Prediction image generating method, method for decoding a moving image, and method for coding a moving image
A prediction image generation apparatus for generating a prediction image by performing motion compensation on multiple reference images, the prediction image generation apparatus includes a prediction image generation unit configured to generate the prediction image with reference to a first reference image, a second reference image and a gradient correction term. A motion compensation unit ( 3091 ) having a motion compensation function using a bi-prediction gradient change (Bi-directional optical flow, BIO) prediction derives the gradient correction term with reference to each pixel value of a horizontal gradient image and a vertical gradient image. An inter-frame-distance weight derivation unit ( 309131 ) included a gradient correction bi-prediction derivation unit ( 30913 ) included in the motion compensation unit determines each weight coefficient to be multiplied by the value of the each pixel, depending on a first picture interval and a second picture interval, the first picture interval being a picture interval between a target image to be predicted and the first reference image, the second picture interval being a picture interval between the target image and the second reference image.
1. A prediction image generating method for generating a prediction image, the prediction image generating method comprising:
generating the prediction image with reference to a first reference image, a second reference image and a gradient correction term, wherein
the gradient correction term is derived with reference to a value of each pixel of
a first gradient image indicating a gradient along a first direction of the first reference image,
a second gradient image indicating a gradient along a second direction of the first reference image,
a third gradient image indicating a gradient along the first direction of the second reference image, and
a fourth gradient image indicating a gradient along the second direction of the second reference image, and
each weight coefficient to be multiplied by the value of the each pixel of the first to fourth gradient images is determined for each block, wherein
each of weight coefficients is either a first value or a second value, a method for determining the first value is different from a method for determining the second value, and
wherein
the first value is determined for each block by using (a) deference of the first reference image and the second reference image, (b) sum of the first gradient image and the third gradient image and (c) sum of the second gradient image and the fourth gradient image, and
the second value is determined for each block by using (a) deference of the first reference image and the second reference image, (b) sum of the first gradient image and the third gradient image and (c) sum of the second gradient image and the fourth gradient image and (d) the first value.
2. A method for decoding a moving image, the method comprising:
generating a prediction image with reference to a first reference image, a second reference image and a gradient correction term; and
generating a local decoded image based on the prediction image and a residual image;
the gradient correction term is derived with reference to a value of each pixel of
a first gradient image indicating a gradient along a first direction of the first reference image,
a second gradient image indicating a gradient along a second direction of the first reference image,
a third gradient image indicating a gradient along the first direction of the second reference image, and
a fourth gradient image indicating a gradient along the second direction of the second reference image, and
each weight coefficient to be multiplied by the value of the each pixel of the first to fourth gradient images is determined for each block, wherein
each of weight coefficients is either a first value or a second value, a method for determining the first value is different from a method for determining the second value, and
wherein
the first value is determined for each block by using (a) deference of the first reference image and the second reference image, (b) sum of the first gradient image and the third gradient image and (c) sum of the second gradient image and the fourth gradient image, and
the second value is determined for each block by using (a) deference of the first reference image and the second reference image, (b) sum of the first gradient image and the third gradient image and (c) sum of the second gradient image and the fourth gradient image and (d) the first value.
3. A method for coding a moving image, the method comprising:
generating a prediction image with reference to a first reference image, a second reference image and a gradient correction term; and
coding a residual picture based on the prediction image;
the gradient correction term is derived with reference to a value of each pixel of
a first gradient image indicating a gradient along a first direction of the first reference image,
a second gradient image indicating a gradient along a second direction of the first reference image,
a third gradient image indicating a gradient along the first direction of the second reference image, and
a fourth gradient image indicating a gradient along the second direction of the second reference image, and
each weight coefficient to be multiplied by the value of the each pixel of the first to fourth gradient images is determined for each block, wherein
each of weight coefficients is either a first value or a second value, a method for determining the first value is different from a method for determining the second value, and
wherein
the first value is determined for each block by using (a) deference of the first reference image and the second reference image, (b) sum of the first gradient image and the third gradient image and (c) sum of the second gradient image and the fourth gradient image, and
the second value is determined for each block by using (a) deference of the first reference image and the second reference image, (b) sum of the first gradient image and the third gradient image and (c) sum of the second gradient image and the fourth gradient image and (d) the first value.