Image processing device and method using adjusted motion vector accuracy between sub-pixels of reference frames
The present technology relates to an image processing device and a method capable of improving encoding efficiency. In a weighted prediction process performed in an image encoding process or in an image decoding process, a motion compensation unit that performs motion compensation of sub-pixel accuracy for each of L0 and L1, a weighted addition unit that applies weighted addition to arithmetic operation results by the motion compensation unit, and a rounding processing unit that suppresses a decrease in arithmetic operation accuracy by performing a necessary rounding process once with respect to an arithmetic operation result by the weighted addition unit are provided. The present disclosure can be applied to an image processing device.
1. An image processing device comprising:
at least one central processing unit (CPU) configured to
perform motion compensation of integer pixel or sub-pixel accuracy for each of reference frames L0 and L1;
apply weighted addition to arithmetic operation results of the motion compensation, wherein the arithmetic operation results are respectively adjusted by multiplying each arithmetic operation result having lower accuracy by an adjustment amount that is determined based on each respective combination of horizontal accuracy and vertical accuracy between the reference frames L0 and L1, including each of the integer pixel or sub-pixel accuracy for horizontal components between each of the reference frames L0 and L1 and the integer pixel or sub-pixel accuracy for vertical components between each of the reference frames L0 and L1; and
apply a rounding process to an arithmetic operation result of the weighted addition,
wherein the adjustment amount is further determined based on a similarity of the accuracy of the horizontal and vertical components between the reference frames L0 and L1 in each respective combination, and
wherein the adjustment amount is determined according to a table indicating a respective adjustment value for each respective combination of the horizontal and vertical components between the reference frames L0 and L1.
2. The image processing device according to claim 1 , wherein the at least one CPU is further configured to adjust higher motion vector accuracy to lower motion vector accuracy between the reference frames L0 and L1.
3. The image processing device according to claim 1 , wherein the at least one CPU does not perform the adjustment of the accuracy when the integer pixel or sub-pixel accuracy of L0 and the integer pixel or sub-pixel accuracy of L1 are equal to each other.
4. The image processing device according to claim 1 , wherein the at least one CPU, in applying a rounding process, is further configured to perform a division process after the arithmetic operation process of the weighted addition including a multiplication process for motion vector accuracy adjustment and a division process for a weighted prediction process.
5. The image processing device according to claim 1 , wherein the weighted addition is performed by Explicit Weighted Prediction.
6. The image processing device according to claim 1 , wherein the weighted addition is performed by Implicit Weighted Prediction.
7. The image processing device according to claim 1 , wherein the weighted addition is performed by Intensity Compensation.
8. A non-transitory computer-readable storage medium having embodied thereon a program, which when executed by a computer causes the computer to execute an image processing method, the method comprising:
performing motion compensation of integer pixel or sub-pixel accuracy for each of reference frames L0 and L1;
applying weighted addition to arithmetic operation results of the motion compensation, wherein the arithmetic operation results are respectively adjusted by multiplying each arithmetic operation result having lower accuracy by an adjustment amount that is determined based on each respective combination of horizontal accuracy and vertical accuracy between the reference frames L0 and L1, including each of the integer pixel or sub-pixel accuracy for horizontal components between each of the reference frames L0 and L1 and the integer pixel or sub-pixel accuracy for vertical components between each of the reference frames L0 and L1; and
applying a rounding process to an arithmetic operation result of the weighted addition,
wherein the adjustment amount is further determined based on a similarity of the accuracy of the horizontal and vertical components between the reference frames L0 and L1 in each respective combination, and
wherein the adjustment amount is determined according to a table indicating a respective adjustment value for each respective combination of the horizontal and vertical components between the reference frames L0 and L1.
9. An image processing method, implemented via at least one processor, the method comprising:
performing motion compensation of integer pixel or sub-pixel accuracy for each of reference frames L0 and L1;
applying weighted addition to arithmetic operation results of the motion compensation, wherein the arithmetic operation results are respectively adjusted by multiplying each arithmetic operation result having lower accuracy by an adjustment amount that is determined based on each respective combination of horizontal accuracy and vertical accuracy between the reference frames L0 and L1, including each of the integer pixel or sub-pixel accuracy for horizontal components between each of the reference frames L0 and L1 and the integer pixel or sub-pixel accuracy for vertical components between each of the reference frames L0 and L1; and
applying a rounding process to an arithmetic operation result of the weighted addition,
wherein the adjustment amount is further determined based on a similarity of the accuracy of the horizontal and vertical components between the reference frames L0 and L1 in each respective combination, and
wherein the adjustment amount is determined according to a table indicating a respective adjustment value for each respective combination of the horizontal and vertical components between the reference frames L0 and L1.
10. The image processing device according to claim 1 , wherein the arithmetic operation results are not adjusted when the respective combination has the same accuracy of the horizontal and vertical components for both of the reference frames L0 and L1.
11. The non-transitory computer-readable storage medium according to claim 8 , wherein the arithmetic operation results are not adjusted when the respective combination has the same accuracy of the horizontal and vertical components for both of the reference frames L0 and L1.
12. The image processing method according to claim 9 , wherein the arithmetic operation results are not adjusted when the respective combination has the same accuracy of the horizontal and vertical components for both of the reference frames L0 and L1.