Facsimile and static presentation processing – Facsimile – Specific signal processing circuitry
Patent
1988-09-02
1990-05-29
Britton, Howard W.
Facsimile and static presentation processing
Facsimile
Specific signal processing circuitry
358105, 358903, G06K 962, H04N 7137
Patent
active
049300130
DESCRIPTION:
BRIEF SUMMARY
The present invention relates to pattern processing and is particularly, though not exclusively, concerned with the application of such processing to the coding of video signals.
One example is in video coders employing inter-frame conditional replenishment coding for the transmission of moving pictures. The difference is taken between picture element values and predicted values (ie those from the preceding frame). The differences are applied to a threshold circuit which for each difference value produces a `1` or a `0` indicating whether or not a threshold value is exceeded, thereby forming a "movement matrix". Only those element values (or, if preferred, element differences) having a `1` in the coresponding element of the movement matrix are coded for transmission.
At the receiver, the received information is used to update an image stored in a local frame store. In order that the receiver can correctly decode the received information side information indicating which coefficients have been sent must be included in the transmission. Sending all bits of the movement matrix represents a substantial overhead, and it has therefore been proposed to divide the picture into 8.times.8 blocks and "vector quantise" each 8.times.8 movement matrix be matching it to the nearest one of a set of (eg) 32 patterns which can be identified to the receiver using only five transmitted bits. Hitherto this matching has been achieved using a correlation method with the patterns; this is computationally extravagant and even with high speed processing 32 patterns is the maximum normally practicable in real time.
The technique is also applicable to so-called hybrid video coders where a combination of transform coding and inter-frame conditional replenishment coding is employed. Transform coding involves applying a two dimensional transform such as the Hadamard or discrete cosine (DCT) transform to (eg 8.times.8) blocks of picture elements to produce corrsponding matrix of transform coefficients (transformation of a whole frame at once is too slow for real time processing). Here the movement matrix is based on the differences between the coefficients of the matrix of the current frame and that from the preceding frame.
In a variant of this procedure, the taking of inter-frame differences may precede the transformation rather than follow it.
According to one aspect of the present invention there is provided a method of classifying a set of values representing a two dimensional pattern comprising forming a weighted sum of those values, modulo x, where x is an integer less than 2.sup.p, p being the number of values in the set, and employing the sum as an address to access a store having x locations each containing a class identification word.
In principle the values referred to could be the value associated with each element of the pattern, but the summation may be simplified and the derivation of the weighting coefficients and modulus facilitated by first forming a set of values each of which represents a number of elements. Thus a preferred embodiment of the invention includes deriving the set of values by identifying, for each of a plurality of groups of elements of the pattern, that one of a set of predefined such groups, each of which has a code associated with it, which that group most closely resembles according to a predetermined criterion, the codes associated with the groups thus identified forming the set of values.
The groups may be the rows (or columns) of a rectangular pattern matrix, though other selections are possible.
One embodiment of the present invention will now be described, by way of example, with reference to the accompanying drawings, in which:
FIG. 1 is a block diagram of an apparatus for vector quantisation of movement matrices;
FIG. 2 illustrates a set of standard line vectors;
FIG. 3 illustrates a set of standard patterns;
FIG. 4 is a diagram illustrating the concept of mapping of pattern domains;
FIGS. 5 and 6 are flowcharts for coefficient and mudulus derivation; and
FIG. 7 is a diagram illustrating mot
REFERENCES:
patent: 4558350 (1985-12-01), Murakami
patent: 4779131 (1988-10-01), Matsumoto
IEEE Transactions on Communications, vol. COM-24, No. 4, Apr. 1986, B. G. Haskell: "Differential Addressing of Clusters of Changed Picture Elements for Interframe Coding of Videotelephone Signals", pp. 140-144.
IEEE Transactions of Communications, vol. 34, No. 4, Apr. 1986, IEEE, (New York, US), M. J. Bage: "Interframe Predictive Coding of Images Using Hybrid Vector Quantization", pp. 411-415.
British Telecommunications public limited company
Britton Howard W.
LandOfFree
Pattern processing does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Pattern processing, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Pattern processing will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-524887