GB2280079A - Chroma signal motion detector - Google Patents

Description

2280079 VIDEO MOTION DETECTOR This invention relates to a video motion

detector, for producing a motion output signal indicating whether motion exists in a colour video signal. The motion detector is suitable for use in a motion-adaptive video processing system.

The invention has been designed for use in conjunction with the proposed PALplus wide-screen television system, described in a paper by Ebner, A., et al., 11PALplus - The European System for Wide-Screen Terrestrial M', International Broadcasting Convention, Amsterdam, 3-7 July 1992, IEE Conference Publication 358, pages 203-207. In that system a 16:9 aspect ratio picture is transmitted so that it can be displayed on an enhanced 16:9 aspect ratio receiver, but can compatibly be displayed by a conventional 4:3 aspect ratio receiver. This is achieved by transmitting the 16:9 aspect ratio signal on only three-quarters of the available lines. These are chosen to be the middle three-quarters of the image, termed the letterbox, leaving black bands at the top and bottom of the picture on a conventional receiver. In these black bands helper signals are transmitted at low amplitude which are used in the enhanced receiver to improve the quality of the display as derived from the letterbox lines. They are invisible on a conventional receiver.

There is complex processing at the transmitter to generate the signals for transmission and at the enhanced receiver to derive signals for display from the transmitted signal. It is proposed in the above paper to apply different processing, dependent upon whether the video signal was originally derived from a camera or from film. If derived from a camera, the two 20ms fields of each 40ms frame (or picture) truly represent the situation at times which are 20ms apart. This is not the case when the video signal has been derived from film, because the two interlaced fields are then derived from a single film frame. Thus they represent information at the same instant in time. Improved display can be obtained by recognising this fact and using different filter functions at both the transmitter and at each enhanced receiver in camera mode and film mode respectively.

It will be appreciated that an improved display can likewise be obtained if a motion detector is used when the equipment is operating in camera mode, so as to detect areas where there is no motion in the image between adjacent fields. These areas can then receive processing analogous to that used in film mode, making use of the fact that the pair of fields correspond to the same image information. This requires a motion detector at the transmitter and a corresponding motion detector at each enhanced receiver.

More specifically, it is proposed in the paper referred to above to use a method of coding based on that described by Kays, R., "Ein Verfahren zur verbesserten PAL-Codierung und -Decodierung". Fernseh- und Kino-Technik, Vol. 44, No. 11, November 1990, pages 595-602, referred to as the ColorPlus system. This works on pairs of lines which are 312 video line periods apart, that is with one line from each field of the frame. The luminance component above about 3MHz is made to be the same on the two lines of a pair, e.g. by averaging them, and similarly the chrominance component is made to be the same on the two lines of a pair. In the decoder, averaging and differencing the lines of each pair yields cleanly- separated luminance and chrominance.

The method proposed for PALplus is an adaptive variant of Color-Plus, which switches between the Color-Plus mode and a fall-back mode under the control of a motion detector. Synchronisation of the switching in the encoder and decoder is achieved by using the same design of motion detector circuit in transmitter (encoder) and enhanced receiver (decoder). Because of the averaging of pairs of lines which are 312 video line periods apart, just discussed, the vertical-temporal bandwidth of the chrominance and of the luminance above 3MHz is limited. Thus in areas of moving colour it is proposed that no luminance above 3MHz should be transmitted, and instead of the averaging of pairs of lines which are 312 video line periods apart, only the unmodified modulated chrominance should be transmitted in this part of the spectrum.

However, we have found that if the motion detector fails to operate properly, and finds movement where none in fact exists, high-frequency luminance detail is unnecessarily softened.

The present invention is defined in claim 1 below, to which reference should now be made. Advantageous features of the invention are set forth in the subclaims.

In a preferred video motion detector, described in more detail below, input chrominance component signals on the current frame are compared with those on the preceding frame. As noted above, this can incorrectly indicate chrominance motion when in fact there is no chrominance in the current frame. To overcome this problem, a determination is made as to whether there is chrominance in the current frame. In the absence of chrominance in the current frame, the motion output signal is forced to its value which indicates the absence of motion.

The invention will now be described in more detail, by way of example, with reference to the accompanying drawings, in which:- Figure 1 is a block diagram illustrating a motion detector circuit as previously proposed; Figure 2 is a corresponding block diagram of a motion detector circuit embodying this invention; Ficrure 3 is a block circuit diagram of an improved motion detector circuit based on the circuit of Figure 2; and Figures 4 and 5 are block circuit diagrams of an encoder and a decoder in which the motion detector can be used.

The previously-proposed motion detector circuit 10 illustrated in Figure 1 processes intra-frame averaged chrominance, i.e. the average of each pair of lines which are 312 video line periods apart, to produce a motion signal M. The motion signal M is intended to be low in areas where the chrominance is stationary or is absent, and high where it is moving.

To this end the motion detector circuit 10 receives the baseband U and V chrominance component signals at inputs 12 and 12A respectively. Similar processing is applied to the two channels and only the U channel is described in detail. Corresponding components in the V channel are given the same references with the suffix A.

The U signal at input 12 is applied to a horizontal low-pass filter 14 having a filter function HF1 and thence passes both through a one-line delay 16 and directly to the two inputs respectively of an adder 18. The adder in effect averages the two inputs applied to it, (the necessary amplitude halving being taken into consideration at any convenient point in the circuit). The output of the adder 18 is then applied both through a one-frame (picture) delay 20 and directly to the two inputs respectively of a subtractor 22, so that the signal from the previous frame is subtracted from the signal for the current frame to provide a frame-difference signal. The frame-difference signal is applied to a rectifying and clipping circuit 24. This may be constituted by a first look-up table (LUT), conveniently implemented in read-only memory (ROM). The U signal output of circuit 24 is then supplied together with the corresponding V signal output of circuit 24A to a function circuit 26, which may also be constituted by a look-up table, again implemented in ROM, and which provides the motion-indicating output signal M.

In operation of the motion detector circuit 10 of Figure 1, after horizontal low-pass filtering with a cut-off frequency of about 1 MHz in filter 14, the chrominance is averaged in adder 18 across the line delay 16 to remove the effects of any static phase error in the received signal, as in a normal delay-line PAL decoding operation. Then the frame delay 20 is used to compute the amplitude difference between the signal for the current frame and for the previous frame. A large amplitude difference signal at the output of subtractor 22 indicates the presence of moving chrominance. This difference signal is then rectified and clipped by the look-up table LUT1 in circuit 24.

The motion signal M is derived by the function circuit 26 from the U and V signals by computing the magnitude of the equivalent PAL chrominance signal. Theoretically this is given by an ellipse in U, V space:

14 = [(0.886U/2.02)2 + (0.701V/1.14)2]1.

an In practice a simpler expression may be used, for example one based on a piece-wise linear approximation of the ellipse.

We have found that the circuit of Figure 1 has one significant drawback. It rightly produces a high value of M when there is chrominance in the current frame but none in the previous frame. However, it also produces a high value of M when there was chrominance in the previous frame but not in the current frame. In the overall system this manifests itself as a softening of essentially uncoloured luminance detail as it is uncovered by a moving coloured object.

This problem is overcome by the motion detector circuit 30 embodying the invention shown in Figure 2. This is closely based on the motion detector 10 of Figure 1, and the corresponding components have been given the same reference numerals and will not be described again in detail.

The additional elements in Figure 2 comprises a further rectifying and clipping circuit 32 connected to the output of the adder (averager) 18. The output of circuit 32 is applied together with the output of the corresponding circuit 32A in the V channel to a function circuit 34. The circuits 32 and 34 are again constituted by look-up tables and are similar to the circuits 24 and 26. The output of the function circuit 34 is applied to another look-up table 36 which provides a threshold function. A multiplier 38 receives the output of the main function circuit 26 and modifies it by multiplication with the output of the circuit 36 to provide the motion signal M at the output of the motion detector.

Thus in the circuit of Figure 2 the motion signal from the function circuit 26 is multiplied by the output of the threshold circuit 36, which generates a control signal in dependence upon the amplitude of its input. The output of threshold circuit 36 is zero when there is no chrominance in the current frame, blocking the false motion signal that is produced by the circuit of Figure 1. The output of threshold circuit 36 is unity when there is significant chrominance in the current frame, allowing the correct motion signal to pass through the multiplier 38 to the output.

Intermediate values may be generated at lower levels of chrominance; this may be helpful in avoiding abrupt and visible mode changes.

While a frame delay is used in Figure 2, in principle a delay based on a field delay, possibly with appropriate interpolation, could be used instead.

We have furthermore appreciated that the circuit of Figure 2 now has a further disadvantage. This arises in particular when a coloured object moves against a complementarily-coloured background. The chrominance component at the edge of the object passes through zero, and because of the softening effect of the horizontal low-pass filters 14, 14A, the chrominance at this edge may be small for a significant number of pixels. In this region the circuit of Figure 2 may erroneously produce a low value of M, as there is apparently no chrominance in the current frame. If this happens the chrominance will be transmitted with reduced verticaltemporal bandwidth. This can produce a 25 Hz flicker on horizontallymoving narrow coloured lines. The circuit of Figure 1 does not have this problem as the edge is moving and so there is a significant difference signal.

This further problem is overcome in the motion detector circuit 50 of Figure 3, in which the horizontal filters 14, 14A located prior to the line delays 16, 16A are omitted and instead four separate horizontal lowpass filters 52, 52A; 54, 54A are included. The filter 52 is located between the rectifying and clipping circuit 24 and the function circuit 26, and the filter 54 is located between the rectifying and clipping circuit 32 and the function circuit 34. By placing the horizontal lowpass filters after the rectifying and clipping circuits in this way, the low amplitude chrominance is retained at the edge of the object, producing the correct value of M.

Many of the processing blocks shown in Figure 3 are identical. These are shown joined by vertical dashed lines, and some economy of hardware can be achieved by using single components for the connected blocks and multiplexing the signals applied through them.

6 - The function of the multiplier 38 could be incorporated into the function circuit 26, which would then have three inputs, or alternatively into the look-up table 36, which would then have two inputs.

Figure 4 is a block circuit diagram of the pertinent components of an adaptive Color-Plus encoder 60 incorporating a motion detector 50 of the type shown in Figure 3 (or alternatively Figure 2). The circuit has been simplified by the omission of delays and filters which are not essential to an understanding of the system but the operation of which will be clear to those skilled in the art.

The inputs 62,64,66 receive the Y, U and V components of an interlaced television signal. By means of delays and switches (not shown), pairs of lines from each frame are input to the encoder simultaneously, one line from each field. The phase of this line pairing is important, as ColorPlus relies on the fact that the phase of the colour subcarrier is inverted after 312 line periods. During the first output field, output line n is encoded from input lines n and (n + 312), e.g. output line 24 is encoded from input lines 24 and 336. During the second output field, output line n is encoded from input lines (n - 312) and n, e.g. output line 336 is encoded from input lines 24 and 336.

The behaviour of the encoder is as follows. The luminance input Y is split into frequencies above and below 3 MHz (approximately) by the horizontal filters. 68 having function HF2 and adjacent subtractors 70. The low frequencies pass straight to the PAL output, via a field rate switch 72. This switch is shown in the "field 2" position. The high frequencies are averaged by an adder 74 (with implied division by 2) before being added back to the low frequencies luminance in adders 76. The amplitude of this part of the signal is controlled by the motion detector; when there is moving chrominance, the OY control" signal falls from unity to zero, cutting off the high frequency luminance, due to the action of multiplier 78. The overall effect of this is to ensure that the luminance above 3 MHz in the PAL output is the same on each line of a pair, and absent when there is moving chrominance.

The chrominance signals U and V are averaged and differenced by adders 80 and subtractors 82. The intra-frame average signals are input to the motion detector circuit 50, and also passed through to the PAL output via field rate switches 84 and subcarrier frequency fsc modulators 86. When there is moving chrominance, the %, V control" signal from the motion detector rises from zero to unity, allowing the intra-frame difference chrominance to be added to the output due to the action of multipliers 88. The overall effect is to ensure that the unmodulated chrominance signals are the same on each line of a pair, except when there is moving chrominance; in this case the entire chrominance signal is sent.

The modified intra-frame difference signal and the average signal are added and subtracted in adders 90 and subtractors 92 before being applied to the selector switches 84. The outputs of the modulators 86 are added in an adder 94, the output of which is added to the luminance signal in an adder 96 to provide the PAL output signal.

The corresponding decoder 100 is shown in Figure 5. The PAL input signal at inputs 102 is formed into pairs of lines, one from each field, as for the encoder of Figure 4. Identical band-splitting filters 104 with filter function HF2 are used to yield low-frequency luminance, which is passed to the luminance output via adders 106 and a field rate switch 108. The low frequency components are subtracted from the input signals in subtractors 110 to yield high frequency luminance plus modulated chrominance. This is averaged and differenced by an adder 112 and a subtractor 114.

As the colour subcarrier on each line of an input pair has opposite phase, the intra-frame average signal will have no residual subcarrier when there is no moving chrominance. Therefore this signal is the clean high frequency luminance and is added to the low frequency luminance by the adders 106, after modification by multiplier 116 in dependence upon a uY controln signal from the motion detector.

at..

Similarly the intra-frame difference signal contains modulated chrominance with no residual luminance. Demodulators 118,120,122 and 124 demodulate the outputs of adder 112 and subtractor 114 with reference to the U and V phases respectively. After demodulation, the intra-frame difference signal gives the clean U and V intra-frame average signals that were encoded. These are applied to the motion detector 50, and are also passed to the U and V outputs through adders 126 and subtractors 128, and also field rate switches 130.

When there is chrominance motion, all the PAL signal above 3 MHz is used for chrominance to provide the full vertical-temporal bandwidth. In this circumstance, the high frequency luminance signal is turned off by multiplier 116, as the 11Y control" signal is reduced to zero. At the same time the intra-frame difference signal is added to the chrominance outputs by the action of multipliers 132, as the "U, V control" signal increases to unity.

More generally, the system could be implemented in whole or in part by software, in which case the figures should be regarded as flow charts or logic diagrams.

It will be appreciated that the example is given above in relation to a 625/50 interlaced PAL colour video signal, but that the principles described are equally applicable of other types of colour video signal.

Claims (5)

1. A video motion detector, comprising: an input (12) for receiving a chrominance component signal; delay means (20) coupled to the input for delaying the chrominance component signal by one video frame or field; comparison means (22, 24, 26) coupled to the inDUt and to the output of the delay means to compare the chrominance component for the current frame with that for the previous frame or field, and for providing a motion signal output which takes a first value when the differences between the current and previous frame or field chrominance components are relatively small indicating an absence of movement and a second value when the said differences are large indicating the presence of movement; and chrominance detection means (32, 34, 36, 38) coupled to the input for detecting whether there is chrominance in the current frame or field, and in the absence of chrominance in the current frame or field forcing the motion signal output to its first value.
2. 2. A video motion detector according to claim 1, comprising a horizontal low-pass filter (14; 52) coupled in series with the delay means.
3. 3. A video motion detector according to claim 1, in which the said horizontal low-pass filter (52) is located in series with the output of the delay means, and in which the chrominance detection means includes a second, corresponding horizontal low-pass filter (54).
4. 4. A video motion detector according to any preceding claim, for use with interlaced video signals, further including means (16, 18) for averaging adjacent lines of the input chrominance component signal.

   g 1^
5. 5. A video motion detector according to any preceding claim, having separate inputs (12, 12A) and delay means (20, 20A), for two chrominance component signals, and in which the comparison means and chrominance detection means operates on the two chrominance component signals.

   11 -