Method of and arrangement for estimating and compensatiing motion in a sequence of pictures and a picture transmission system provided with such an arrangement

Claim 1

A method of estimating and compensating motion in a sequence of pictures in which the brightness of each picture element is expressed digitally, characterized in that each picture is divided into blocks of I×J elements, each of these blocks being defined by two coordinates (m, n) which denote the rank m of the picture line in which a block is located and the position n of the block in this m th line, respectively, and in that said method comprises: (A) preserving three consecutive pictures in the sequence of initial pictures, denoted by the respective ranks 2k, 2k+1, 2k+2; (B) defining for the blocks three groups in which they are classified, a first group wherein said blocks are estimated by corrected horizontal recursion, a second group wherein said blocks are estimated by non-corrected horizontal recursion and a third group wherein said blocks are estimated by vertical recursion, in which groups the blocks will be classified, with a delay of L blocks, in accordance with the progressing picture scan, by examining, to effect said classification, a predetermined number of sequences of possible groups for the L blocks subsequent to the considered current blocks X, the criterion on which this classification is based being the difference in brightness between two pictures in the sequence of pictures, taking the displacement which occurs between one picture and another into account; (C) forming a classifying tree whose three branches seen at each branching point are equal in number to those of the possible groups and whose 3 L branches occurring as a result thereof at each successive level 1 to L correspond to the sequences of possible groups for the L consecutive considered blocks; (D) assigning two parameters to each branch of each level: (a) a first parameter, the displacement vector, which is representative of any displacement of the current block of the picture 2k+1 relative to the preceding and subsequent pictures 2k and 2k+2 and is calculated for the element which is representative of said block with the aid of a recursive displacement estimation method; (b) a second parameter, the accumulated distortion parameter, which is representative of the accumulated sum of the approximation error of the current block and those of the preceding blocks in the number L of blocks, each approximation error being rendered minimal by taking into account the influence of an estimated displacement for a block on the displacement estimated for the L subsequent blocks, and said influence possible not being investigated, among the 3 L possible branches of the classifying tree, with the exception of M branches of the weakest accumulated distortion, referred to as surviving branches, M being a limited number less than the maximum number of branches 3 L ; (E) taking, at the end of this method, as regards L consecutive blocks in the scanning direction, a majority decision as regards the effective group of the block which, relative to the current block, is located L blocks earlier, said decision being taken by examining the surviving branches and by classifying said blocks situated L blocks earlier in the group to which the majority of surviving branches correspond, and said decision procedure being repeated in an identical manner with the exception of the L last blocks for which no majority decision is taken.

Claim 2

A method of estimating and compensating motion in a sequence of pictures in which the brightness of each picture element is expressed digitally, characterized in that each picture is divided into blocks of I×J elements, each of these blocks being defined by two coordinates (m, n) which denote the rank m of the picture line in which a block is located and the position n of the block in this m th line, respectively, and in that said method comprises: (A) preserving three consecutive pictures in the sequence of initial pictures, denoted by the respective ranks 2k, 2k+1, 2k+2; (B) defining for the blocks three groups in which they are classified, a first group wherein said blocks are estimated by corrected horizontal recursion, a second group wherein said blocks are estimated by non-corrected horizontal recursion and a third group wherein said blocks are estimated by vertical recursion, in which groups the blocks will be classified, with a delay of L blocks, in accordance with the progressing picture scan, by examining, to effect said classification, a predetermined number of sequences of possible groups for the L blocks subsequent to the considered current blocks X, the criterion on which this classification is based being the difference in brightness between two pictures in the sequence of pictures, taking the displacement which occurs between one picture and another into account; (C) forming a classifying tree whose three branches seen at each branching point are equal in number to those of the possible groups and whose 3 L branches occurring as a result thereof at each successive level 1 to L correspond to the sequences of possible groups for the L consecutive considered blocks; (D) assigning two parameters to each branch of each level: (a) a first parameter, the displacement vector, which is representative of any displacement of the current block of the picture 2k+1 relative to the preceding and subsequent pictures 2k and 2k+2 and is calculated for the element which is representative of said block with the aid of a recursive displacement estimation method; (b) a second parameter, the accumulated distortion parameter, which is representative of the accumulated sum of the approximation error of the current block and those of the preceding blocks in the number L of blocks, each approximation error being rendered minimal by taking into account the influence of an estimated displacement for a block on the displacement estimated for the L subsequent blocks, and said influence possible not being investigated, among the 3 L possible branches of the classifying tree, with the exception of M branches of the weakest accumulated distortion, referred to as surviving branches, M being a limited number less than the maximum number of branches 3 L ; (E) taking, at the end of this method, as regards L consecutive blocks in the scanning direction, a majority decision as regards the effective group of the block which, relative to the current block, is located L blocks earlier, said decision being taken by examining the surviving branches and by classifying said blocks situated L blocks earlier in the group to which the majority of surviving branches correspond, and said decision procedure being repeated in an identical manner with the exception of the L last blocks for which no majority decision is taken.