Process for controlling the outflow rate of a coder of digital data representative of sequences of images

• US 5,724,099 A
• Filed: 06/04/1996
• Issued: 03/03/1998
• Est. Priority Date: 07/10/1995
• Status: Expired due to Term

• Current Assignee: Gula Consulting Limited Liability Company (Intellectual Ventures LLC)
• Original Assignee: Orange S.A.
• Inventors: Roberts, James, Hamdi, Maher, Pierre, Rolin
• Primary Examiner(s): Chin, Tommy P.
• Assistant Examiner(s): DIEP, NHON THANH
• Application Number:
  US08/658,050
• Time in Patent Office: 637 Days
• Field of Search: 348/405, 348/419, 348/466
• US Class Current: 375/240.05
• CPC Class Codes: G06T 9/007 Transform coding, e.g. discrete cosine transform

Abstract

A process for controlling the output flow rate of a coder of digital data representative of sequences of images, the coding effected by the coder including a quantization stage of quantization parameter Q, the sequences of images being made up of successive groups of images. The process consists of determining, at the moment of the presence of the i group of images at the output of the coder, the quantization parameter Q(i+1) for the coding of the (i+1) group of images so that the flow rate at the output of the coder conforms with a version of a leaky bucket defined by a size M and by a leakage rate λ in terms of the activity of the scene represented by the i group of images.

Claims

1. A process for controlling a data output flow rate of a coder of digital data which is representative of a sequence of images, said coder including a quantization stage having a quantization parameter Q, said sequence of images comprising successive groups of images, said process comprising the steps of:

setting a quantization parameter Q(i+1) for coding an (i+1) group of images at a moment when data of an i group of images is at an output of the coder, said digital data having a flow rate at the output of said coder conforming to a version of a leaky bucket defined by a size M and by a leakage rate λ

; and

setting said quantization parameter Q(i+1) for the coding of the (i+1) group of images in terms of an activity of a scene represented by the i group of images, the scene of said i group of images being active when its open-loop flow rate is greater than the leakage rate λ

of the leaky bucket and being inactive when said open-loop flow rate is less than the leakage rate λ

, said open-loop flow rate of a group of images being the flow rate which said scene of images would have at an output of said coder if it had been quantized with a constant quantization parameter Qo at an average rate D over an entire sequence of images to which said scene of images belongs is equal to the leakage rate of said leaky bucket, the average rate D being an average rate when each group of the images is quantized with a constant parameter.

2. The process according to claim 1, and further steps of setting the quantization parameter Q(i+1) wherein:

active scenes have a flow rate at the output of the coder which is equal to (a) the open-loop flow rate when a rate of filling the bucket is zero, (b) the leakage rate of said leaky bucket when the rate of filling the bucket is equal to a maximum rate for filling said bucket, and (c) a value between the open-loop flow rate and the leakage rate when the rate of filling said bucket takes a non-zero value which is lower than its maximum rate of filling; and

,

inactive scenes have a flow rate at the output of the coder which is equal to (d) the leakage rate of said leaky bucket when the rate of filling the bucket is zero, (e) the open-loop flow rate when the rate of filling of the bucket is equal to the maximum rate of filling of said bucket, and (f) a value included between the leakage rate and the open-loop flow rate when the rate of filling said bucket takes a non-zero value which is lower than its maximum rate of filling.

3. The process according to either claim 1 or 2 and further comprising the steps of

setting the active scene quantization parameter by calculating, a function having (a) a variable rate of filling of the leaky bucket and (b) a parameter according to the leakage rate of said bucket (c) the maximum rate of filling said bucket, and (d) a predetermined quantization parameter;

and

setting the inactive scenes quantization parameter by using another function having variables and parameters which are the same.

4. The process according to either claim 1 or 2 further comprising the steps of setting a quantization parameter wherein:

(a) an active scene flow rate at the output of the coder is equal to a calculation of a continuously decreasing function f having a variable responsive to the rate of filling the leaky bucket which takes;

(i) a value equals to the open-loop flow rate as a zero value of the rate of filling of the bucket,

(ii) a value equals to the leakage rate of said leaky bucket as a value of the rate of filling the bucket is equal to the maximum rate of filling said bucket, and

(iii) a value included between the open-loop flow rate and the leakage rate as an intermediate value of the rate of filling; and

(b) the inactive scenes flow rate at the output of the coder is equal to a calculation of a continuous decreasing function g having a variable responsive to the rate of filling the leaky bucket which takes;

(i) a value equals to the leakage rate of said leaky bucket as a zero value of the rate of filling,

(ii) a value equals to the open-loop flow rate as a value of the rate of filling the bucket equal to the maximum rate of filling said bucket, and

(iii) a value included between the leakage rate and the open-loop flow rate as an intermediate value of the rate of filling.

5. The process according to claim 4 wherein said calculation includes a determination of the quantization parameter by calculating of the following functions:

(a) for active scenes;

space="preserve" listing-type="equation">Q(i+1)=Qo×

Go(i)/f; and

(b) for inactive scenes;

space="preserve" listing-type="equation">Q(i+1)=Qo×

Go(i)/g

where Go(i) is the open-loop flow rate, and said (i) is a group of images.

6. The process of claim 4 wherein said calculation further comprises a derivation of the function f as a value decreasing responsive to the rate of the filling of the bucket and the function g as a value decreasing responsive to the rate of the filling of the bucket.

7. The process of claim 5 wherein said calculation further comprises a derivation of the function f as a value which is decreasing responsive to the rate of filling the bucket and the function g is a value which is decreasing responsive to the rate of filling the bucket.

8. The process according to claim 6, wherein the function f is:

space="preserve" listing-type="equation">f=G.sub.o (i).(1-ε

.sub.1 (x))+λ

.ε

.sub.1 (x)); and

the function g is;

space="preserve" listing-type="equation">g=λ

.(1-ε

.sub.2 (x))+G.sub.o (i).ε

.sub.2 (x))

wherein ε

₁ (x), and ε

₂ (x), are arbitrary functions, and x defines the filling of the bucket in terms of information relating to the number of bits present in the bucket and the size of the bucket.

9. The process according to claim 7, wherein the function f is:

space="preserve" listing-type="equation">f=G.sub.o (i).(1-ε

.sub.1 (x))+λ

.ε

.sub.1 (x)); and

the function g is;

space="preserve" listing-type="equation">g=λ

.(1-ε

.sub.2 (x))+G.sub.o (i).ε

.sub.2 (x))

wherein ε

₁ (x), and ε

₂ (x), are arbitrary functions, and x defines the filling of the bucket in terms of information relating to the number of bits present in the bucket and the size of the bucket.