GB2126450A - Time compression of sampled signals - Google Patents

Abstract

The illustrated apparatus is for time compressing a television picture in the horizontal direction. Input samples are fed into an interpolator 5 which produces lower frequency output samples, each formed from a plurality of input samples multiplied by coefficients of an interpolation window function. The low frequency samples are fed to a line store 6 and read out under control of a read address generator at the original samples rate to produce the compressed output samples. A write address generator for the interpolator and line store operates at a rate inversely proportional to the compression factor so that the interpolation aperture function is constant relative to the output samples and the interpolator has a low pass filter characteristic which varies according to the compression factor. To take account of the varying number of input samples contributing to each output sample, the input samples may be multiplied by a factor inversely proportional to the compression factor. Compression may be effected in the vertical direction by using a field store, the input samples being successive lines, or in both directions. <IMAGE>

Description

SPECIFICATION Time compression of sampled signals The present invention relates to signal compression, particularly, though not exclusively, of television signals to compress the picture in the horizontal and/or vertical direction.

The invention is defined in the appended claims, to which reference should now be made.

The invention will be further described, by way of example, with reference to the accompanying drawings, in which: Figure 1 shows a number of frequency spectra illustrating some of the considerations involved in the present invention; Figure 2 illustrates graphically the processing and interpolation of input samples; Figure 3 is a block diagram of one embodiment of apparatus according to the invention; Figures 4 to 6 are more detailed block diagrams of parts of the apparatus of Fig. 3; and Figure 7 illustrates a typical interpolation aperture function.

In order to time compress a waveform, a series of samples representing the waveform must be processed to produce a new set containing a smaller number of samples. To avoid aliasing, the new samples must be produced by an interpolator having a low-pass filtering effect which attenuate unwanted frequencies in the signal.

Fig. 1 demonstrates this process in the frequency domain. Fig. 1 a represents the spectrum of a signal sampled at a sampling frequency f,. The interpolation process can be considered as separate processes of low-pass filtering (Fig. 1 b), followed by resampling at a rate fas', Fig. 1 c.

To avoid aliasing, the interpolator response must be resonably low beyond i5'. Time compression then has the effect of stretching the spectrum so that fs' in Fig. 1 C becomes fin Fig. 1 d.

If the resampling process is to retain the maximum amount of information at all compression factors, without causing aliasing, the low-pass filter characteristic must scale according to the compression factor. In the time domain this means that more and more input samples must contribute to each output sample as the compression factor increases; ultimately, when a whole line of television has been compressed to a single sample, this must be made up of contributions from all the input samples.

The filter characteristic should be designed to give good rejection of components at both the old and new sampling frequencies and their harmonics. The effectiveness of the filter, as indicated by the width of its transition band in proportion to the passband and stopband is determined by the width of the aperture. When expressed as a number of low frequency output sample periods Tsl (tis' = 1 /fas'), this is a constant for all compression factors. An interpolation aperture of width nTs' is used, then each input sample contributes to only n output samples.

Fig. 2 illustrates a situation in which input samples i" i2 . ., (row a) are received at a sampling rate f,, sampling period Ts= 1 /fas, and are processed by an interpolator to produce samples at a lower rate s' (period Tis'). The compression factor fs/fs' is assumed to have been set at 2 2/3 and the interpolator is assumed to have a linear interpolation aperture function and an interpolation aperture of 2Ts'. Row b shows that each of the scaled interpolation aperture functions encompasses several input samples, but each input sample i contriputes to only n = 2 output samples.Row b also indicates the coefficient values c for the output samples ol, o2. . (shown in row c) where 01 = cadiz + c12i2 + C3i3 + c14 i4 + c15 i5 + C16i6 2 = C24i4 + c25i5 + C26i6 + C27 i7 + C28 i8 04 = C37i7 + C38i8 + C39i9 + C310i10 + C311ill Because the aperture function is constant relative to the output period Tis', the low-pass cutoff frequency (e.g. T 5') is proportional to the frequency 5' of the lower rate samples.

To effect compression, retiming must take place such that the output samples are output at the same sampling rate as the original input. Although in principle it would be possible to retime the input signal to increase its sample rate by a factor equal to the compression factor, in practice it is preferable to carry out compression after interpolation. Thus row d of Fig. 2 illustrates the output samples Oi, 2 retimed to the input rate f,.

A block diagram of a suitable apparatus is shown in Fig. 3. The inputs comprise an input 1 for video signals sampled at f,, an input 2 for clock pulses at fs, an input 3 for receiving a digital signal equal to the reciprocal of the compression factor, and start address input 4. The various registers referred to in the detailed description below are fed with clock pulses from the input 2.

An interpolator 5 receives the video samples from the input 1 and produces the interpolated lower rate samples (of Fig. 2; row c) which are stored in a line store 6, from which the samples are read out at fas and forwarded to a compressed signal output 7.

The interpolation and storage process is controlled by a write address generator 8 which is shown in more detail in Fig. 4, and consists of two address/register pairs each of which is based on our U.K. patent no 1455821. The receprocal of the compression factor (i.e. 3/8 in the above example) is repeatedly added, via an adder 10 to the contents of a register 11, at video clock rate. The register 11 contains only the fractional part of the sum, any carry being forwarded to a shift enable output 12. The carry occurs every 2 2/3 clock pulses (in the example) i.e., at f:, the output rate of the interpolator and is also forwarded to the second pair of adder 13 and register 1 4 which, essentially, counts them to produce a write address (output 15) for entering the lower rate samples from the interpolator 5 into the line store 6.

The output of register 11 indicates the phase of the input samples i, i2. . relative to the output samples Oi, 02... and is fed via output 1 6 for use as a coefficient address in the interpolator.

The interpolator 5 is shown in Fig. 5 and has. in general, n stages as discussed above. Four stages A-D are shown.

The input 1 is fed via a common multiplier 1 8 to a multiplier 20 of each stage whose other input is supplied from a respective read-only-memory 21 addressed by the coefficient address from output 1 6 of the write address generator.

Each stage has a recirculating arrangement with an adder 22 and register 23 arranged repeatedly to add weighted samples from the multiplier 20 into the register 23; a changeover switch 24 is operable, however, to break the loop and connect one input of the adder 22 to the register 23 of the preceeding stage instead of its own. The first stage A (is supplied with zero.

The switches 24 are commonly controlled by the shift enable line 1 5 and when actuated, cause the summation process to be shifted along to the next stage. The output of last register, D, forms the output of the interpolator..

The aperture function illustrated in Fig. 2 requires only two stages, and the operation of the interpolator will therefore be described initially assuing only two stages i.e. the output taken from point 'x'. Assuming initially that the coefficient address is zero, the first sample o, is generated as follows. In stage A, register 20 contains zero, switch 24 in its recirculating position. The address zero causes readout from the r.o.m. 21 of coefficient cl, which is multiplied by input sample i, in multiplier 20 and added in adder 22 to register 23. The coefficient address advances to 3/8, so that coefficient c21 is addressed in r.o.m. 21, and C12 i2 is added to the register. The next term is added similarly.The third addition of the increment 3/8 in the write address generator causes a carry, producing a shift enable signal at 1 5 changing the positions of the switches 24 in both stages A and B of the interpolator.

The next clock pulse then causes (c"i1 + c,2i2 + C13i3) from stage A being transferred to stage B and loaded with the addition of C14i4 into the register 23 of that stage. Addition of the remaining terms occurs in stage B, with C14, C15 and c16 being supplied by the r.o.m. 21 of stage B. Meanwhile, the first three terms of output sample 2 are being accumulated in stage A, and so the process continues.

It will be noted that output sample o1 is not available at the output of the interpolator until i6 has occurred at the input. The write address generator has an input for loading the registers with a start address to determine the position which the compressed signal will occupy in the line store (and hence in the output signal) and thus can be adjusted to take account of the line shift in the interpolator.

In practice, a linear aperture function is not ideal: a more practical function, having an aperture of 4 Ts (i.e. n= 4) is illustrated in Fig. 6, and hence requires the form interpolator stages shown in Fig. 5. As Fig. 6 shows, coefficients for the four sections of the function are provided by the form read-only memories 21 of the four stages A,B,C,D respectively.

The read address generator 30 (Fig. 3) for readout from the line store may simply be a counter incremented at clock rate to read out the stored samples in sequence.

As mentioned, a multiplier is included in the input to the interpolator. As the compression factor increases, the number of samples contributing to each output sample increases proportionately, so, in order to prevent a corresponding increase in output amplitude, the input is multiplied by the reciprocal of the compression factor in the multiplier. Of course, the correction could be applied at the output, but the arrangement shown provides a smoother result in cases where movement of the compressed area of picture is effected-as may be achieved, for example, by progressively changing the start address supplied to the write address generator.

The line store 6 is shown in Fig. 6 and requires the input samples to be assembled separately from the samples to be read out. So two separate stores 30, 31 are used on alternate lines, being switched at half line frequency by address switches 32, 33 and output switch 34.

The arrangement described provides for accurate interpolation at any compression ratio, with the degree of aliasing determined by the shape of the interpolation aperture. The interpolator has a convenient modular form and is relatively economical in components.

Although described in relation to horizontal compression, the method can also be applied to compression in the vertical direction (or, of course. both directions). The aDDaratus described would then have to be modified by substituting line stores for the registers in the interpolator and using a field store to store the compressed picture.

Colour signals can of course be handled by providing similar arrangements for the components of the signal.

Claims (8)

CLAIMS 1. A signal compression apparatus comprising an input for receiving a sampled input signal and an interpolator arranged in operation in response to signals supplied to the apparatus to indicate the compression factor to produce from a plurality of successive input samples a correspondingly smaller number of output samples, the aperture function of the interpolator relative to the output samples being constant whereby it has a low-pass filter characteristic which varies according to the compression factor. 2. An apparatus according to claim 1, in which the interpolator is followed by retiming means to store the interpolated samples and output them at the same sampling rate as that of the input signal. 3. An apparatus according to claim 1 or 2, in which the interpolator comprises: a plurality of cascaded stages, each having means for generating weighting coefficients, a multiplier to multiply successive input samples by the coefficients, and an accumulator to add the successive products; and control means with a counter incremented successively by a value equal to the reciprocal of the compression factor, the non-integral part of the count providing a control signal for coefficient generation, indicating the relative phase of the input and output samples, and the integral part serving when incremented, to effect a shift of the accumulator contents from one stage of the interpolator to the next succeeding stage. 4. An apparatus according to any one of the preceding claims for altering the apparent dimension of a television picture or part thereof by compressing the television signal. 5. An apparatus according to claim 4, for compression in the horizontal direction, in which the input samples are successive samples of a television line and the retiming means is a line store. 6. An apparatus according to claim 4, for compression in the vertical direction, in which the input samples are successive lines, the accumulators of the interpolator are line stores, and the retiming means is a field store. 7. A signal compression apparatus substantially as herein described with reference to the accompanying drawings. CLAIMS (31 Aug 1983)

1. Apparatus for variably compressing a signal, comprising an input for receiving a sampled input signal and an interpolator arranged to respond to a control signal supplied to the apparatus to indicate the compression factor to produce from a plurality of successive input samples a correspondingly smaller number of output samples, the interpolator having an aperture function which is so controlled by the control signal as to be constant relative to the output samples, whereby the interpolator has a low-pass filter characteristic which varies according to the compression factor.

2. Apparatus according to claim 1, in which the interpolator is followed by retiming means arranged to store the interpolated samples and output them at the same sampling rate as that of the input signal.

3. Apparatus according to claim 1 or 2, in which the interpolator comprises a plurality of cascaded stages, each having means for generating weighting coefficients, a multiplier to multiply successive input samples by the coefficients, and an accumulator to add the successive products; and control means with a counter incremented successively by a value equal to the reciprocal of the compression factor, the non-integral part of the count providing a control signal for coefficient generation, indicating the relative phase of the input and output samples, and the integral part serving when incremented, to effect a shift of the accumulator contents from one stage of the interpolator to the next succeeding stage.

4. Apparatus according to claim 1, 2 or 3, comprising means for scaling the magnitude of the input samples in accordance with the reciprocal of the compression factor.

5. Apparatus according to any of the preceding claims adapted for altering the apparent dimension of a television picture or part thereof by compressing the television signal.

6. Apparatus according to claims 2 and 3, for compression in the horizontal direction, in which the input samples are successive samples of a television line and the retiming means is a line store.

7. Apparatus according to claims 2 and 5, for compression in the vertical direction, in which the input samples are successive lines, the accumulators of the inerpolator are line stores, and the retiming means is a field store.

8. Signal compression apparatus substantially as herein described with reference to the accompanying drawings.