Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
1998-11-30
2002-05-21
Diep, Nhon T (Department: 2713)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
C382S250000
Reexamination Certificate
active
06393060
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to video signal coding and decoding, and more particularly an apparatus and method for coding and decoding low transfer rate video images.
2. Description of the Related Art
The international standards for a conventional video coding include the Joint Photographic Coding Experts Group (JPEG) for still picture coding/decoding, the Moving Picture Experts Group (MPEG) for motion picture coding/decoding, and the H.261 or H.263 for low transfer rate video coding/decoding. In response to an increasing demand for video communication over the existing Public Switched Telephone Network (PSTN), extensive studies have been made on the low bit video coding.
For example, the H.263 recommends the International Telecommunication Union-Telecommunication Standardization Sector (ITU-T, previously known as CCITT), which employs a motion compensated hybrid Differential Pulse Code Modulation/Discrete Cosine Transform (DPCM/DCT) coding method suitable to very low bit video transmission for realization of video telephone system. Particularly, the video coding method involves a DCT transformation and quantization of an input digital video signal; a restoration of the quantized video signal to detect a difference between the quantized video signal and the original video signal thereby estimating a motion; and a control of the quantizing step to attain a desired bit rate.
FIG. 1
is a block diagram showing a conventional motion picture coding apparatus based on DCT. The video coding is generally divided into an Intra (I) frame coding and a Predictive (P) frame coding. If an input video bit stream is an I frame, it is directly output to a DCT unit
101
without change. For a P frame, a difference between a motion-estimated data and the current input bit stream is output to the DCT unit
101
through the subtractor
110
.
The DCT unit
101
eliminates the correlation of data by a dimensional coordinate transform. The DCT unit
101
decomposes the input frame in block units for the coordinate transform such that a portion of each picture block is transformed from a spatial domain to a frequency domain. Thus, the DCT transformed data are inclined to be driven in one direction towards the lower frequency band and are quantized at the quantizing (Q) unit
102
. The quantization parameters such as a weight matrix and a quantization scale code are used for quantizing wherein the weight matrix indicates the weight of each DCT coefficient and the quantization scale code determines the quantizing step.
After quantization, each coefficient is output to an entropy coding unit
103
, which performs a Variable Length Coding (VLC). Through the VLC, a frequently occurring value is represented by a smaller number of bits and an occasionally occurring value is represented by a larger number of bits, thereby reducing the entire number of bits to be transmitted to channel
104
. The quantized data is also subjected to dequantization at a dequantizing unit
105
before an Inverse Discrete Cosine Transform (IDCT) at an IDCT unit
106
. An adder
107
sums a motion-estimated data from a motion prediction unit
109
and the IDCT data, storing the summed value in a frame memory
108
.
A continuous picture in a time axis usually includes a motion of a human or object in the center of the image. Based upon this idea, the motion prediction unit
109
eliminates redundancy by replacing an unchanged or similarly moving portion of the picture with the related portion of the previous picture. Thus, the amount of data to transmit is reduced to a large extent.
When the adder
107
stores the summed value in the frame memory
108
, the data stored in the frame memory
108
forms the previous picture while the motion prediction unit
109
estimates the motion of a currently input picture. Motion estimation is performed by searching for the most similar blocks between the previous and current pictures, and a motion vector (MV) represents a degree of motion. The motion vectors are transmitted as well as the information concerning the VLC transform coefficients via the channel
104
. The motion vectors also undergo VLC at the entropy coding unit
103
in order to attain a maximize the coding efficiency.
The MVs must first be obtained in order for the motion prediction unit
109
to perform a motion estimation. Up to four MVs are produced per one macro block, but only a difference between the current and previous MV is subjected to VLC for transmission because four MVs have a large number of bits to transmit. In the motion estimation, the motion prediction unit
109
uses forward and backward predicted blocks as well as involving two types of motion compensated frame.
The P frame is an estimated motion through the forward prediction and is used to predict the next P frame. The P frame is also usable for the forward and backward predictions of a bi-directionally (B) predicted frame. The B frame itself is not usable for prediction of other frames. On the other hand, an I frame is an image used as a criterion for performing an image compressive coding. Thus the original signals of an I frame is input to DCT transformation and quantization steps, thereby eliminating redundancy in spatial direction only.
The first frame is generally for I-frame coding, and when a transmission packet loss occurs, a transmitter may send an I frame at any other time by the request of a receiver. The I frame is used in the motion estimation for the P and B frames. Accordingly, the I frame coding also determines the coding efficiency for the following P and B frames. Especially, the background portion of the P frame image is not coded until the I-frame coding step has terminated and a scene change has occurred. As a result, the I-frame coding has an influence on the image quality. Thus, a better result of I-frame coding provides better coding of the subsequent P frame.
For the P frame coding, two coding steps are used with the input frames, namely a motion compensated prediction coding using a highly visible correlation between adjacent frames, and a displaced frame difference (DFD) coding involving an estimation error after performing a motion compensation. The DFD is the output of the subtractor
110
, i.e. the signal difference between the current frame and the previous frame, the difference being as much as a motion vector. The DFD coding is the bulk of the P frame bits.
In most standardized coding, the DFD is coded in a same manner as the I frame video coding, which does not utilize the characteristic difference between a natural image and a DFD image. The difference arises because the DFD image has less spatial correlation, having much more mid and high frequency components compared to the natural image, including primarily smoothing regions. Thus, the DFD image provides a lower energy compression efficiency than the natural image, resulting in a deterioration of the overall efficiency in the run-length coding method using the existing zigzag scanning.
Furthermore, the number of DCT coefficients to be coded in low bit rate transmissions is too small. Yet, the coefficients are expressed by high quantizing levels, resulting in block and ringing effect.
Another issue of the motion picture coding is a bit rate control at the bit rate control unit
111
. Because bits are generated by simply fixed quantizing parameters and coded by coefficients, a repetitive coding method is required to adjust to a specific target bit rate. Even though control is possible, an accurate bit rate control is hard to achieve.
The I-frame coding involves a structure for bit rate control by simply controlling the quantization step size (quantization parameter×2). Thus, the DCT error increases due to a large quantization distance in low bit rate transmissions, resulting in a continuous effect on the coding including motion estimation and compensation of P frames. As a result, the entire coding performance is deteriorated.
Moreover, the bits required for the I-frame coding when transmitting at a low bit ra
LandOfFree
Video coding and decoding method and its apparatus does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Video coding and decoding method and its apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Video coding and decoding method and its apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2882369