three-dimensional shape parameters determined by the three-dimensional shape refinement module and the three-dimensional motion parameters determined by the three-dimensional structure recovery module. 10. A method for generating a three-dimensional model of an object from a two-dimensional image sequence, comprising:capturing a sequence of images of a scene, the scene including the object;determining a plurality of two-dimensional motion parameters for the object from the sequence of images; andestimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters from the two-dimensional motion parameters using a rank 1 factorization of a matrix. 11. The method of claim 10, wherein capturing includes capturing a sequence of digitized images of the scene. 12. The method of claim 10, wherein capturing includes capturing a sequence of analog images of the scene, and further comprising digitizing the analog images prior to determining the pluralit y of two-dimensional motion parameters. 13. The method of claim 10, wherein:determining includes determining a plurality of two-dimensional motion parameters from the digitized image sequence for each of a plurality of surface patches of the object; andestimating a set of three-dimensional shape parameters includes estimating a set of three-dimensional shape parameters for each of the plurality of surface patches from the two-dimensional motion parameters using a rank 1 factorization of a matrix. 14. The method of claim 10, further comprising assigning confidence weights to certain of the two-dimensional motion parameters, and wherein estimating the set of three-dimensional shape parameters and the set of three-dimensional motion parameters includes estimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters from the two-dimensional motion parameters using a weighted rank 1 factorization of a matrix based on the confidence weights. 15. The method of claim 10, further comprising generating a set of texture parameters for the object based on the set of three-dimensional shape parameters and the set of three-dimensional motion parameters. 16. The method of claim 10, further comprising generating a refined set of three-dimensional shape parameters from the set of three-dimensional shape parameters. 17. The method of claim 10, wherein generating a refined set of three-dimensional shape parameters includes generating the refined set of three-dimensional shape parameters using an iterative multiresolution continuation method based on a maximum likelihood estimation technique. 18. The method of claim 16, further comprising determining a set of texture parameters for the object based on the refined set of three-dimensional shape parameters. 19. A method for generating a three-dimensional model of an object from a sequence of digitized images of a scene, the scene including the object, comprising:determining a plurality two-dimensional motion parameters for the object from the sequence of digitized images of the scene; andestimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters from the two-dimensional motion parameters using a rank 1 factorization of a matrix. 20. The method of claim 19, wherein:determining includes determining a plurality of two-dimensional motion parameters from the digitized image sequence for each of a plurality of surface patches of the object; andestimating a set of three-dimensional shape parameters includes estimating a set of three-dimensional shape parameters for each of the plurality of surface patches from the two-dimensional motion parameters using a rank 1 factorization of a matrix. 21. The method of claim 19, further comprising assigning confidence weights to certain of the two-dimensional motion parameters, and where estimating the set of three-dimensional shape parameters and the set of three-dimensional motion parameters includes estimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters from the two-dimensional motion parameters using a weighted rank 1 factorization of a matrix based on the confidence weights. 22. The method of claim 19, further comprising generating a set of texture parameters for the object based on the set of three-dimensional shape parameters and the set of three-dimensional motion parameters. 23. The method of claim 19, further comprising generating a refined set of three-dimensional shape parameters from the set of three-dimensional shape parameters. 24. A device for generating a three-dimensional model of an object from a sequence of digitized images of a scene, the scene including the object, comprising:a two-dimensional motion filter module for determining a plurality two-dimensional motion parameters for the object from the sequence of digitized images; anda three-dimensional structure recovery module in communication with the two-dimension al motion filter module for estimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters from the two-dimensional motion parameters using a rank 1 factorization of a matrix. 25. The device of claim 24, wherein:the two-dimensional motion filter module is for determining a plurality of two-dimensional motion parameters for each of a plurality of surface patches of the object; andthe three-dimensional structure recovery module is for estimating the set of three-dimensional shape parameters from the set of two-dimensional motion parameters for each of the plurality of surface patches of the object using a rank 1 factorization of a matrix. 26. The device of claim 24, wherein:the two-dimensional motion filter module is further for assigning confidence weights to certain of the two-dimensional motion parameters; andthe three-dimensional structure recovery module is for estimating the set of three-dimensional shape parameters and the set of three-dimensional motion parameters from the two-dimensional motion parameters using a weighted rank 1 factorization of a matrix based on the confidence weights. 27. The device of claim 24, further comprising a texture recovery module in communication with the three-dimensional structure recovery module for recovering a set of texture parameters for the object based on the set of three-dimensional shape parameters and three-dimensional motion parameters. 28. The device of claim 24, further comprising a three-dimensional shape refinement module in communication with the three-dimensional structure recovery module for generating a refined set of three-dimensional shape parameters. 29. A device for generating a three-dimensional model of an object from a sequence of digitized images of a scene, the scene including the object, comprising:means for determining a plurality two-dimensional motion parameters for the object from the sequence of digitized images of the scene; andmeans for estimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters from the two-dimensional motion parameters using a rank 1 factorization of a matrix. 30. The device of claim 29, wherein:the means for determining includes means for determining a plurality of two-dimensional motion parameters for each of a plurality of surface patches of the object; andthe means for estimating the set of three-dimensional shape parameters and the set of three-dimensional motion parameters includes means for estimating the set of three-dimensional shape parameters from the two-dimensional motion parameters for each of the plurality of surface patches of the object using a rank 1 factorization of a matrix. 31. The device of claim 29, further comprising means for assigning confidence weights to certain of the two-dimensional motion parameters, and wherein the means for estimating the set of three-dimensional shape parameters and the set of three-dimensional motion parameters includes means for estimating the set of three-dimensional shape parameters and the set of three-dimensional motion parameters from the two-dimensional motion parameters using a weighted rank 1 factorization of a matrix based on the confidence weights. 32. The device of claim 29, further comprising means for recovering a set of texture parameters for the object based on the set of three-dimensional shape parameters and three-dimensional motion parameters. 33. The device of claim 29, further comprising means for generating a refined set of three-dimensional shape parameters. 34. A computer-readable medium, having stored thereon instructions, which when executed by a processor, cause the processor to:determine a plurality of two-dimensional motion parameters for an object from a sequence of digitized images of a scene, wherein the scene includes the object; andestimate a set of three-dimensional shape parameters and a set of three-dimensional motion parameters from the two-dimensional motion para meters using a rank 1 factorization of a matrix. 35. The computer-readable medium of claim 34, having further stored thereon instructions, which when executed by the processor, cause the processor to:determine a plurality of two-dimensional motion parameters from the digitized image sequence for each of a plurality of surface patches of the object; andestimate a set of three-dimensional shape parameters for each of the plurality of surface patches from the two-dimensional motion parameters using a rank 1 factorization of a matrix. 36. The computer-readable medium of claim 34, having further stored thereon instructions, which when executed by the processor, cause the processor to:assign confidence weights to certain of the two-dimensional motion parameters; andestimate a set of three-dimensional shape parameters and a set of three-dimensional motion parameters from the two-dimensional motion parameters using a weighted rank 1 factorization of a matrix based on the confidence weights. 37. The computer-readable medium of claim 34, having further stored thereon instructions, which when executed by the processor, cause the processor to generate a set of texture parameters for the object based on the set of three-dimensional shape parameters and the set of three-dimensional motion parameters. 38. The computer-readable medium of claim 34, having further stored thereon instructions, which when executed by the processor, cause the processor to generate a refined set of three-dimensional shape parameters from the set of three-dimensional shape parameters. 39. A computer-readable medium, having stored thereon instructions, which when executed by a processor, cause the processor to:generate a matrix of two-dimensional motion parameters, wherein the two-dimensional motion parameters describe motion of an object in an image sequence; andestimate a set of three-dimensional shape parameters and a set of three-dimensional motion parameters for the object from the two-dimensional motion parameters using a rank 1 factorization of the matrix. 40. The computer-readable medium of claim 39, having further stored thereon instructions, which when executed by the processor, cause the processor to:assign confidence weights to certain of the two-dimensional motion parameters; andestimate the set of three-dimensional shape parameters and the set of three-dimensional motion parameters from the matrix using a weighted rank 1 factorization of the matrix based on the confidence weights. 41. The computer-readable medium of claim 39, having further stored thereon instructions, which when executed by the processor, cause the processor to generate a set of texture parameters for the object based on the set of three-dimensional shape parameters and the set of three-dimensional motion parameters. 42. The computer-readable medium of claim 39, having further stored thereon instructions, which when executed by the processor, cause the processor to generate a refined set of three-dimensional shape parameters from the set of three-dimensional shape parameters. 43. A three-dimensional structure recovery module, comprising:means for generating a matrix of two-dimensional motion parameters, wherein the two-dimensional motion parameters describe motion of an object in an image sequence; andmeans for estimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters for the object from the two-dimensional motion parameters using a rank 1 factorization of the matrix. 44. The three-dimensional recovery module of claim 43, wherein certain of the two-dimensional motion parameters have confidence weights assigned thereto, and wherein the means for estimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters includes means for estimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters from the two-dimensional motion parameters using a weighted rank 1 factori zation of the matrix based on the confidence weights. 45. A method for generating a three-dimensional model of an object from a two-dimensional image sequence of a scene including the object, wherein the object includes at least one occlusion, comprising:generating a matrix of two-dimensional motion parameters for the object from the sequence of images;generating an initial estimate of a set three-dimensional shape parameters and a set of three-dimensional motion parameters for the object based a submatrix of the matrix of two-dimensional motion parameters, wherein the submatrix includes no missing elements; anditeratively re-estimating the set three-dimensional shape parameters and the set of three-dimensional motion parameters based the initial estimate to generate a final estimate of the set three-dimensional shape parameters and a set of three-dimensional motion parameters for the object. 46. The method of claim 45, wherein generating an initial estimate includes generating an initial estimate of the set three-dimensional shape parameters and the set of three-dimensional motion parameters of the object using a rank 1 factorization of the submatrix. 47. The method of claim 45, wherein iteratively re-estimating includes:computing unknown three-dimensional shape parameters for the object based on known three-dimensional motion parameters; andcomputing unknown three-dimensional motion parameters for the object based on known three-dimensional shape parameters. 48. A device for generating a three-dimensional model of an object from a sequence of digitized images of a scene including the object, wherein the object includes at least one occlusion, comprising:a two-dimensional motion filter module for determining a plurality two-dimensional motion parameters for the object from the sequence of digitized images; anda three-dimensional structure recovery module in communication with the two-dimensional motion filter module for:generating a matrix of two-dimensional motion parameters for the object from the sequence of images;generating an initial estimate of a set three-dimensional shape parameters and a set of three-dimensional motion parameters for the object based on a submatrix of the matrix of two-dimensional motion parameters, wherein the submatrix includes no missing elements; anditeratively re-estimating the set three-dimensional shape parameters and the set of three-dimensional motion parameters based the initial estimate to generate a final estimate of the set three-dimensional shape parameters and a set of three-dimensional motion parameters for the object. 49. The device of claim 48, wherein the three-dimensional structure recovery module is for generating an initial estimate of a set three-dimensional shape parameters and a set of three-dimensional motion parameters using a rank 1 factorization of the submatrix. 50. A communications network, comprising:a two-dimensional motion filter module for determining a plurality two-dimensional motion parameters for an object from a sequence of digitized images of a scene including the object;a three-dimensional structure recovery module in communication with the two-dimensional motion filter module for estimating a set of three-dimensional shape parameters and a set of three-dimensional motion parameters for the object from the two-dimensional motion parameters using a rank 1 factorization of a matrix; andan end-user terminal in communication with the three-dimensional structure recovery module. 51. The communications network of claim 50, wherein the end-user terminal is in communication with the three-dimensional structure recovery module via an IP-based network. 52. The communications network of claim 50, wherein the end-user terminal is in communication with the three-dimensional structure recovery module via at least one of a wireless communications link and a wireline communications link.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.