Unit for and method of motion estimation and image processing apparatus provided with such motion estimation unit
View Patent ↗The motion estimation unit ( 100 ) comprises a first summation means ( 106 ) for calculating match errors of a number of candidate motion vectors of a segment ( 116 ) of a first image ( 118 ). The motion vector with the lowest match error can be assigned to the segment ( 116 ) as the estimated motion vector. The match error is based on summation of absolute differences between pixel values of the segment ( 116 ) and pixel values of a second image ( 120 ). In order to estimate the deviation of the estimated motion from the true motion vector, i.e. the motion vector error ( 130 ), the motion estimation unit ( 100 ) further comprises a second summation means ( 108 ) for calculating a variance parameter by summation of absolute differences between pixel values of the segment ( 116 ) and pixel values of the first image ( 118 ) and estimation means ( 110 ) for estimating a motion vector error ( 130 ) by comparing the match error with the variance parameter.
1. A motion estimation unit that generates motion vectors, each of the motion vectors corresponding to a respective segment of a first image, the motion estimation unit comprising:
a first summer that calculates a match error of the motion vector of the respective segment by summation of absolute differences between pixel values of the respective segment and pixel values of a second image;
a second summer that calculates a variance parameter by summation of absolute differences between pixel values of the respective segment and pixel values of the first image; and
an estimator that estimates a motion vector error by comparing the match error with the variance parameter, said motion vector error being used to assess a quality of the motion vector, wherein an expectation value of the motion vector error is proportional to the match error divided by the variance parameter, wherein the expectation value of the motion vector error is approximately ⅗ of the match error divided by the variance parameter.
2. The motion estimation unit of claim 1 , wherein the second summer calculates the variance parameter by adding:
absolute differences between pixel values of the respective segment and pixel values of a second segment which corresponds to the respective segment being shifted at least one pixel in a first direction; and
absolute differences between pixel values of the respective segment and pixel values of a third segment which corresponds to the respective segment being shifted at least one pixel in a second direction, the first direction being cross to the second direction.
3. The motion estimation unit of claim 2 , wherein the second summer adds absolute differences between pixel values of the respective segment and pixel values of a fourth segment which corresponds to the respective segment being shifted at least one pixel in the first direction and at least one pixel in the second direction.
4. The motion estimation unit of claim 1 , characterized in generating the motion vector of the respective segment, the respective segment being a block of pixels.
5. The motion estimation unit of claim 1 , characterized in generating the motion vector of the respective segment based on luminance values as pixel values.
6. A method of generating motion vectors, each of the motion vectors corresponding to a respective segment of a first image, the method comprising:
calculating a match error of a motion vector of the respective segment by summation of absolute differences between pixel values of the respective segment and pixel values of a second image;
calculating a variance parameter by summation of absolute differences between pixel values of the respective segment and pixel values of the first image;
estimating a motion vector error by comparing the match error with the variance parameter; and
determining a quality of the motion vector based on the motion vector error, wherein an expectation value of the motion vector error is proportional to the match error divided by the variance parameter, wherein the expectation value of the motion vector error is approximately ⅗ of the match error divided by the variance parameter.
7. The method of claim 6 , wherein the variance parameter is calculated by adding:
absolute differences between pixel values of the respective segment and pixel values of a second segment which corresponds to the respective segment being shifted at least one pixel in a first direction; and
absolute differences between pixel values of the respective segment and pixel values of a third segment which corresponds to the respective segment being shifted at least one pixel in a second direction, with the first direction cross to the second direction.
8. The method of claim 7 , further comprising:
adding absolute differences between pixel values of the respective segment and pixel values of a fourth segment which corresponds to the respective segment being shifted at least one pixel in the first direction and at least one pixel in the second direction.
9. An image processing apparatus comprising:
a motion estimation unit that generates motion vectors, each of the motion vectors corresponding to a respective segment of a first image, the motion estimation unit comprising:
a first summer that calculates a match error of a motion vector of the respective segment by summation of absolute differences between pixel values of the respective segment and pixel values of a second image; and
a second summer that calculates a variance parameter by summation of absolute differences between pixel values of the respective segment and pixel values of the first image; and
an estimator that estimates a motion vector error by comparing the match error with the variance parameter, said motion vector error being used to assess a quality of the motion vector, wherein an expectation value of the motion vector error is proportional to the match error divided by the variance parameter, wherein the expectation value of the motion vector error is approximately ⅗ of the match error divided by the variance parameter.
10. The image processing apparatus of claim 9 , wherein the motion compensated image processing unit reduces noise in the first image.
11. The image processing apparatus of claim 9 , wherein the motion compensated image processing unit de-interlaces the first image.
12. The image processing apparatus of claim 9 , wherein the motion compensated image processing unit performs an up-conversion.
13. The motion estimation unit of claim 1 , wherein the motion vector error is used to detect object edges.
14. The method of claim 8 , further comprising:
detecting object edges based on the motion vector error.
15. The motion estimation unit of claim 1 , wherein a standard deviation of the motion vector error is approximately 7/10.
16. The method of claim 6 , wherein a standard deviation of the motion vector error is approximately 7/10.
17. The image processing apparatus of claim 9 , wherein a standard deviation of the motion vector error is approximately 7/10.