Highly efficient picture coding method with motion-adaptive zonal sampling enabling optimized image signal compression

Claim 2

A picture coding method comprising the steps of: receiving an uncompressed image signal; translating a sample point in the uncompressed image signal from each of at least K-frames into a 3-dimensional block in the form of an M×N×K sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 3-dimensional block and represent by f (x, y, t) (x=0 to M-1, y=0 to N-1, t=0 to K-1) is orthogonally transformed to a transform coefficient of 3-dimension expressed as F (u, v, w) (u=0 to M-1, v=0 to N-1, w=0 to K-1); calculating the movement of the 2-dimensional block which is a part of the 3-dimensional block during a period of time between t=0 and t=K-1; determining whether or not the transform coefficient is to be coded based on the calculated movement; and outputting a compressed image signal in accordance with the coded transform coefficients, wherein spatial resolution is reduced and the time resolution is increased in regions where the calculated movement is high and wherein the spatial resolution is increased and the time resolution decreased in regions where the calculated movement is low.

Claim 3

A picture coding method according to claim 1, wherein the movement of the 2-dimensional block is determined as a combination of the horizontal movement mx and the vertical movement my, by means of block matching between t=0 and t=k-1, with reference to a 2-dimensional block, at t=0, of the 3-dimensional block f (x, y, t).

Claim 4

A picture coding method according to claim 1, wherein assuming that the movement of a 2-dimensional block at t=K-1 is a combination of the horizontal movement mx and the vertical movement my, by means of block matching on the basis of a 2-dimensional block, at t=0, of the 3-dimensional block, the movement of the 2-dimensional block is determined as a combination of the horizontal movement mx'=mx-(cx'-cx) and the vertical movement my'=my-(cy'-cy) if the coordinate center of the 2-dimensional block is expressed at t=0 as (cx, cy) and at t=K-1 as (cx', cy').

Claim 5

A picture coding method comprising the steps of: receiving an uncompressed image signal; translating a sample point in the uncompressed image signal from each of at least K-frames into a 3-dimensional block in the form of an M×N×K sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 3-dimensional block and represented by f (x, y, t) (x=0 to M-1, y=0 to N-1, t=0 to K-1) is orthogonally transformed to a transform coefficient of 3-dimensions expressed as F (u, v, w) (u=0 to M-1, v=0 to N-1, w=0 to K-1); and identifying the 3-dimensional block as a motion block if the energy ratio E0/E1 is smaller than a predetermined value or a still block if the energy ratio E0/E1 is greater than the predetermined value where E0 is the entire energy of the transform coefficient F (u, v, 0) which is present on a first plane extending across a point w=0 on the w-axis provided in the transform domain corresponding to the reverse axis of the time base t in a real space, expressed as ##EQU4## and E1 is the entire energy of the transform coefficient F (u, v, w) (w≠0) which is present on any plane after the first place associated with w≠0, expressed as ##EQU5## determining whether or not the transform coefficient is to be coded based on whether or not the block is a motion block or a still block; and outputting a compressed image signal in accordance with the coded transform coefficients, wherein spatial resolution is reduced and the time resolution is increased in motion block regions and wherein the spatial resolution is increased and the time resolution decreased in the still block regions.

Claim 6

A picture coding method comprising the steps of: receiving an uncompressed image signal; translating a sample point in the uncompressed image signal from each of at least K-frames into a 3-dimensional block in the form of an M×N×K sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 3-dimensional block and represented by f (x, y, t) (x=0 to M-1, y=0 to N-1, t=0 to K-1) is orthogonally transformed to a transform coefficient of 3-dimensions expressed as F (u, v, w) (u=0 to M-1, v=0 to N-1, w=0 to K-1); and identifying the 3-dimensional block as a motion block if the energy ratio E0/E1 is smaller than a predetermined value or a still block if the energy ratio E0/E1 is greater than the predetermined value where E0 is the entire energy of the transform coefficient F (u, v, 0) which is present on a first plane extending across a point w=0 on the w-axis provided in the transform domain corresponding to the reverse axis of the time base t in a real space, expressed as ##EQU4## and E1 is the entire energy of the transform coefficient F (u, v, w) (w≠0) which is present on any plane after the first place associated with w≠0, expressed as ##EQU5## determining whether or not the transform coefficient is to be coded based on whether or not the block is a motion block or a still block; and outputting a compressed image signal in accordance with the coded transform coefficients, wherein spatial resolution is reduced and the time resolution is increased in motion block regions and wherein the spatial resolution is increased and the time resolution decreased in the still block regions.

Claim 7

A picture coding method according to claim 4, wherein the movement of the 3-dimensional block is determined as a combination of the horizontal movement mx=Ec and the vertical movement my=a×Ec, with reference to the energy ratio Ec=E1/E0 of the transform coefficient in the 3-dimensional block.

Claim 8

A picture coding method according to claims 2, 3, or 5 wherein the orthogonal transform coefficient F (u, v, w) satisfying u<u1, v<v1, or w<w1 is coded if |m|x>mxh, |my|>myh, or |mz|<mzh respectively, where |my| is the absolute horizontal movement, |my| is the absolute vertical movement, mz is a square sum of the horizontal and vertical movements expressed as mz=mx 2 +my 2 , mxh, myh, and mzh are the thresholds of given movements, and u1, v1, and w1 are the thresholds of transform coefficients.

Claim 9

A picture coding method according to claims 2, 3, or 5, wherein assuming that the thresholds of transform coefficients up, vp, and wp are determined as up u1 if |mx|>mxh, up=u2 (where u1<u2) if |mx|≦mxh, vp=v1 if |my|>myh, vp=v2 (where v1<v2) if |my|≦myh, wp=w1 if mz≦mzh, and wp=w2 ( where w1<w2) if mz>mzh, where |mx| is the absolute horizontal movement, |my| is the absolute vertical movement, mz is a square sum of the horizontal and vertical movements expressed as mz=mx 2 +my 2 , mxh, myh, and mzh are the thresholds of given movements, and u1, v1, and w1 are the thresholds of transform coefficients, the orthogonal transform coefficient F (u, v, w) satisfying p (u, v, w)<c is coded where p (u, v, w) is the notation of a plane extending across the points (up, 0, 0), (0, vp, 0), and (0, 0, wp) in the 3-dimension orthogonal transform domain and c is a constant.

Claim 10

A picture coding method according to claims 2, 3, or 5, wherein assuming that the thresholds of transform coefficients up, vp, and wp are determined as up=u1 if |mx|>mxh, up=u2 (where u1<u2) if |mx|≦mxh, vp=v1 if |my|>myh, vp=v2 (where v1 v2) if |my|≦myh, wp=w1 if mz<mzh, and wp w2 (where w1<w2) if mz≦mzh, where |mx| is the absolute horizontal movement, |my| is the absolute vertical movement, mz is a square sum of the horizontal and vertical movements expressed as mz=mx 2 +my 2 , mxh, myh, and mzh are the thresholds of given movements, and u1, v1, and w1 are the thresholds of transform coefficients, the orthogonal transform coefficient F (u, v, w) satisfying g (u, v, w)<c is coded where g (u, v, w) is the notation of a curved plane extending across the points (up, 0, 0), (0, vp, 0), and (0, 0, wp) in the 3-dimension orthogonal transform domain and c is a constant.

Claim 11

A picture coding method comprising the steps of: receiving an uncompressed image signal; translating a sample point in the uncompressed image signal at the time t into a 2-dimensional block in the form of an M×N sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 2-dimensional block and represented by f (x, y) (x=0 to M-1, y=0 to N-1) is orthogonally transformed to a transform coefficient of 2-dimension expressed as F (u, v) (u=0 to M-1, v=0 to N-1); calculating the horizontal movement mx and the vertical movement my of the 2-dimensional block at the time t+1 by means of block matching; determining whether or not the transform coefficient is to be coded based on the calculated movement; and outputting a compressed image signal in accordance with the coded transform coefficients, whether spatial resolution is reduced in motion block regions and is increased in still block regions.

Claim 12

A picture coding method comprising the steps of: receiving an uncompressed image signal; translating a sample point in the uncompressed image signal at the time t into a 2-dimensional block in the form of an M×N sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 2-dimensional block and represented by f (x, y) (x=0 to M-1, y=0 to N-1) is orthogonally transformed to a transform coefficient of 2-dimension expressed as F (u, v) (u=0 to M-1, v=0 to N-1); calculating the horizontal movement mx and the vertical movement my of the 2-dimensional block at the time t+1 by means of block matching; determining whether or not the transform coefficient is to be coded based on the calculated movement; and outputting a compressed image signal in accordance with the coded transform coefficients, whether spatial resolution is reduced in motion block regions and is increased in still block regions.

Claim 13

A picture coding method according to claim 9, wherein the orthogonal transform coefficient F (u, v) satisfying u<u1 or v<v1 is coded if |mx|>mxh or |my|<myh respectively, where |mx| is the absolute horizontal movement, |my| is the absolute vertical movement, mxh and myh are the thresholds of given movements, and u1 and v1 are the thresholds of transform coefficients.

Claim 14

A picture coding method according to claim 9, wherein assuming that the thresholds of transform coefficients up and vp are determined as up=u1 if |mx|>mxh, up=u2 (where u1<u2) if |mx|<mxh, vp=v1 if |my|>myh, and vp=v2 (where v1<v2) if |my|≦myh, where |mx| is the absolute horizontal movement, |my| is the absolute vertical movement, mxh and myh are the thresholds of given movements, and u1 and v1 are the thresholds of transform coefficients, the orthogonal transform coefficient F (u, v) satisfying p (u, v)<c is coded where p (u, v) is the notation of a line extending across the points (up, 0) and (0, vp) in the 2-dimension orthogonal transform domain and c is constant.

Claim 15

A picture coding method according to claim 9, wherein assuming that the thresholds of transform coefficients up and vp are determined as up=u1 if |mx|>mxh, up=u2 (where u1<u2) if |mx|≦mxh, vp=v1 if |my|>myh, and vp=v2 (where v1<v2) if |my|≦myh, where |mx| is the absolute horizontal movement, |my| is the absolute vertical movement, mxh and myh are the thresholds of given movements, and u1 and v1 are the thresholds of transform coefficients, the orthogonal transform coefficient F (u, v) satisfying g (u, v)<c is coded where g (u, v) is the notation of a curved line extending across the points (up, 0) and (0, vp) in the 2-dimension orthogonal transform domain and c is a constant.

Claim 16

A picture coding method according to any of claims 1 to 3, 5, 9-12, wherein the transform coefficient to be omitted from coding in the transform domain is designated as 0.

Claim 17

A picture coding method according to any of claims 1 to 3, 5, 9-12, wherein the transform coefficients to be omitted from coding are quantified at greater steps than those to be coded.

Claim 18

A picture coding method according to any of claims 1 to 3, 5, 9-12, wherein the transform coefficients to be omitted from coding in the transform domain are not output to the transmission line.

Claim 19

A picture coding method comprising the steps of: receiving an uncompressed image signal; translating a sample point in the uncompressed image signal from each of at least K-frames into a 3-dimensional block in the form of an M×N×K sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 3-dimensional block and represented by f (x, y, t) (x=0 to M-1, y=0 to N-1 t=0 to K-1) is orthogonally transformed to a transform coefficient of 3-dimension expressed as F (u, v, w) (u=0 to M-1, v=0 to N-1, w=0 to K-1); separating a group of the F (u, v, w) coefficients into motion blocks and still blocks according to results of a motion detection; and after separation into the motion and still blocks, dividing the group of the F(u, v, w) coefficients into k1 and k2 classes by the amount of energy of each 3-dimensional block expresses as ##EQU6## and determining the allocation of bits in each class in proportion to the variance of transform coefficient of the 3-dimensional block in the class, the variance being represented by ##EQU7## determining whether or not the transform coefficient is to be coded based on whether or not the block is a motion block or a still block; and outputting a compressed image signal in accordance with the coded transform coefficients, wherein spatial resolution is reduced and the time resolution is increased in motion block regions and wherein the spatial resolution is increased and the time resolution decreased in the still block regions.

Claim 20

A picture coding method comprising the steps of: receiving an uncompressed image signal; translating a sample point in the uncompressed image signal from each of at least K-frames into a 3-dimensional block in the form of an M×N×K sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 3-dimensional block and represented by f (x, y, t) (x=0 to M-1, y=0 to N-1 t=0 to K-1) is orthogonally transformed to a transform coefficient of 3-dimension expressed as F (u, v, w) (u=0 to M-1, v=0 to N-1, w=0 to K-1); separating a group of the F (u, v, w) coefficients into motion blocks and still blocks according to results of a motion detection; and after separation into the motion and still blocks, dividing the group of the F(u, v, w) coefficients into k1 and k2 classes by the amount of energy of each 3-dimensional block expresses as ##EQU6## and determining the allocation of bits in each class in proportion to the variance of transform coefficient of the 3-dimensional block in the class, the variance being represented by ##EQU7## determining whether or not the transform coefficient is to be coded based on whether or not the block is a motion block or a still block; and outputting a compressed image signal in accordance with the coded transform coefficients, wherein spatial resolution is reduced and the time resolution is increased in motion block regions and wherein the spatial resolution is increased and the time resolution decreased in the still block regions.

Claim 21

A picture coding method according to claim 16, wherein the movement of a 2-dimensional block of the 3-dimensional block f (x, y, z) is determined as a combination of the horizontal movement mx and the vertical movement my, by means of block matching between t=0 and t=K-1, with reference to a 2-dimensional block at t=0.

Claim 22

A picture coding method according to claim 16, wherein assuming that the movement of a 2-dimensional block at t=K-1 is a combination of the horizontal movement mx and the vertical movement my, by means of block matching on the basis of a 2-dimensional block, at t=0, of the 3-dimensional block, the movement of the 2-dimensional block is determined as a cobination of the horizontal movement mx'=mx-(cx'-cx) and the vertical movement my'=my-(cy'-cy) if the coordinate center of the 2-dimensional block is expressed at t=0 as (cx, cy) and at t=K-1 as (cx', cy').

Claim 23

A picture coding method comprising the steps of: receiving an uncompressed image signal; translating a sample point in the uncompressed image signal at the time t into a 2-dimensional block in the form of an M×N sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 2-dimensional block and represented by f (x, y) x=0 to M-1, y=0 to N-1) is orthogonally transformed to a transform coefficient of 2-dimension expressed as F (u, v) (u=0 to M-1, v=0 to N-1); calculating the horizontal movement mx and the vertical movement my of the 2-dimensional block at t=1 by means of block matching; separating a group of the F (u, v) coefficients into motion blocks and still blocks according to results of a motion detection; and after separation into the motion and still blocks, dividing the group of the F (u, v) coefficients into k1 and k2 classes by the amount of energy of each 2-dimensional block expressed as ##EQU8## and determining the allocation of bits in each class in proportion to the variance of a transform coefficient of the 2-dimensional block in the class, the variance being represented by ##EQU9## determining whether or not the transform coefficient is to be coded based on whether or not the block is a motion block or a still block; and outputting a compressed image signal in accordance with the coded transform coefficients, wherein spatial resolution is reduced in motion block regions and is increased in still block regions.

Claim 24

A picture coding method comprising the steps of: receiving an uncompressed image signal; translating a sample point in the uncompressed image signal at the time t into a 2-dimensional block in the form of an M×N sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 2-dimensional block and represented by f (x, y) x=0 to M-1, y=0 to N-1) is orthogonally transformed to a transform coefficient of 2-dimension expressed as F (u, v) (u=0 to M-1, v=0 to N-1); calculating the horizontal movement mx and the vertical movement my of the 2-dimensional block at t=1 by means of block matching; separating a group of the F (u, v) coefficients into motion blocks and still blocks according to results of a motion detection; and after separation into the motion and still blocks, dividing the group of the F (u, v) coefficients into k1 and k2 classes by the amount of energy of each 2-dimensional block expressed as ##EQU8## and determining the allocation of bits in each class in proportion to the variance of a transform coefficient of the 2-dimensional block in the class, the variance being represented by ##EQU9## determining whether or not the transform coefficient is to be coded based on whether or not the block is a motion block or a still block; and outputting a compressed image signal in accordance with the coded transform coefficients, wherein spatial resolution is reduced in motion block regions and is increased in still block regions.

Claim 25

A picture coding method comprising the steps of: receiving an uncompressed image signal translating a sample point in the uncompressed image signal from each of at least K-frames into a 3-dimensional block in the form of an M×N×K sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 3-dimensional block and represented by f (x, y, t) (x=0 to M-1, y=0 to N-1, t=0 to K-1) is orthogonally transformed to a transform coefficient of 3-dimension expressed as F (u, v, w) (u=0 to M-1, v=0 to N-1, w=0 to K-1); coding the transform coefficients in a sequence from a first plane extending across a point w=0 on the w-axis provided in the transform domain corresponding to the reverse axis of the time base in a real domain, a second plane extending across a point w=1 on the w-axis, a third plan, . . . to a (K-1)th plane; and outputting a compressed image signal in accordance with the coded transform coefficients.

Claim 26

A picture coding method comprising the steps of: receiving an uncompressed image signal translating a sample point in the uncompressed image signal from each of at least K-frames into a 3-dimensional block in the form of an M×N×K sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 3-dimensional block and represented by f (x, y, t) (x=0 to M-1, y=0 to N-1, t=0 to K-1) is orthogonally transformed to a transform coefficient of 3-dimension expressed as F (u, v, w) (u=0 to M-1, v=0 to N-1, w=0 to K-1); coding the transform coefficients in a sequence from a first plane extending across a point w=0 on the w-axis provided in the transform domain corresponding to the reverse axis of the time base in a real domain, a second plane extending across a point w=1 on the w-axis, a third plan, . . . to a (K-1)th plane; and outputting a compressed image signal in accordance with the coded transform coefficients.

Claim 27

A picture coding method comprising the steps of: receiving an uncompressed image; translating a sample point in the uncompressed image signal from each of at least K-frames into a 3-dimensional block in the form of an M×N×K sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 3-dimensional block and represented by f (x, y, t) x=0 to M-1, y=0 to N-1, t=0 to K-1) is orthogonally transformed to a transform coefficient of 3-dimension expressed as F (u, v, w) (u=0 to M-1, v=0 to N-1, w=0 to K-1); transmitting the F (u, v, w) transform coefficients in a sequence from the smallest thereof in the sum of u, v, and w are also, in which w is smaller for coding the F (u, v, w) transform coefficients in an order from the lowest sequency transform coefficient; and outputting a compressed image signal in accordance with the coded transform coefficients.

Claim 28

A picture coding method comprising the steps of: receiving an uncompressed image; translating a sample point in the uncompressed image signal from each of at least K-frames into a 3-dimensional block in the form of an M×N×K sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 3-dimensional block and represented by f (x, y, t) x=0 to M-1, y=0 to N-1, t=0 to K-1) is orthogonally transformed to a transform coefficient of 3-dimension expressed as F (u, v, w) (u=0 to M-1, v=0 to N-1, w=0 to K-1); transmitting the F (u, v, w) transform coefficients in a sequence from the smallest thereof in the sum of u, v, and w are also, in which w is smaller for coding the F (u, v, w) transform coefficients in an order from the lowest sequency transform coefficient; and outputting a compressed image signal in accordance with the coded transform coefficients.

Claim 29

An image processing method for detecting motion in picture formed by the image signal comprising the steps of: receiving an image signal; sampling the image signal and translating a sample point in the image signal from each of at least K-frames into a 3-dimensional block in the form of an M×N×K sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 3-dimensional block and represented by f (x, y, t) (x=0 to M-1, y=0 to N-1, t=0 to K-1) is orthogonally transformed to a transform coefficient of 3-dimensions expressed as F (u, v, w) u=0 to M-1, v=0 to N=1, w=0 to K-1); identifying the 3-dimensional block as a motion block if the energy ratio E0/E1 is smaller than a predetermined value or a still block if the energy ratio E0/E1 is greater than the predetermined value where E0 is the entire energy of the transform coefficient F (u, v, o) which is present on a first plane extending across a point w=0 on the w-axis provided in the transform domain corresponding to the reverse axis of the time base t in a real space, expressed as ##EQU10## is E1 and the entire energy of the transform coefficient F and E1 is the entire energy of the transform coefficient F (u, v, w) (w≠0) which is present on any plane after the first place associated with w≠0, expressed as ##EQU11## outputting results of the identification of the 3-dimensional block as a motion block or as a still block.

Claim 30

An image processing method for detecting motion in picture formed by the image signal comprising the steps of: receiving an image signal; sampling the image signal and translating a sample point in the image signal from each of at least K-frames into a 3-dimensional block in the form of an M×N×K sample point containing horizontal, vertical, and time-base components, in which the sample point which is a 3-dimensional block and represented by f (x, y, t) (x=0 to M-1, y=0 to N-1, t=0 to K-1) is orthogonally transformed to a transform coefficient of 3-dimensions expressed as F (u, v, w) u=0 to M-1, v=0 to N=1, w=0 to K-1); identifying the 3-dimensional block as a motion block if the energy ratio E0/E1 is smaller than a predetermined value or a still block if the energy ratio E0/E1 is greater than the predetermined value where E0 is the entire energy of the transform coefficient F (u, v, o) which is present on a first plane extending across a point w=0 on the w-axis provided in the transform domain corresponding to the reverse axis of the time base t in a real space, expressed as ##EQU10## is E1 and the entire energy of the transform coefficient F and E1 is the entire energy of the transform coefficient F (u, v, w) (w≠0) which is present on any plane after the first place associated with w≠0, expressed as ##EQU11## outputting results of the identification of the 3-dimensional block as a motion block or as a still block.