Motion estimation or P-type images using direct mode prediction

1. A process for a differential coding of a current image block in a first reference image on a basis of a predicted image block in a second reference image, the reference images flanking one or more images of bidirectional type or type B whose differential coding can take account of the blocks of these reference images, comprising steps of using a coding device to perform:

calculating, for a candidate block in the second reference image, a difference between the current block and the candidate block designated by a motion vector Vp which is allocated to the current block;

calculating, for said candidate block, a difference between a co-located block of the current image block in an image of type B and the block in a reference image, the latter block being designated by a colinear motion vector, which is allocated to the co-located block and is colinear with the motion vector Vp;

selecting a predicted image block as the one, among the candidate blocks, minimizing a function of at least the two calculated differences, to give a selected predicted image block; and

differential coding of the current image block by using said selected predicted image block.

2. The process according to claim 1 , wherein the colinear motion vector is a backward motion vector between the image of type B and the first reference image.

3. The process according to claim 1 , wherein the colinear motion vector is a forward motion vector between the image of type B and the second reference image.

4. The process according to claim 1 , wherein the selecting step minimizes a function of three calculated differences, a difference between the current block and the candidate block, a difference between the co-located block and a block in the second reference image corresponding to a first colinear motion vector between the image of type B and the second reference image, and a difference between the co-located block and a block in the first reference image corresponding to a second colinear motion vector between the image of type B and the first reference image.

5. The process according to claim 1 , the reference images flanking two images B 0 and B 1 of bidirectional type, wherein the selecting step minimizes a function of:

a difference between the current block and the candidate block,

a difference between a co-located block within B 0 and a block in a reference image corresponding to a first colinear motion vector between the first image of type B, B o , and a reference image, and

a difference between a co-located block within B 1 and a block in a reference image corresponding to a second colinear motion vector between the second image of type B, B 1 , and a reference image.

6. The process according to claim 1 , wherein a difference is calculated on a basis of sums of absolute values of differences (SAD) between the luminance values of the pixels of the blocks.

7. The process according to claim 1 , wherein a difference is calculated on a basis of sums of absolute values of coefficients of a transform of a residual which is a pixelwise difference between the luminance values of the pixels of the blocks.

8. The process according to claim 1 , wherein the candidate blocks are image blocks of a same size as the current block, that can be extracted from a predefined search window.

9. The process according to claim 1 , wherein one of the coding modes of images of the B type is a temporal direct mode of the MPEG 4 standard.

10. The process according to claim 1 , wherein said function is an affine function.

11. The process according to claim 10 , wherein said affine function is a weighted sum of the differences.