Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
2000-08-03
2002-05-07
Lee, Richard (Department: 2613)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
Reexamination Certificate
active
06385247
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates to an encoding circuit for transform coding of a picture signal. The invention also relates to a decoding circuit for decoding an encoded signal supplied by the encoding circuit.
An encoding and a decoding circuit of this type may form part of a television broadcasting system, in which case the encoding system forms part of a television transmitter and each television receiver is provided with a decoding circuit. The encoding and decoding circuits may also form part of a video recorder.
2. Description of the Related Art
As is generally knows, a television picture may be assumed to be a two-dimensional array of pixels. In a 625-line television system, the picture comprises 576 visible picture lines, and each picture line comprises 720 visible pixels. The television picture thus comprises 576*720 pixels. If the luminance of each pixel is represented by, for example, 8 bits, the transmission of 25 pictures per second requires a bit-rate of approximately 83 Mbit/sec for the luminance information only. This is found to be inadmissibly high in practice.
By subjecting each picture to a two-dimensional transform, the number of bits per picture, and hence, the bit-rate can be limited considerably. To perform such a transform, the picture is partitioned into sub-pictures of N*N pixels each, for example, into 72*90=6480 sub-pictures of 8*8 pixels each. Each sub-picture is subsequently converted into a coefficient block of N*N coefficients by two-dimensional transform. The transform is intended to obtain a block of coefficients which are mutually uncorrelated. Among the known transform methods, the discrete cosine transform (DCT) is generally considered to be the best alternative.
The following is a representation providing insight into the two-dimensional transform. Associated with the chosen transform is a collection of N
2
mutually orthogonal basic pictures B(i,k) with i,k=0, 1, 2, . . . N, each comprising N*N pixels. Of these basic pictures, B(0,0) has a uniform luminance. As the index k increases, the basic picture B(i,k) has higher spatial frequencies in the horizontal direction, hence more detail. As the index i increases, the basic picture has higher spatial frequencies in the vertical direction. In the two-dimensional transform, each sub-picture is considered as the weighted sum of said basic pictures B(i,k) each with its own weighting factor y(i,k) i,k=0, 1, 2, . . . N. The weighting factors y(i,k) correspond to the previously-mentioned coefficients. It is these coefficients which are transmitted instead of the original pixels.
A reduction of the number of bits to be transmitted per picture is now achieved by transmitting only those coefficients which have a significant value. For example, the coefficient y(0,0), being the weighting factor of the basic picture B(0,0) and thus a measure of the average luminance of the sub-picture, is always transmitted. This coefficient y(0,0) is also referred to as “dc coefficient”. The other coefficients, referred to as “ac coefficients”, are only transmitted when their absolute value is larger than a predetermined threshold value. This is referred to as threshold coding. The coefficients may also be subjected to a coarser quantization as the corresponding basic picture comprises more details. This is because the human eye cannot observe fine details very well. The latter is also referred to as frequency-dependent quantization. In practice, frequency-dependent quantization and threshold coding are often combined. Then, only those coefficients are transmitted which still have a value which is unequal to zero after quantization.
The transmission of only those coefficients having a value which is unequal to zero implies that the address of the location of these coefficients in the two-dimensional coefficient block should also be transmitted. In practice, the coefficient block is read in a predetermined sequence for this purpose so that, for each coefficient block, a series of coefficients is produced in which said address is represented by a scanning sequential number. Reference 1 describes a method of scanning the coefficient block in accordance with a zigzag pattern, starting with the do coefficient y(0,0). Generally, the largest part of the signal energy in a sub-picture is concentrated in the low spatial frequencies. The significant coefficients are therefore often the coefficients y(i,k) with a small value for i and k. In the known zigzag scanning method, the significant coefficients acquire low scanning sequential numbers, the zero value coefficients are clustered for the greater part, and acquire high scanning sequential numbers. Such a series of coefficients can be transmitted efficiently.
However, the zigzag scanning pattern is not efficient for moving pictures. In fact, when there is motion within a Sub-picture, the value of the coefficients representing high spatial frequencies in the vertical direction increases drastically. These are the coefficients y(i,k) having a large value for i. The vertical scanning shown in FIG. 1
c
of Reference 1 has now proven to be more efficient.
SUMMARY OF THE INVENTION
It is an object of the invention to provide an encoding circuit with which an even further bit rate reduction is achieved.
According to the invention, the scanning pattern provides at least one jump to a predetermined coefficient which is not contiguous to a previously read coefficient.
The invention is based on the recognition that a block of coefficients is transmitted most efficiently if the coefficients are less significant as their scanning sequential nether is higher. It is achieved, with the encoding circuit according to the invention, that the predetermined coefficient, for example, y(7,0) which is representative of the motion in the sub-picture, is read earlier than the coefficients contiguous to this coefficient and is thus transmitted earlier than is the case with the known scanning patterns. As a result, the Coefficients are transmitted as much as possible in the order of their significance. More particularly, fewer and longer sub-series of zero value coefficients are then produced, which contributes considerably to the coding efficiency. As is indicated in Reference 2, one compact run-length code is assigned to sub-series of zero value coefficients. These sub-series further occur with greater probability at the end of the scanning process. In that case, it is even unnecessary to transmit said run-length code, but a code indicating the end of the block is then sufficient.
A favorable embodiment of the encoding circuit in which the sub-picture comprises two interlaced sub-fields, is characterized in that the transform circuit is adapted to partition each coefficient block into sub-blocks by separately transforming the sub-fields, and in that the predetermined coefficient forms part of a different sub-block than the previously read coefficients. It is thereby achieved that only one series of coefficients is obtained for each sub-picture, and that in said series, the lowest scanning sequential numbers are assigned to the most significant coefficients. It is to be noted that it is known per se from Reference 3, that two interlaced sub-fields are separately transformed into sub-blocks in the case of motion. However, in this known encoding circuit, each sub-block is transmitted separately. Since the total number of sub-blocks per picture is dependent on the amount of motion in the picture, and is thus variable, the number of transmitted coefficient series is also motion-dependent. This is found to have detrimental effects, in practice. Inter alia, each series of coefficients is terminated by a multi-bit “end-of-block” code, which, in the case of much motion, hence many series, is at the expense of the coding efficiency. In the encoding circuit according to the invention, the total number of series per picture remains constant.
REFERENCES
1. Variable control method for dcT coding using motion-compensated prediction. Lecture at the Gener
Borgers Stephanus M. C.
De With Peter H. N.
Goodman Edward W.
Lee Richard
US Philips Corporation
LandOfFree
Encoding circuit for transform coding of a picture signal... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Encoding circuit for transform coding of a picture signal..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Encoding circuit for transform coding of a picture signal... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2899149