Model based bit rate control for a macroblock encoder

• US 7,720,145 B2
• Filed: 05/05/2005
• Issued: 05/18/2010
• Est. Priority Date: 05/13/2004
• Status: Active Grant

• Current Assignee: Ittiam Systems Pte Ltd
• Original Assignee: Ittiam Systems Pte Ltd
• Inventors: Raman, Arvind, Sethuraman, Sriram, Parameswaran, Sankaranarayanan, Tamia, Dileep Kumar, Rao, Bhavani Gopalakrishna, Muthukrishnan, Murali Babu
• Primary Examiner(s): WONG, ALLEN C
• Application Number:
  US11/122,928
• Publication Number:
  US 20050254578A1
• Time in Patent Office: 1,839 Days
• Field of Search: 375/240, 375/240.01, 375/240.06, 375/240.12, 375/240.14, 375/240.16, 375/241, 375/240.02, 375/240.03
• US Class Current: 375/240.03
• CPC Class Codes: H04N 19/124 Quantisation
• CPC Class Codes: H04N 19/132 Sampling, masking or truncation of coding units, e.g. adaptive resampling, frame skipping, frame interpolation or high-frequency transform coefficient masking
• CPC Class Codes: H04N 19/149 by estimating the code amount by means of a model, e.g. mathematical model or statistical model
• CPC Class Codes: H04N 19/152 by measuring the fullness of the transmission buffer
• CPC Class Codes: H04N 19/176 the region being a block, e.g. a macroblock
• CPC Class Codes: H04N 19/587 involving temporal sub-sampling or interpolation, e.g. decimation or subsequent interpolation of pictures in a video sequence
• CPC Class Codes: H04N 19/61 in combination with predictive coding

Abstract

A method for achieving bit rate controlled encoding, e.g., constant bit rate, uses a bit rate control model based on bits allocated per macroblock and based on error in estimating bits consumed in an encoding process in a macroblock. The method computes bits consumed per macroblock as a function of the formed bit rate control model, and allocates bits for the macroblock. To this end, a quadratic (second order) equation is used for the model, which equation might default into a first order equation requiring fewer computations, in the event that no second order solutions exist. In one form, the bit rate control model computes the bit rate to satisfy video buffer verifier (VBV) compliance in accordance with MPEG requirements. The bit rate control model makes allowances for the type of macroblock, i.e., inter, intra or bidirectional, and provides for any frame-skips.

Claims

1. An article comprising a computer readable medium having instructions thereon which when executed by a computing platform result in execution of a method for computing a quantizer value (Q) in a video encoder, comprising:

modeling bits (B) required to code a current macroblock in a frame, in a rate control unit of the video encoder, using an equation

B=a*S/Q+b*S/Q²,

where a and b are coefficients of the equation, S is signal energy of the current macroblock, and Q is the quantizer value used to code the current macroblock;

determining values of the coefficients a and b of the equation using a set of values of R_i, S_i, and Q_i in the rate control unit, where R_i is the actual number of bits consumed by an i^th coded macroblock with signal energy (S_i) at a quantizer value (Q_i), and i varies from 0 to N−

1, where N is a number of coded macroblocks in the frame up to the current macroblock; and

,

computing the quantizer value (Q) for the current macroblock, in the rate control unit, by solving the equation with the determined values of coefficients a and b, the signal energy (S) for the current macroblock, and the bits (B) allocated for the current macroblock.

2. A method for computing a quantizer value (Q) in a video encoder, comprising:

modeling bits (B) required to code a current macroblock in a frame, in a rate control unit of the video encoder, using an equation

B=a*S/Q+b*S/Q²,

where a and b are coefficients of the equation, S is signal energy of the current macroblock, and Q is the quantizer value used to code the current macroblock;

determining values of the coefficients a and b of the equation using a set of values of R_i, S_i, and Q_i in the rate control unit, where R_i is the actual number of bits consumed by an i^th coded macroblock with signal energy (S_i) at a quantizer value (Q_i), and i varies from 0 to N−

1, where N is a number of coded macroblocks in the frame up to the current macroblock; and

,

computing the quantizer value (Q) for the current macroblock, in the rate control unit, by solving the equation with the determined values of coefficients a and b, the signal energy (S) for the current macroblock, and the bits (B) allocated for the current macroblock.

3. The method of claim 2, wherein the values of the coefficients a and b are determined using the equations:

a={Σ

(R_i/S_i)*Σ

(1/Q_i)−

Σ

(R_iQ_i/S_i)*Σ

(1/Q²_i)}/{Σ

(1/Q²_i)−

N*Σ

(1/Q²_i)}, and

b={Σ

(R_iQ_i/S_i)*Σ

(1/Q_i)−

N*Σ

(R_i/S_i)}/{Σ

(1/Q²_i)−

N*Σ

(1/Q²_i)},

Wherein summation extends over i=0 to i=N−

1, where N is the number of coded macroblocks in the frame up to the current macroblock.

4. The method of claim 2, wherein the bits (B) allocated for the current macroblock is computed using the equation:

bits allocated per macroblock=bits per frame/number of macroblocks.

5. The method of claim 4, wherein the bits per frame is computed based on bits per second desired by a user, a ratio of I pictures to P pictures in a second and a factor by which bit allocation for the I pictures is higher than for P pictures.

6. The method of claim 5, wherein the computed bits per frame is modified based on the video buffering verifier (VBV) requirement for the MPEG standards.

7. The method of claim 2, wherein the step of allocating the bits for the macroblock is computed using the equation:

bits allocated per macroblock=(bits in reservoir-actual bits consumed by the previous macroblock)/remaining number of macroblocks.

8. The method of claim 2 further comprising:

computing a difference in bits between the bits (B) used for computing the quantizer value (Q) of the current macroblock and bits estimated for the current macroblock at a quantizer value (Q_old) of the previous macroblock;

if the difference in bits is greater than or equal to a threshold value, then quantizing the current macroblock using the computed quantizer value (Q) of the current macroblock; and

if the difference in bits is below the threshold value, then quantizing the current macroblock using the quantizer value (Q_old) of the previous macroblock.

9. The method of claim 2, wherein the step of computing a quantizer value further comprises:

if the computed quantizer value is not a real number, then degenerating the equation to an equation in Q, given by

B=c*S/Q

where c is coefficient of the equation, S is the signal energy of the current macroblock, Q is the quantizer value used to code the current macroblock, and B is the bits required to code the current macroblock in the frame;

determining the coefficient c of the linear equation as a least squares estimate using the set of values of R_i, S_i, and Q_i, where R_i is the actual number of bits consumed by the i^th coded macroblock with signal energy (S_i) at a quantizer value (Q_i), and i varies from 0 to N−

1, where N is the number of coded macroblocks in the frame up to the current macroblock; and

,

computing the quantizer value (Q) for the current macroblock by solving the linear equation with the determined value of coefficient c, the signal energy (S) for the current macroblock, and the bits (B) allocated for the current macroblock.

10. The method of claim 2, wherein the step of determining the values of the coefficients of the equation comprises:

determining values of model coefficients for intra and inter (or bi-directional) macroblocks separately.

11. The method of claim 10, wherein, for coding intra coded frames, signal energy is assumed to be a sum of all the pixel values in the current macroblock to be encoded.

12. The method of claim 10, wherein forming the model for an inter macroblock comprises:

dividing the allocated bits per macroblock into bits for motion vector data and coefficient data.

13. The method of claim 12, wherein the bits for the motion vector data are computed by a predetermined ratio of the allocated bits per macroblock.

14. The method of claim 13, wherein the model is used for computing the quantizer value based on the allocated bits per coefficient data.

15. The method of claim 14, further comprising:

scaling motion vector bits based on a number of frame-skips.

16. The method of claim 15, where, in case of P pictures, the method includes using frame level average to determine bits consumed by motion vectors, and using a predetermined factor of allocated bits per frame for allocating motion vector bits for an initial P frame.

17. The method of claim 2, wherein the allocation of bits per macroblock is done for a group of pictures (GOP).

18. The method of claim 17 further comprising changing a GOP structure dynamically to suit scene changes such that a scene change is coded as an Intra frame, and a new GOP begins from the Intra frame.

19. The method of claim 2 further comprising resetting the values of coefficients a and b at the beginning of every frame and the average quantizer value of the previous frame is used for the first few macroblocks, and resulting error in estimation is used to determine the values of a and b.