GB2044036A - Colour Separation Overlay of Television Signals - Google Patents

Abstract

In colour separation overlay of television signals the value of the keying signal is defined on the chrominance phase diagram by two straight-line generators which pass through the origin, enabling the keying signal to be derived from encoded PAL. The keying signal is also used to change any chrominance components within a keying sector thus defined and which contains the backdrop colour to a colour at the edge of the sector. The chrominance components are separated out (7) from an encoded PAL input signal and demodulated (8, 9) with respect to two conveniently- chosen phase axes. A measure of the vector difference of the instantaneous chrominance phase from the generators in a direction parallel to a key colour axis is obtained (10 to 19) and used to control a selector switch (25) which switches between the foreground and background signals. In the presence of a keying signal, the foreground signal has its luminance reduced and its colour changed so as to be outside the keying sector (21 to 24). <IMAGE>

Description

SPECIFICATION Colour Separation Overlay of Television Signals This invention relates to colour separation overlay of television signals, and in particular to the generation of a keying signal for use in such overlay.

In colour separation overlay, two scenes are electrically combined into a single composite picture in the following manner. A first scene consists of the elements of a foreground picture typically including actors, in front of a background of a specified colour, usually blue. The second scene consists of a background picture, typically of a distant location. By use of colour separation overlay the foreground picture is made to appear in front of the background picture, thus the actors can be made to appear to be at the distant location. To do this a camera views the first scene and provides a first video signal, and second video signal is generated representative of the second scene by any suitable means such as a camera, telecine machine, or video tape recorder, synchronised with the camera viewing the first scene.From the first scene there is also produced a waveform, known as the keying signal, which indicates whether the first video signal instantaneously corresponds to a back-drop area of the first scene or a foreground area of the scene. This keying signal can then be used to operate a fast selector switch so as to select the first video signal when this represents foreground information, but at other times to select the second, background video signal, to provide the composite signal.

The keying signal or a signal related to it can also be used to modify the first video signal by reducing the signal amplitude in those areas corresponding to backdrop information. In this way any tendency for the final composite picture to have fringes of the backdrop colour round the foreground picture elements is at least substantially reduced.

Originally the keying signal consisted of the blue component of the first camera output after passing through a threshold circuit. In accordan'ce with our British Patent No. 1,189,402 it is now preferred to use either the B-Y signal or preferably a signal B--21(R+G) where B, R and G are the three camera signals represenative of the blue, red and green signal components, and Y represents the luminance component. These keying signals have the advantage that they are at least substantially independent of the scene brightness or the luminance of the video signal.

Hitherto colour separation overlay has operated with R G B signals, that is with uncoded signals. There is however a demand for a system in which a keying signal can be derived from an encoded PAL or N.T.S.C. signal, without first having to decode the signal down to its R, G and B components.

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

In accordance with this invention, therefore, we provide a colour separation overlay system in which the value of the keying signal is defined on a phase diagram of the chrominance components of the television signal at least in part in terms of two non-coincident substantially straight-line generators which pass through the origin. By defining the keying signal in this way it then becomes possible to derive the keying signal from an encoded PAL or N.T.S.C. signal with relatively little processing. This means that colour separation overlay can now more easily be used when the foreground scene has been recorded on video tape or is produced by a telecine machine.

It will be appreciated that the phase diagram, also known as the R-Y/B-Y graph, displays (at least approximately) hue in terms of phase, and saturation in terms of radial distance, and that it is substantially independent of luminance. By defining the keying signal in this way we are therefore defining it essentially in terms of hue.

This particular form of definition therefore has advantages not only as to the way in which the keying signal can be derived from an encoded signal, but also because the keying signal obtained is accurately related to colour and is unlikely to produce spurious "breakthrough", this being caused by hues in the foreground picture which are interpreted as being of the backdrop colour.

According to a further feature of the invention, the keying signal, preferably obtained as above, is used to change the chrominance components of an encoded video signal when these indicate a colour within a specified region of a phase diagram of the chrominance components, for example to a colour at the edge of the said region.

This then suppresses hues corresponding to the backdrop colour, thereby reducing any colour fringing effect.

The invention will now be described in more detail, by way of example, with reference to the drawings, in which: Fig. 1 is a graph showing the R-Y/B-Y phase plane for a PAL colour television signal; and Fig. 2 is a block circuit diagram of a colour separation overlay system embodying the invention.

Referring first to Fig. 1, the R-Y/B-Y phase plane is shown and there are two orthogonal axes termed the "key colour axis" and the "key+900 axis". Typically these will respectively be the blue and blue +900 axes, i.e. at 130 and +770 relative to B-Y, but in principle they could be any pair of orthogonal axes in this plane and could be at any fixed angle to the B-Y and R-Y axes, subject only to the key colour axis being related to the colour of the backdrop in the foreground scene. Fig. 1 is thus essentially equivalent to a phase/amplitude diagram for the chrominance subcarrier of the PAL signal, the PAL switch being ignored, subject to rotation by 130.

From the origin two straight lines are drawn, namely OA and OA'. These generators will define the keying signal. Thus if the signal from the first scene falls in the region containing the positive key colour axis to one side of the line A-O-A', the corresponding area of the first scene will be taken to be backdrop, while otherwise it will be taken to be foreground information.

It will be seen that the lines OA and OA' are non-coincident, that is to say they are at an angle to one another and do not form a single straight line.

It is assumed in this example that the lines OA and OA' are equally anguiarly spaced to either side of the key colour axis by an angle 0. This is not essential; the lines OA and OA' could be spaced by different angles 0, and , from the key colour axis. The backdrop colour is the colour defined by the key colour axis. In a practical example the angle (k could bs in the range 400 to 750, and typically about 650.

It will now be assumed that the colour of the first scene at any instant is represented by the point X of radius x and angle 6. The keying signal depends on whether or not the vector PX is positive or negative.

The vector OX can be resolved into two components: Along the key colour axis: OR=x cos 6 Along the key +900 axis: XR=x sin 6 The ratio of these two components gives tan 6, or the slope of OX. This could be tested against two fixed values corresponding to the slopes of OA and OA' to determine on which side of the line A-O-A' the point X lies. If the lines OA and OA' lie symmetrically about the key colour axis, then the modulus of tan 6 could be tested against a single fixed value.

Alternatively, the vector PX can be evaluated and its polarity determined. This vector PX can be used to effect "hue suppression" to remove colour fringes round foreground objects by removing the offending colour from the signal and reducing the luminance. From Fig. 1 it can be seen that: PX=QR=(OR-OQ) and PQ XR OQ=------= tan tan xsin 6 PX=x cos 0 tan 5) In fact, this is the length of the vector from point X parallel to the key colour axis to the line A-O A", which is incorrect for colours with negative 6.

To take account of the discontinuity in A-O-A', the modulus of sin 0 must be taken, to give: sin 61 PX=x(cos 6- tan 6 in the case where A-O-A' is symmetrical about the key colour axis. Were it not, two alternative relations would have to be used depending on whether 6 were positive or negative.

In Fig. 1 the sector A-O-A' does not in practice extend to infinity but is bounded by the practical values corresponding to maximum saturation.

Furthermore the keying region could possibly have an inner bound given byx > K3 oryx cos 6 > K3 so that unsaturated colours very close to white would not produce a spurious keying waveform.

The system illustrated in Fig. 2 will now be described.

The apparatus has an input 30 for receiving a first encoded PAL signal representative of the foreground scene from a colour camera or other video source. This is applied to a chrominance bandpass filter 7 the output of which is applied to two demodulators 8 and 9 which are fed with chrominance subcarrier in quadrature phase relation. The filter 7 can be of any suitable type, e.g. active or passive, but preferably operates as a comb filter to separate the chrominance components from the high-frequency luminance components, and in this case is conveniently a transversal filter.

The chrominance subcarrier can be derived from the PAL signal by means of a burst locked oscillator 2 and a PAL square wave switching generator 3, both connected to the input 30. A key axis synthesiser 4 receives the output of the osciliator 2 or alternatively an independent but synchronised source of reference subcarrier from an input 32 after limiting in a limiter 1, as selected by a switch 34. In either event the subcarrier signal and the PAL switching signal are applied to the synthesiser 4 the output of which is a subcarrier signal of defined predetermined phase. This is achieved by deriving U and V subcarrier phases and recombining them in the correct proportions.The output of synthesiser in fact has the phase of the key +900 axis and this is changed by 900 in a phase shift circuit 6 to produce key axis subcarrier, and is also inverted by 1 800 on alternate lines in a PAL switch circuit 5 to produce key +900 axis subcarrier alternately.

The output of phase shift circuit 6 is connected to demodulator 8 and the output of the PAL switch 5 is connected to demodulator 9. The output of demodulator 8 is low-pass filtered in a filter 10 and represents x cos 6, while the output of demodulator 9 is low-pass filtered in a filter 11 and represents x sin 6. The filters 10 and 11 remove residual and harmonics of subcarrier. The sin 6 path now has to be split to take account of the modulus of sin 6. As shown this is done by an amplifier 12 with differential outputs and which supplies positive signals at a first output connected to a clamping and clipping circuit 13, and inverts its input to supply negative signals at a second output connected to a clamping and clipping circuit 14. In circuits 13 and 14 the signals are clamped to black level, and clipped to remove any residual negative-going components.

Attenuators 1 5 and 16, which may be adjustable, are connected respectively to the circuits 1 3 and 14 and attenuate by a factor representing tan 0. If the generators OA and OA' in Fig. 1 are not symmetrically arranged about the key colour axis OR, then the attenuation factors of attenuators 15 and 1 6 will be different. The attenuator outputs are combined in an adder 17, to give a signal representing (ixsin 6l}/tan 0, and the output of adder 17 is applied to the inverting input of a subtractor 19. If both attenuators 1 5 and 16 are set at equal, e.g. zero (0 dB), attenuation, then the summation in adder 17 would produce a conventional full-wave rectified signal.The noninverting input of subtractor 1 9 receives the x cos 6 signal from filter 10 after passing through a compensating delay 18 to allow for the delay in the x sin 0 channel. It will be appreciated that the adder 1 7 and subtractor 1 9 can of course be constituted by a single combining circuit.

The output of subtractor 1 9 is thus representative of: lxsin el x cos 6- tan which is the desired vector PX. When this signal is positive, the first video signal instantaneously contains backdrop information. When it is negative, the video signal contains foreground information. This signal is applied to a further clamping and clipping circuit 20 the output of which constitutes the keying signal. The circuit 20 again clamps at black level and removes any negative-giving components. The keying signal is used to control a video selector switch 25 which selects either the foreground signal or the background encoded video signal received at a terminal 26 as an output 27.The switch 25 preferably combines the signals in proportions dependent on the keying signal, and is arranged to give a "soft edge" or gradual change at transitions between the foreground and background scene.

The keying signal is also applied through an attenuator 22 to an inverting input of a combining circuit 24 to suppress the luminance components of the foreground scene. A modulator 21 modulates the keying signal onto key colour subcarrier with the appropriate sense, this being 1 80O out of phase with the key colour, and this is also applied to the circuit 24 to enable colour suppression. Finally the combining circuit receives the signal from input 30 after passing through a compensating delay 23. In the presence of the keying signal, the luminance is reduced and the colour altered so as to be outside the sector A-O A' in Fig. 1. The output of circuit 24 is applied to the switch 25 as the video signal representing the first or foreground scene.

As noted above, in operation of the apparatus the key axis subcarrier is synthesised from the U and switched V component and is therefore correctly PAL phased. For N.T.S.C. it would be possible to generate the key +90 axis subcarrier simply by phase shifting the key axis subcarrier with a quarter-cycle delay, but the PAL this would mean that the sine output of demodulator 9 would alternate in sign on each line. This could be corrected by a full-wave rectifier, provided that the keying sector defined by OA and OA' in Fig. 1 is symmetrical about the key colour axis. This can be disadvantageous since the primary and secondary colours are not distributed evenly around the vector diagram, due to the U, V weighting factors.If the key +90 axis subcarrier is generated using the PAL switch, as shown, then the position of the colour X in Fig. 1 will be correct on every line, and the rectifier formed by elements 12 to 1 7 can be designed with unequal gains in the positive and negative paths so that the angle 4 is different on each side of the key axis. This proves to be valuable for accurate hue suppression.

The keying signal represented by PX in Fig. 1 is a measure of how much key colour ought to be removed from the foreground picture for hue suppression. This signal is then used to drive modulator 21, also fed with reference subcarrier in anti-phase with the key colour (correctly switched for PAL), whose output is added to the composite foreground signal in circuit 24.

Consequently, if the foreground scene contained pure key colour, the chrominance in the foreground signal would be reduced to zero. This will not matte the areas of key colour in the foreground to black, since the luminance component is still present, so a proportion of the keying signal determined in attenuator 22 is subtracted from the foreground signal. This proportion should change for different key colours according to the matrix coefficients of PAL system I. Referring back to Fig. 1, all colours in the keying sector will be moved, parallel to the key colour axis, until they lie on the boundary marked out by the generators A-O-A', and the luminance will be suitably reduced.

If the asymmetrical system with unequal control of the fullwave rectifier outlined above is used, O-A can be set to correspond to the nearest primary or secondary colour in the positive sense, and O-A' similarly in the negative sense. In this way the system can emulate most accurately the present equipment in broadcasting use, despite the weighting factors used in System It will be appreciated that the circuit shown can be modified in many ways. For example, to determine whether the foreground signal represents a colour in the keying sector it would be possible simply to demodulate about two (nonorthogonal) axes which were perpendicular respectively to the generators OA and OA', and look for two positive resultants.

A particular advantage of this system is that it is not restricted to using the primary and secondary colours as the key colours, but can use any permissible hue, so that the system can be adjusted to suit the particular foreground scene, rather than the scene being adjusted to suit the system.

The system can either operate in a 'Simple PAL' mode, or be modified with a delay line to use 'Delay PAL' type decoding, so reducing the effects of phase errors causing, for example, jitter on the edges of the key waveform.

Because the system operates with encoded television signals, the bandlimiting inherent in the chrominance signal will mean that the keying waveform is less sharply defined than would be the case if it were derived from R, G and B before encoding. Nevertheless, we have found that the results obtained are subjectively quite acceptable.

Claims (21)

Claims

1. A method of generating from a colour television signal a keying signal for use in colour separation overlay, in which the value of the keying signal is defined on a phase diagram of the chrominance components of the television signal at least in part in terms of two non-coincident substantially straight-line generators which pass through the origin.

2. A method according to claim 1, in which the chrominance components of the television signal are demodulated with respect to two axes on the phase diagram.

3. A method according to claim 2, in which the said axes are orthogonal.

4. A method according to any preceding claim, in which any chrominance components which fall within a keying sector at least partially defined by said generators are modified so as to represent a colour at the edge of or outside the keying sector.

5. A method according to any preceding claim, in which the chrominance components are separated by comb filtering.

6. A method according to any preceding claim, in which when a keying signal is present the luminance of the television signal is altered.

7. A method according to any preceding claim, in which the expression: Ix sin el x cos 6- tan 0 is evaluated, where x and 6 are the amplitude and phase of the chrominance components of the input signal, and 0 is an angle less than 90 degrees.

8. A method of generating a keying signal for use in colour separation overlay, substantially as herein described with reference to the accompanying drawing.

9. A method of colour separation overlay in which the chrominance components of the television signal are changed when they indicate a colour within a specified region of-a phase diagram of the chrominance components.

10. A method according to claim 9, wherein the chrominance components are changed to a colour at the edge of the said region.

11. Apparatus for use in the method of claim 1, comprising an input terminal for receiving a colour television signal, and means connected to the input terminal for deriving a keying signal with a value defined on a phase diagram of the chrominance components of an input television signal at least in part in terms of two noncoincident substantially straight-line generators which pass through the origin.

12. Apparatus according to claim 11, including two demodulators for demodulating the chrominance components with respect to two axes on the phase diagram.

13. Apparatus according to claim 12, in which the said axes are orthogonal.

14. Apparatus according to any of claims 11 to 13, including means for modifying any chrominance components which fall within a keying sector at least partially defined by said generators so as to represent a colour at the edge of or outside the keying sector.

15. Apparatus according to any of claims 11 to 14, including a comb filter to separate the chrominance components from the television signal.

1 6. Apparatus according to claim 15, in which the comb filter is constituted by a transversal filter.

1 7. Apparatus according to any of claims 11 to 1 6, including means for altering the luminance of the television signal in the presence of a keying signal.

1 8. Apparatus according to any of claims 11 to 17, including means for evaluating the expression: Ix sin 61 x cos 6- tan s where x and 6 are the amplitude and phase of the chrominance components of the input signal, and (kis an angle less than 90 degrees.

1 9. Colour separation overlay apparatus substantially as herein described with reference to the drawing.

20. Colour separation overlay apparatus, including means for changing the chrominance components of the television signal when the chrominance components indicate a colour within a specified region of a phase diagram of the chrominance components.

21. Apparatus according to claim 20, in which the said means changes the colour to a colour at the edge of the said region.