System and method for spatio temporal video image enhancement

1. A method for motion improvement of low light video images, the method comprising:

receiving a current video frame;

identifying moving rigid objects in the received current video frame from the current video frame and a previous or future video frame;

selecting one or more cardinal pixels in each identified moving rigid object in the received current video frame;

identifying a patch around each cardinal pixel, in the received current video frame;

calculating a matched patch in the previous or future video frame for each identified patch matching said each identified patch;

defining a location of the cardinal pixel of each matched patch in the previous or future video frame;

identifying matched rigid objects in the previous or future video frame corresponding to each defined location of each cardinal pixel; and

extracting each identified matched rigid moving object from the previous or future frame and processing each of the extracted matched rigid moving objects with a corresponding rigid object in the received current video frame to generate an improved quality video image frame.

2. The method of claim 1 , wherein the cardinal pixel is a centroid pixel.

3. The method of claim 1 , wherein identifying moving rigid objects further comprises finding the difference between the current video frame and the previous or future video frame.

4. The method of claim 1 , wherein selecting one or more cardinal pixels further comprises dividing rigid objects into two or more regions.

5. The method of claim 1 , wherein identifying a patch around each cardinal pixel further comprises determining an amount of spatial variation in the rigid object.

6. The method of claim 1 , wherein calculating a matched patch further comprises finding the difference between a patch in the current frame with similarly sized patches in the previous or future frames.

7. The method of claim 1 , wherein defining a location of the cardinal pixel of each matched patch in the previous or future video frame further comprises finding a pixel in a center of the matched patch.

8. The method of claim 1 , wherein identifying matched rigid objects in the previous or future video frame further comprises using geometric size or shape information about the rigid object in the current frame.

9. The method of claim 1 , wherein processing of the rigid object from the current frame further comprises weighted averaging of the rigid object from the current frame and the rigid object from the previous or future frame.

10. A method for improving quality of low light video images, the method comprising:

receiving a current video frame;

temporally enhancing the received video frame by applying a first weight matrix including higher weight factors for stationary regions and lower weight factors for moving regions in the received current video frame to the received current video frame and a reference frame, to generate an enhanced temporal video frame;

spatially enhancing the enhanced temporal video frame by applying a second weight matrix including higher weight factors for stationary regions and lower weight factors for moving regions, in the enhanced temporal video frame, to generate an enhanced spatial video frame; and

motion enhancing the enhanced temporal video frame by extracting matched rigid moving object in a previous or future frame and processing each of the extracted matched rigid moving objects with a corresponding rigid object in the enhanced temporal video frame to generate an improved quality video image frame.

11. The method of claim 10 , wherein generating a reference video frame further comprises weighted averaging of the current video frame with the previous or future video frame.

12. The method of claim 10 , further comprising computing a likelihood metric to determine the likelihood of a pixel to correspond to stationary regions and moving regions in the received current video frame from the reference frame.

13. The method of claim 12 , wherein computing a likelihood metric further comprises finding the difference between the current video frame and the reference frame.

14. The method of claim 10 , wherein computing a first weight matrix further comprises finding the inverse of the likelihood metric.

15. The method of claim 10 , wherein said generating an enhanced temporal video frame further comprises weighted averaging the current video frame and the reference frame by using the first weight matrix.