Motion video signal processing for recording or reproducing – Local trick play processing – With randomly accessible medium
Reexamination Certificate
1998-08-18
2001-03-06
Nguyen, Huy (Department: 2615)
Motion video signal processing for recording or reproducing
Local trick play processing
With randomly accessible medium
C386S349000, C386S349000
Reexamination Certificate
active
06198878
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to a method and to an apparatus for encoding and decoding digital video data for and from, respectively, storage media having a given data storage capacity.
BACKGROUND OF THE INVENTION
ISO/IEC 13818 “Information Technology—Generic coding of moving pictures and associated audio information: Video” (MPEG2) discloses an International standard concerning coded digital video and audio data streams. Due to varying picture content the generated amount of code varies over time. However, in case of fixed channel capacity normally a constant bit rate (CBR) MPEG encoder control is desired which can be achieved by using an encoder buffer at transmission side and a decoder buffer at receiver side which is explained in more detail in annexes C and D.4 of said International standard. The other mode is variable bit rate (VBR) coding.
SUMMARY OF THE INVENTION
If there is any digital storage medium like DVD (digital video disc) which is capable of variable bit rate operation it might be possible to improve the coding quality by using variable bit rate coding instead of constant bit rate coding.
It is one object of the invention to disclose a method of variable bit rate encoding for a fixed-capacity storage medium in which the coding quality is basically improved compared to that of constant bit rate encoding.
It is a further object of the invention to disclose a method of variable bit rate decoding from a fixed-capacity storage medium in which the decoding quality is basically improved compared to that of constant bit rate decoding.
It is a further object of the invention to disclose an encoding and a decoding apparatus which utilises the inventive encoding and decoding method, respectively.
It is a further object of the invention to disclose a digital video signal which is encoded or decoded according to the inventive method.
“Constant bit rate” shall mean a constant number of bits output to the channel (or e.g. the Systems multiplexer) in each field period (2 Oms in a 50 Hz TV signal), while “variable bit rate” shall cover all other bit delivery schedules. In a CBR environment the kernel MPEG encoder circuit is followed by a buffer of a certain size. This buffer is used to smooth the inherently variable bit rate output of the kernel MPEG encoder. The video buffering verifier VBV is a concept used in encoder control that gives an upper bound for the usable encoder buffer size as well as the minimum buffer size needed by a decoder to decode all bit streams compliant to this profiles @ levels (P@L). For all defined profiles @ levels maximum VBV_buffer_size values are specified in clause 8 of ISO/IEC 13818-2. Smaller values are also allowed and can be signalled in each bitstream by the VBV_buffer_size element. In a variable bit rate application there is no urgent need to use the encoder buffer to smooth the bit rate, however, it might still be useful to retain the buffer, as will become evident below. So the change is that the channel bit rate is variable now.
In both, CBR and VBR applications, the MPEG encoder and the MPEG decoder operate by using the well-known hybrid DPCM/DCT structures.
The bit rate control goals for applications using storage devices with the capability to process a variable bit rate data stream differ from the CBR case. It can now be desirable to:
maintain uniform quality over time;
maximise storage time on a medium of a given capacity.
Therefore the encoder control strategy needs to be adapted to this situation.
To ease understanding the encoder control for the CBR case shall be reviewed first.
The conventional encoder control strategy consists of three steps or control levels:
1) bit allocation (global control):
The allocation of bits to the different coded frame types (I, P, B frames) is determined according to their relative complexity X
i
, X
p
, X
b
. The goal is to equalise the subjective quality of the frame types. In the following feedback implementation, which does not carry out any pre-analysis, it is assumed that video scenes are sufficiently stationary for some time, so that information from the past frames can be used for the bit allocation to the coming frame or field, depending on the coding structure. In the remainder of the specification, “frame” may generally denote a frame or a field or any other section of a GOP which is greater than a unit (to be defined below).
The complexity can be estimated by any criterion reflecting the relative coding cost for I, P, and B frames at a given image quality. However, it should be considered whether the necessary measurements can be obtained without too much burden on the encoder complexity.
First the “image quality” should be measured. However, it is well known that no commonly agreed, sufficiently simple, objective measure exists. Therefore, usually the signal-to-noise ratio (SNR) of the reconstructed frame is taken as an approximation to image quality. In the context of an MPEG encoder control even this measure is somewhat costly as it involves computation of SNR for each encoded frame. Therefore the average quantiser step size (Q
i
, Q
p
, Q
b
) used to code a frame is taken as a quality indicator instead.
Now, having defined a measure for “quality”, the “complexity” of a frame can be expressed as the product of bits generated and the quantiser step size used. In other words: a frame is complex to code if it produces many bits even if the quantiser step size is coarse. This product is formed for each unit m (e.g. macroblock or block or slice or any section which is smaller than a frame) and then summed over the frame:
x
m
=b
m
*q
m
(1a)
X
c
=&Egr;x
m
, (1b)
wherein b
m
is the number of bits for coding unit m, q
m
is the average quantiser step size used in unit m, and c=i, p or b is the coding type.
At the beginning of each Group Of Pictures GOP the bit budget B
G
(g) available for the current GOP g, considering also a carry B
G
(g−1) from the previous GOP, is
B
F
(
n
32 0)
=B
G
(
g
)=(bit rate)*(frames per GOP)/(frame rate)−
B
G
(
g−
1) (2)
For each frame n in the current GOP, a target number of bits T
c
, depending on its coding type c, is determined according to the computed complexity and the remaining number of frames of each coding type (N
i
, N
p
, N
b
) within the current GOP.
Considering the remaining bits B
F
(n) after coding frame n−1 of the GOP, the following formula can be used:
T
c
=B
F
(
n
)*(
X
c
*K
c
)/(
N
i
*X
i
*K
i
+N
p
*X
p
*K
p
+N
b
*X
b
*K
b
) (3)
The factors K
c
are used to reflect different visual importance of frame coding types, typically K
i
=K
p
=1 and K
b
<1.
The complexity computation according to equations (1a) and (1b) is normally updated after each frame, using data from the most recent I, P and B frames, to ensure a reasonable response to scene changes. The bit budget
B
F
(
n+
1)=
B
F
(
n
)−&Egr;
b
m
(4)
for the remaining frames of the GOP, after subtracting the number of bits &Egr;b
m
used for the current frame, is re-distributed using formula (3) with N
i
, N
p
, N
b
set to the remaining number of frames of each type.
After coding the last frame N of the GOP, the remaining bits B
G
(g)=B
F
(N) are carried to the next GOP g+1.
It is also possible to explicitly allow a non-constant amount of bits per GOP, keeping only the long-time average bit rate constant. Practically this can be implemented by re-filling the bit budget B
F
(n) after each frame using
B
F
(
n
)=
B
F
(
n−
1)−&Egr;
b
m
+(bit rate)/(frame rate) (5)
instead of re-filling the budget only at the start of a GOP using equation (2). This budget will then always be distributed according to the number of I, P and B frames in a GOP, rather than only on the remaining frames in the current GOP. The effect of this change in case of a scene change in the middle of a GOP is that a scene change P frame might obtain more bits which impr
Blawat Meinolf
Herpel Carsten
Deutsche Thomson-Brandt GmbH
Herrman Eric. P.
Nguyen Huy
Tripoli Joseph S.
LandOfFree
Method and apparatus for encoding and decoding digital video... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for encoding and decoding digital video..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for encoding and decoding digital video... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2538935