Method and apparatus for motion compensation prediction

1. A video decoding method for motion compensation performed under an inter-frame prediction to decode a target picture, the method comprising computer executable steps executed by a processor of a video decoder to implement:

decoding a residual and a motion vector received from an encoder;

referencing to the motion vector to retrieve a reference sample from a reference picture stored in a reference picture memory, wherein the reference picture stored in the reference picture memory and the reference sample retrieved from the reference picture are both represented with a first bit depth;

performing a scaling-up operation and a first fractional sample interpolation in a first direction on the retrieved reference sample to generate a first set of fractional samples represented with a second bit depth to which the first bit depth is scaled up by a scaling-up factor, wherein the second bit depth is constant and set equal to a number of bits available to represent the fractional sample, and the scaling-up factor is set equal to the second bit depth minus the first bit depth and is variable to keep the second bit depth constant and independent from a change of the first bit depth;

performing a second fractional sample interpolation on the first set of fractional samples in a second direction to generate a second set of fractional samples represented with the second bit depth;

referencing fractional parts of the motion vector to derive a prediction sample from the first and second sets of fractional samples, wherein the prediction sample is represented with the second bit depth;

scaling down and clipping the prediction sample to generate a prediction picture scaled down from the second bit depth to the first bit depth; and

adding the prediction picture and the residual to reconstruct the target picture represented with the first bit depth.

2. The method according to claim 1 , wherein the fractional sample interpolation applies an 8-tap FIR (Finite Impulse Response) filter having one of the following three sets of coefficients to generate a quarter-pel sample:

[−1, 4, −10, 58, 17, −5, 1, 0];

[−1, 4, −11, 40, 40, −11, 4, −1]; and

[0, 1, −5, 17, 58, −10, 4, −1].

3. The method according to claim 2 , wherein the fractional sample interpolation comprises truncating each filtered result so that the fractional samples are represented with the second bit depth.

4. The method according to claim 3 , wherein the second bit depth is maintained throughout the first and second fractional interpolation processes.

5. The method according to claim 3 , wherein truncating each filtered result comprises rounding truncated results towards minus infinity.

6. The method according to claim 1 , further comprising subtracting an offset from the scaled up samples.

7. A video encoding method for motion compensation performed under an inter-frame prediction, the method comprising computer executable steps executed by a processor of a video encoder to implement:

comparing a target picture and a reference picture stored in a reference picture memory to estimate a motion vector, wherein the reference picture stored in the picture memory is represented with a first bit depth;

referencing the motion vector to retrieve a reference sample from the reference picture, wherein the reference sample retrieved from the reference picture is represented with the first bit depth;

performing a scaling-up operation and a first fractional sample interpolation in a first direction on the retrieved reference sample to generate a first set of fractional samples represented with a second bit depth to which the first bit depth is scaled up by a scaling-up factor, wherein the second bit depth is constant and set equal to a number of bits available to represent the fractional sample, and the scaling-up factor is set equal to the second bit depth minus the first bit depth and is variable to keep the second bit depth constant and independent from a change of the first bit depth;

performing a second fractional sample interpolation on the first set of fractional samples in a second direction to generate a second set of fractional samples represented with the second bit depth;

referencing fractional parts of the motion vectors to derive a prediction sample from the first and second sets of fractional samples, wherein the prediction sample is represented with the second bit depth;

scaling down and clipping the prediction sample to generate a prediction picture scaled down from the second bit depth to the first bit depth; and

subtracting the prediction picture from the target picture to derive a residual for encoding.

8. The method according to claim 7 , wherein the fractional sample interpolation applies an 8-tap FIR (Finite Impulse Response) filter having one of the following three sets of coefficients to generate a quarter-pel sample:

[−1, 4, −10, 58, 17, −5, 1, 0];

[−1, 4, −11, 40, 40, −11, 4, −1]; and

[0, 1, −5, 17, 58, −10, 4, −1].

9. The method according to claim 8 , wherein the fractional sample interpolation comprises truncating each filtered result so that the fractional samples are represented with the second bit depth.

10. The method according to claim 9 , wherein the second bit depth is maintained throughout the first and second interpolation processes.

11. The method according to claim 9 , wherein truncating each filtered result comprises rounding truncated results towards minus infinity.

12. The method according to claim 7 , further comprising subtracting an offset from the scaled up samples.