Method for Scalable Video Coding

• US 20090304090A1
• Filed: 09/29/2005
• Published: 12/10/2009
• Est. Priority Date: 09/29/2005
• Status: Active Grant

• Current Assignee: Telecom Italia S.p.A.
• Original Assignee: Telecom Italia S.p.A.
• Inventors: Francini, Gianluca, Lepsoy, Skjalg, Cordara, Giovanni
• Granted Patent: US 8,625,678 B2
• Time in Patent Office: Days
• US Class Current: 375/240.260
• CPC Class Codes: H04N 19/53 Multi-resolution motion estimation; Hierarchical motion estimation
• CPC Class Codes: H04N 19/537 Motion estimation other than block-based

Abstract

A method for estimating motion for the scalable video coding, includes the step of estimating the motion field of a sequence of photograms which can be represented with a plurality of space resolution levels including computing the motion field for the minimum resolution level and, until the maximum resolution level is reached, repeating the steps of: rising by one resolution level; extracting the photograms for such resolution level; and computing the motion field for such resolution level. The motion field is computed through an optical flow equation which contains, for every higher level than the minimum resolution level, a regularization factor between levels which points out the difference between the solution for the considered level and the solution for the immediately lower resolution level. A more or less high value of the regularization factor implies more or less relevant changes of the component at the considered resolution during the following process iterations.

Claims

1-17. -17. (canceled)

18. A method for a scalable video coding, comprising the step of estimating a motion field (p) of a sequence of photograms (f₁, f₂, . . . , f_T) which can be represented with a plurality of space resolution levels (K, . . . , K−

m, . . . , K−

M) included between a minimum resolution level (K−

M) and a maximum resolution level (K), wherein the step of estimating the motion field comprises the step of computing the motion field for each one of said space resolution levels by the minimization of a function, wherein, for each level which is different from the minimum space resolution level, said function comprises a regularization term between levels which represents a difference between the solution of said minimization for the considered level and the solution of said minimization for the space resolution level which is immediately lower than the considered level.

19. The method according to claim 18, wherein said step of computing the motion field for every level different from the minimum space resolution level is performed sequentially passing from the minimum space resolution level to the maximum space resolution level.

20. The method according to claim 18, wherein, for each space resolution level, the motion field comprises an initial motion field (p⁰) and an incremental motion field (q) and said step of computing the motion field (p) comprises computing the incremental motion field (q) and summing the incremental motion field (q) to the initial motion field (p⁰).

21. The method according to claim 20, wherein, for a generic space resolution level with order K−

m different from the minimum space resolution level, said regularization term between levels can be expressed as;

F₃(q<sub>K−

m</sub>)=∥

E(q<sub>K−

m</sub>+p<sub>K−

m</sub>⁰)−

p<sub>K−

m−

1</sub>∥

²,

where E is a filtering and sub-sampling operator, p⁰ is the initial motion field and q is the incremental motion field, said incremental motion field q in a generic point with coordinates x, y being suitable to be expressed as

$q<br><br>\left(x,y\right)=\left[\begin{array}{c}\mathrm{\delta <br><br>}<br><br><br><br>u<br><br>\left(x,y\right)\\ \mathrm{\delta <br><br>}<br><br><br><br>v<br><br>\left(x,y\right)\end{array}\right],$

wherein δ

u and δ

v are the horizontal and vertical components of the incremental motion field.

22. The method according to claim 21, wherein said filtering and sub-sampling operator E is a low-pass component of a wavelet transform.

23. The method according to claim 21 wherein, for each level different from the minimum space resolution level, said minimization can be expressed as

q=arg min{F₁(q)+λ

F₂(q)+λ

F₃(q)}

where F₁(q) is a term designating the approximation error, F₂(q) is a term designating the motion field regularity and λ and

μ

are two regularization factors.

24. The method according to claim 23, wherein, for said minimum space resolution level, said minimization can be expressed as:

q=arg min{F₁(q)+λ

F₂(q)}

25. The method according to claim 23, wherein, taking into account two photograms g_t and g_t+1 at following instants t and t+1 at the generic resolution level K−

m, the term designating the approximation error can be expressed as;

F₁(q)=∥

A<sub>K−

m</sub>q−

(M(g_t,u₀,v₀)−

g_t+1)∥

², where;

u₀ and v₀ are the horizontal and vertical components of the initial motion field,

A<sub>k−

m</sub> is a matrix comprising two diagonal sub-matrixes containing space derivatives of photogram g_t+1, expressed as A<sub>K−

m</sub>=[diag(g_t+1^(x))diag(g_t+1^(y))], and

M is a motion operator adapted to distort a generic photogram g in the following way;

M(g,u,v)(x,y)=g(x−

u(x,y),y−

v(x,y)).

26. The method according to claim 23, wherein the term designating the motion field regularity can be expressed as

F₂(q)=∥

Hq∥

², where H is a second derivative operator which can be expressed as;

$H=\left[\begin{array}{cc}\frac{{<br><br>}^2}{<br><br>x^2}& \\ \frac{{<br><br>}^2}{<br><br>y^2}& \\ & \frac{{<br><br>}^2}{<br><br>x^2}\\ & \frac{{<br><br>}^2}{<br><br>y^2}\end{array}\right]$

27. The method according to claim 18, comprising, before the steps of computing the motion fields, the further step of building photograms at different space resolution levels starting from photograms at maximum space resolution level.

28. The method according to claim 27, wherein, for each space resolution level, the step of building photograms comprises filtering and sub-sampling photograms at maximum space resolution level, said step of filtering and sub-sampling being performed a number of times equal to the number of levels separating the considered space resolution level from the maximum space resolution level.

29. The method according to claim 28, wherein the step of filtering and sub-sampling comprises the following steps:

low-pass filtering along the rows;

sub-sampling the columns;

low-pass filtering along the columns; and

sub-sampling the rows.

30. The method according to claim 18, wherein said function can be expressed through an optical flow equation.

31. The method for transmitting coded video images, comprising the steps of coding a video image according to the method of claim 18, transmitting the coded image, receiving the coded image and decoding the received coded image.

32. The method according to claim 31, wherein the step of decoding comprises applying to said coded video image at least one transform of the wavelet type for obtaining a sequence of photograms at a desired resolution level.

33. A computer program capable of being directly loaded in a memory of a data processing system and adapted to implement, when run, the method according to claim 18.

34. A system comprising means for performing the steps of the method according to claim 18.