Method for analyzing movements in temporal sequences of digital images

• US 5,365,603 A
• Filed: 01/25/1993
• Issued: 11/15/1994
• Est. Priority Date: 07/27/1990
• Status: Expired due to Fees

• Current Assignee: Siemens AG
• Original Assignee: Siemens AG
• Inventors: Karmann, Klaus-Peter
• Primary Examiner(s): Couso, Jose L.
• Application Number:
  US07/969,212
• Time in Patent Office: 659 Days
• Field of Search: 382/17, 382/48, 382/54, 358/133, 358/135, 358/136, 358/138
• US Class Current: 382/291
• CPC Class Codes: G06T 7/254 involving subtraction of images

Abstract

Temporal sequences of predicted measurement data (Y), among which there are predicted movement data (W) relating to temporal sequences of digital images (I), are compared with temporal sequences of measured measurement data (S), among which there are measured movement data (V) relating to the same temporal sequences of digital images (I), by forming differences (D) from measured measurement data (S) and predicted measurement data (Y). These differences (D) are fed to a time-recursive filter (TRF) for estimating temporal sequences of state variables (X). The predicted measurement data (Y), including the predicted movement data (W) are determined from these state variables (X) with the aid of a measurement map (H). These predicted movement data (W) are used to determine (ME) measured movement data (V) relating to the temporal sequences of digital images (I). The method facilitates the determination of spatial structures and the analysis of spatial object movements, including intrinsic movement. Moving objects can be detected and segmented reliably.

Claims

1. A method for analyzing temporal sequences of digital images, comprising the steps of:

providing temporal sequences of digital images from at least one sensor;

deriving measured movement data from the temporal sequences of digital images;

deriving predicted movement data from the temporal sequences of digital images;

providing temporal sequences of predicted measurement data and temporal sequences of measured measurement data;

determining temporal sequences of state variables in a temporally recursive fashion using a comparison of the temporal sequences of predicted measurement data, that includes the predicted movement data, with the temporal sequences of measured measurement data, that includes the measured movement data;

said temporal sequences of state variables being input to a measurement map, and said predicted movement data and said temporal sequences of predicted measurement data being output from said measurement map;

said predicted movement data and said temporal sequences of digital images being input to a formation device, measured movement data and said temporal sequences of measured movement data being output from said formation device;

said temporal sequences of measured movement data and said temporal sequences of predicted measurement data being into to a comparator, and temporal sequences of resultant measurement data being output from said comparator; and

said temporal sequences of resultant measurement data and said temporal sequences of state variables being input to a time-recursive filter, and said temporal sequences being output from said time-recursive filter.

2. The method as claimed in claim 1, wherein the predicted and measured movement data are displacement vectors.

3. The method as claimed in claim 1, wherein the predicted measurement data is determined from the state variables.

4. The method as claimed in claim 1, wherein the comparison is performed by forming differences between the measured measurement data and the predicted measurement data.

5. The method according to claim 1, wherein the state variables are determined using at least one time-recursive filter.

6. The method as claimed in claim 5, wherein a measure of correspondence between the predicted measurement data and the measured measurement data is determined from the comparison.

7. The method as claimed in claim 6, wherein characteristics of the filter depend on the measure of the correspondence between the predicted measurement data and the measured measurement data.

8. The method as claimed in claim 1, wherein the measured movement data is determined taking account of the predicted movement data.

9. The method as claimed in claim 5, wherein in determining the measured measurement data, the significance of the measured measurement data is estimated using a measure of quality.

10. The method as claimed in claim 9, wherein characteristics of the filter depend on the significance of the measured measurement data, expressed by the measure of quality.

11. The method as claimed in claim 1, wherein the state variables are determined using at least one time-recursive filter, wherein a measure of correspondence between the predicted measurement data and the measured measurement data is determined from the comparison, wherein characteristics of the filter depend on the measure of the correspondence between predicted measurement data and the measured measurement data, wherein in determining the measured measurement data, the significance of the measured measurement data is estimated using a measure of quality, wherein characteristics of the filter depend on the significance of the measured measurement data, expressed by the measure of quality, and wherein the at least one time-recursive filter is a Kalman filter having system noise covariances given by the measure of the correspondence between the predicted measurement data and the measured measurement data, and having measurement noise covariances given by the measure of quality for estimating the significance of the measured measurement data.

12. The method as claimed in claim 1, wherein the state variables describe intrinsic movement of at least one imaging sensor.

13. The method as claimed in claim 1, wherein the state variables describe spatial structures relative to images of at least one imaging sensor.

14. The method as claimed in claim 1, wherein the state variables describe object movements in the temporal sequences of digital images.

15. The method as claimed in claim 1, wherein there are image signals of at least one temporal sequence of digital images among the measured measurement data, and wherein there are predicted values for said image signals among the predicted measurement data.

16. The method as claimed in claim 1, wherein at least one temporal sequence of movement-compensated images is determined using the predicted movement data.

17. The method as claimed in claim 16, wherein at least one temporal sequence of reference images is determined from the movement-compensated images.

18. The method as claimed in claim 17, wherein moving objects are detected and segmented by comparing the movement-compensated images with the reference images.

19. The method as claimed in claim 1, wherein a measure of correspondence between the predicted measurement data and the measured measurement data is determined from the comparison.

20. The method as claimed in claim 1, wherein in determining the measured measurement data, the significance of the measured measurement data is estimated using a measure of quality.