GB2198008A - Limiting component video signal - Google Patents

Abstract

Component colour video signals Y, R-Y and B-Y (Y, U and V are limited in such a manner that the corresponding RGB components lie within specified limits. A two-dimensional range of values of valid U and V components is defined for values of constant luminance. An input signal is compared with the maximum and minimum U and V values in the valid range for the appropriate luminance value. If the U, V values of the input signals are outside the valid range they are moved within the valid range by multiplying both signals by the same attenuating factor K determined in accordance with the results of the comparison. <IMAGE>

Description

LIMITING COMPONENT VIDEO SIGNALS This invention concerns a method of and apparatus for limiting Y,R-Y, B-Y component colour video signals in such a manner that the corresponding RGB components lie within specified limits.

The need for such limiting may arise with electronic digital video signal generators which can generate signals in Y,R-Y,B-Y form which correspond to non-existent colours. Similar results can occur if two signals are added for example. The theoretical principles of the conversion of RGB limits to Y,R-Y,B-Y limits have been described in a paper entitled "Determining Valid Component Analog Video Signals with a 3-D Vector Representation" by Earl Matney and Dan Baker published in SMPTE Journal, May 1986, pages 550 to 556.

This paper was concerned with providing a vectorscope type display for the signals.

For convenience in this specification the B-Y and R-Y signals will be referred to as U and V signals respectively. These signals are also sometimes referred to as the CB and CR signals.

The known method of limiting YUV component signals such that the corresponding RGB components lie within specified limits is first to convert the YUV components into RGB form, secondly limit the resulting RGB signals and then lastly convert back into YUV components. However this method requires somewhat complex signal processing and the results obtained are not entirely satisfactory, e.g. the hue (colour) of picture areas affected by this limiting process is normally altered. We have appreciated that it would-be very advantageous to have a method of limiting based on the conversion of the RGB limits into corresponding YUV limits which can be applied directly to a YUV video signal without any conversion of the video components into RGB form.

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

The method operates by testing against allowed U,V values for levels of constant luminance and if an invalid colour is found changing the colour without changing the luminance. This takes advantage of the fact that although the valid Y,U,V values form a parallelepiped in Y,U,V space, the sections taken through this block at values of constant luminance have a large number of sides parallel to the U,V axes. Only one face is not thus parallel and this always has a constant slope. As a result, locating and changing invalid colours is made very much easier that would otherwise have been the case.

A preferred embodiment of the invention to be described in more detail below provides a method of limiting the magnitudes of the U and V colour difference components of a YUV component video signal with the object of obtaining YUV components which provide valid RGB components, valid components being defined as those whose R,G and B magnitudes lie within specified RGB limits.

In the preferred method, the instantaneous magnitudes, Uin and Vin, of the input U and V components are multiplied by the same factor K to give limited output values K*Uin and K*Vin where the factor K is equal to: (a) zero if the instantaneous magnitude Yin of the luminance component combined with zero magnitudes of Uin and Vin would provide invalid RGB components, i.e. if Yin is above the luminance level for white or below the luminance level for black: or (b) unity if Yin, Uin and Vin provide valid values of R, G and B; or otherwise (c) the maximum positive value of K for which components Yin, K*Uin and K*Vin would provide valid R, G and B compents.

Preferably three intermediate values of K are determined and the least of these is selected as the required value of K unless this minimum value is greater than unity in which case the final value of K is equal to unity.

Except for certain conditions under which they are set to zero or unity, the three intermediate values of K are given by: First value: the positive value of K equal to Vmax(Yin)/Vin or Vmin(Yin)/Vin where Vmax(Yin) and Vmin(Yin) are the maximum and minimum values of V for the given value of Yin which would provide valid values of R,G and B, the value of Uin being ignored.

Second value: the positive value of K equal to Umax(Yin)/Uin or Umin(Yin)/Uin where Umax(Yin) and Umin(Yin) are the maximum and mimimum values of U for the given value of Yin which would provide valid values of R, G and B, the values of Vin being ignored.

Third value: the positive valueof K for which components Yin, K*Uin and K*Vin would provide RGB components with G=O or G=1.

Apparatus for use in the method can comprise, firstly, separate means for deriving the three intermediate K factors from input Y, U and V values, secondly means for selecting the minimum of these three K factors to give a final K factor which is then restricted to a maximum value of unity, and thirdly means for multiplying the input U and V values by the final K factor to give the required output values of U and V.

The apparatus could alternatively comprise a read only memory for which each input address consists of Y, U and V input values and whose output is the corresponding value of K, and secondly means for multiplying the input U and V values by the value of K thus derived to give the required output U and V values.

A computer which operates on stored data representing a YUV component video signal could be used to produce the required limited output values of the U and V components by means of a computer programme.

The system is particularly suitable for processing digital signals and is described in that context but in principle could be applied to analogue video signals.

The invention will now be described in more detail, by way of example, with reference to the drawings, in which: Figure 1 illustrates the valid colours in 3-D RGB space; Figure 2 illustrates the block of valid colours in YUV space; Figure 3 shows three mutually perpendicular views of the YUV valid colour parallelepiped block, namely (a) the "top" view on the Y axis, (b) a "side" view on the V axis and (c) a "side" view on the U axis; Figure 4 shows a series of cross-sections through the YUV valid colour block taken at various constant-Y values, showing at (a) the cross-sections at the Y values given by the primary and secondary colours (y,c,g,m,r and b) and (b) a range of intermediate values in the bottom half of the block (Y less than 128);; Figure 5 shows a cross-section through the block taken at a constant luminance level which intersects the G=O or G=l surface (the "top" and "bottom" sloping surfaces); Figure 6 illustrates the preferred method of determining the limiting factor K; Figure 7 is a series of block diagrams showing how the signals necessary to implement the method of Figure 6 can be obtained; Figure 8 is a block diagram showing an alternative way of implementing the method of Figure 6, in which (a) is the basic circuit and (b) shows optional additions for reducing quantising effects; and Figure 9 is a diagram based on Figure 5 illustrating an alternative simpler method of limiting the U and V signals while still retaining luminance constants.

For analogue video signals, the luminance and colour difference signal magnitudes Y,B-Y and R-Y are related to R,G and B signals by the equations: Y - .299R + .587G + .114B B-Y - .886B - .587G - .299R R-Y = .701R - .587G - .114B Assuming that values of R,G and B are to be confined to the range from O(zero) to 1 volts, the minimum and maximum values of Y,B-Y and R-Y are given by: : O < Y < 1 -0.886 < B-Y < 0.886 -0.701 < R-Y < 0.701 When Y, U and V component signals are digitally encoded according to CCIR Recommendation 601 (Encoding parameters of digital television for studios, Geneva 1982, Vol XI Part 1 pages 271 to 273), the maximum and minimum 8-bit values of the Y component are defined as being 235 (white level) and 16 (black level) while those of the U and V signals are defined as being +112 and -112.It follows that the 8-bit values of these digital YUV components are related to RGB voltages limited to the range 0 to 1 volts by the equations: Y = 16 + 219 * (.299R + .587G + .114B) U = 112 * (.886B - .587G - .299R) / 0.886 V - 112 * (.701R - .587G - .114B) / 0.701 The resulting YUV values for the colours in a 100 saturated colour bar signal are shown in Table 1.

TABLE 1.

8-bit YUV quantum levels for 100% saturated colours Colour R G B Y U V Colour PG 3 Y U V White (w) 1 1 1 235.0 0 0 Yellow (y) 1 1 0 210.0 -112.0 18.2 Cyan (c) O 1 1 169.5 37.8 -112.0 Green (g) 0 1 0 144.5 -74.2 -93.8 Magenta (m) 1 0 1 106.5 74.2 93.8 Red (r) 1 0 1 81.5 -37.8 112.0 Blue (b) 0 0 1 41.0 112.0 -18.2 Black (bk) 0 0 0 16.0 0 0 In three dimensional (3-D) RGB space, the R,G and B values of the colours in Table 1 give the co-ordinates of the corners of a rectangular block as illustrated in Fig.1. The six faces of this block are planes given by R=0, G=0, B=0, R=1, G=1 and B=1. The volume enclosed by this block defines the limits of all points in RGB space which have 'valid' combinations of R,G and B.

By transforming the RGB points on the surface of this block into YW values, a block can be constructed in 3-D YUV space which encloses all YUV values corresponding to valid RGB values. This block will be referred to as the 'YUV valid-colour' block and its shape is illustrated in Fig.2. Since the RGB to YW transformation is a linear process, the corners of the YW valid-colour block are, like the RGB block of Fig.1, given by the co-ordinates of the colours in Table 1 and the surfaces are flat planes corresponding to R=G=B=0 and R=G=B=1. Further details of the shape of this YUV block can be conveniently illustrated by three, mutually perpendicular,views or projections obtained by viewing along the Y,U and V axes. These views are obtained by plotting U versus V, Y versus V and Y versus U as shown in Fig.3.Further useful information is obtained from cross-sections of the YUV block at different constant values of Y as shown in Fig 4.

It should be noted that the four saturated colours having 3=0 (black, red, green and yellow) and the four having B=1 (blue, magenta, cyan and white) lie along straight lines in the Y-U plane of Fig.3(b). This means that the faces of the valid-colour block given by 3=0 and B=1 are perpendicular to the Y-U plane. Similarly, the faces given by P=0 and R=1 are perpendicular to the Y-V plane. As a result the edges of the cross-sections of the valid-colour block as shown in Fig.4 are predominantly parallel to the U or V axes; the slanting edges occur where the cross-sections meet the G=0 and G=1 planes of the block. Note that all slanting edges have the same slope.

The proposed YUV limiting technique has the effect of moving any invalid combination of Y,U and V to a point on the surface of the YUV valid colour block. The direction of movement can be defined as the direction which maintains: (a) constant hue; this corresponds to moving directly towards the origin in the U-V plane of Fig.3(a) (b) constant luminance: this corresponds to moving horizontally towards the Y axis in the U-Y and V-Y planes of Figs.3(b) and 3(c) Using these rules of constant hue and luminance. the video components Yout,Uout and Vout given out by the limiter are related to the input components Yin,Uin and Vin by: Yout = Yin Uout = K * Uin Vout = K * Vin where K = 1 for valid colours and 0 < = K < 1 for invalid colours.In the following discussions, K will be referred as the W limiting factor The effect of these rules on pictures containing invalid colours has been found te be very satisfactory from a subjective point of view.

To obtain the value of the W limiting factor K for any combination of Y.U and V, mathematical details of the edges of the cross-section of the valid-colour block for the given Y are required.

These details can be defined in terms of five parameters which are dependant on Y plus a constant giving the slope of the slanting edges. These parameters are illustrated in Fig.5 for a value of Y lying between that of 100% red and magenta and are defined as follows: (a) Umax(Y) = maximum positive valid U for given Y (b) Umin(Y) = maximum negative valid U for given Y (c) Vmax(Y) = maximum positive valid V for given Y (d) Vmin(Y) = maximum negative valid V for given Y (e) Vo(Y) = value of V at the point where a slanting edge of a cross-section (caused by the G=C limit for Y < 128 or G=l limit fcs Y > 128) or an extension of this edge. crosses the V axis.

(f) Slope. dV/dU of slanting edge = -0.482 for the definitions cf digital YJV magnitudes given in Section 2.

From the above definitions, the equation of a slanting edge is: V = -0.482*U + Vo(Y) The value of Vo(Y) for a given Y and the slope of the slanting edges can be calculated by setting U=O and G=O or G=1 in the equations for Y,U and V given above. Note that, as indicated in Fig.4, the slanting edge given by the G=0 limit encroaches on the rectangular limits given by Umax(Y) ,Umin(Y) ,Vmax(Y) and Vmin(Y) only if Y is less than that for 100% saturated magenta,Ymag. Thus the G=0 limit need be considered only if Y < Ymag. Similarly, the 6=1 limit need be considered only if Y > Ycyan.However, the instrumentation of a limiter may be more convenient if the G=O and 6=1 limits are determined for the wider range of Y values given by Y < 128 and Y > =128 respectively.

For values of Y between black level (Y=16) and white level (Y=235). values of Umax(Y) and Umin(Y) can readily be derived from the Y-U projection of the valid-colour block shown in Fig.3(b). Similarly, Vmax(Y) and Vmin(Y) can be derived from the Y-V projection shown in Fig.3(b).

For all Y values below black level and above white level, the values of Umax(Y), Umin(Y), Vmax(Y), Vmin(Y) and Vo(Y) are all made equal to zero.

A procedure which has been devised for calculating the value of the UV limiting factor K from the parameters given above may be explained as follows. Suppose that the cross-section of the valid YUV block at Yin is as shown in Fig.6 and that Uin and Vin correspond to point A in this figure. Points B, C and D correspond to three possible values of Uout and Vout, i.e. they are the points where the line joining point A to the origin 0 crosses the limits given by U=Umax(Y), V=Vmax(Y) and V=-0.482U+Vo(Y). Point F is the intersection of the V-axis with a line passing through point A and drawn parallel to the slanting edge of the cross-section. As a result, V=Vin+0.482Uin at. point F.

Remembering that K is the ratio Uout/Uin or Vout/Vin, it can be seen that points B, C and D correspond to three different values of K given by: Kb = OB/OA = Vmax(Y) / Vin Kc = OC/OA = Umax(Y) / Uin Kd = OD/OA = OE/OF = Vo(Y) / (Vin+0.482Uin) For invalid colours, the required value of K is lowest of these three values. Let this value be Kmin. If Kmin > 1. point A must lie inside the valid cross-section and therefore no limiting is required: this is acheived by setting K=1 if Kmin > 1.

A method of handling all possible combinations of Y,U and V will be explained using the following procedure. It is assumed that the values of Umax(Yin) , Umin(Yin), Vmax(Yin), Vmin(Yin) and Vo(Yin) required by this procedure would be calculated first and stored in look-up tables, so that the procedure can be implemented in the form of a computer programme.

input Yin,Uin,Vin, if Vin > 0 then Kb = Vmax(Yin) / Vin else if Vin(O Kb - Vmin(Yin) / Vin else (if Vin=0) Kb = 1 if Uin > 0 Ke = Umax(Yin) / Uin else if Uin < 0 Kc = Umin(Yin) / Uin else (if Uin=0) Kc = 1 if Yin'128 AND (Vin+0.482Uin) > 0 then /*Test for 6=0 limit*/ Kd = Vo(Y) / (Vin+0.482Uin) else if Yin > =128 AND (Vin+0.482Uin) < 0 then /*Test for G=1 limit*/ Kd = Vo(Y) / (Vin+0.482Uin) else Kd = 1 Kmin = lowest(Kb,Kc,Kd) if Kmin > 1 then Kmin=1 Uout =Emin*Uin Vout =K > in*Vin Greater speed could be achieved by means of a look-up table containing pre-calculated values of Uout and Vout for each possible combination of Yin,Uin and Vin.

This technique has the disadvantages of requiring a very large look-up table and/or of giving coarsely quantised values of Uout and Vout. As an example, for high-accuracy 8-bit input and output YUV values.this table would be required to give a 16-bit output for each of the 17 million possible 24-bit YUV input addresses, giving a total storage requirement of about 33 Mbytes. This size of table is too large to be acceptable. More realistically, the use of 6,5 and 5 bits for Yin, Uin and Vin respectively and 8 bits for each of Uout and Vout might give an acceptable compromise between quantising accuracy and storage capacity (131Kbytes). Processing errors for valid colurs could be avoided by setting the output for these colours to be a fixed invalid code which is used to signal that Uout and Vout should be identical to Uin and Vin respectively.

The size of the store could be halved by storing 8-bit values of K rather than the two 8-bit values of Uout and Vout. The only disadvantage would be an increase in processing time required for the multiplication processes Uout=K*Uin and Vout=K*Vin.

A further halving of the required storage could be achieved by making use of the fact that the YUV valid colour block is skew symmetrical about the cross-section for mid-grey luminance,Ygrey, where Ygrey lies midway between the luminance levels for black and white. This means that the value of K for components (Ygrey+X),U,V is the same as the value of K for (Ygrey-X) ,-U,-V and therefore the K factor for any luminance level above Ygrey can be easily found from stored values for luminance levels below Ygrey.

All the software routines discussed above can be instrumented in hardware. The best technique for a given application depends on the cost, complexity and performance requirements. Various compromises between these requirements are discussed below.

For any of the possible techniques, it is convenient to make extensive use of PROMs (Programmable Read Only Memories) to store parameters such as Umax(Y) and to perform various processions operations. MOS EPROMs are now available at a reasonable cost which have capacities up to 128K*8 bits i.e. they can have 17 bit input addresses and give 8 bit outputs. Somewhat unfortunately, the access time of currently available large MOS EPROMs, which is typically 150 to 250 nsecs, is slightly greater than the 148 nsec sampling interval of the 6.75 Mhz clock specified for U and V signals in CCIR Rec.601.

This problem can be overcome by demultiplexing the U and V signals by a factor of two and by doubling the number of EPROMs but results in increased circuit components. For storage capacities up to about 8K*8 bits bipolar EPROMs with access times of only about 60 nsec can be employed.

Diagrams indicating possible circuitry are shown in Figs.7 and 8.

The number of bits shown for signal paths in these figures are suggested compromises for good performance without undue circuit complexity. For simplicity, no demultiplexing of U and V signals to allow for practical access times has been shown. It has also been assumed that, if necessary, the Y signals have been sub-sampled to the same rate as that of the U and V signals.

Fig.7 shows a method of determining the three possible K factors and selecting the lowest of these as in the psuedo computer programme given above. A desirable feature of this circuit is that good quantising accuracy can be obtained without the need for excessively large ROMs. The main disadvantage is circuit complexity.

Fig.8(a) shows a circuit arrangement based on a single, large look-up table. The main advantage of this ciruit is its basic simplicity. The main disadvantage is the very large storage capacity required for good quantising accuracy. Improvements in quantising performance for a given size of PROM can be obtained using the additional circuitry shown in Fig.8(b). The purpose of the inverters and switches shown in this figure is to take advantage of the skew-symmetry of the valid YUV block about mid-srey level, Ygrey.

This processing is based on a re-defined valid YUV block for which the luminance of white level, Ywhite, is quantum level 239 rather than 235. With this value of Ywhite, the resulting value of Ygrey (=255/2) is such that (Ygrey+X) can be converted to (Ygrey-X) by simply inverting its binary code. This change of Ywhite to level 239 is for limiting purposes only and does not contravene any video signal specifications.

The purpose of the error feedback circuits in Fig.8(b) is to minimise the visibility of quantising errors resulting from the reduced number of bits in the YUV signals fed to the PROM.

A simpler, though less preferred method of ensuring valid U and V values is provided by a routine in which U and V are limited independentlv of one another in much the same way that R, G and B signals are normally individually limited. Like RGB limitation, this routine has the disadvantage of altering hue, but does not change luminance. The technique is as follows: input Y U V if U > Umax(Y) then U=Umax(Y) else if U < Umin(Y) then U=Umin(Y) if V > Vmax(Y) then V=Vmax(Y) else if V < Vmin(Y) then V=Vmin(Y) if YYrnaa AND VYv'o(Y)-O.4B2U then V=Vo (Y) -o 482U else if Y > Ycyan At V < Vo(Y)-0.482U then V=Vo(Y)-C.482U The effect of the above routine is illustrated in Fig.9. In this figure.the arrowed lines show the direction of movement of W co-ordinates caused by the limiting process. Thus invalid UV values given by point A would be changed to valid values at point D via points B, where U-Umax(Y) and C where V-Vmax(Y)

Claims (12)

1. CLAIMS 1. A method of limiting Y, R-Y and B-Y component colour video signals (Y,U,V signals) such that corresponding RGB components lie within specified limits, comprising the steps of: defining a twodimensional range of valid U,V values for values of constant luminance, comparing an input signal with the range of valid U,V values appropriate to its luminance value and if the U,V values of the input signal are outside the valid range changing them until they are within the valid range.

   2. A method according to Claim 1, in which the comparing step comprises comparing the input signal with maximum and minimum permitted values of U and V respectively for that luminance.

   3. A method according to claim 2, in which the comparing step additionally comprises comparing the input signal with one (U,V) function appropriate to that luminance.

   4. A method according to Claim 1, 2 or 3, in which the changing of invalid U,V values is achieved by multiplying U and V by the same attenuating factor to retain the hue represented by the input signal.

   5. A method according to Claim 4, in which the attenuating factor is determined as the smallest positive values of (i) Vmax/Vin or Vmin/Vin where Vmax and Vmin are the maximum and minimum values of V for a given value of luminance which would provide valid values of R, G and B, and Vin is the value of V in the input signal, (ii) Umax/Uin or Umin/Uin where Umax and Umin are the maximum and minimum values of U for a given value of luminance which would provide valid values of R, G and B, and U is the value of U in the input signal, and (iii) the value of K for which components Yin, K*Uin and K*Vin would provide RGB components with G=O or G=1.

   6. Apparatus for use in the method of Claim 1, comprising means defining a two-dimensional range of valid U,V values for values of constant luminance, means for comparing an input signal with the large bf valid U, values appropriate to iks luminance values, and means dependent upon the result of the comparison operative if the U,V values of the input signal are outside the valid range to change them until they are within the valid range.

   The preceding claims have been superseded by the following claims:1. A method of limiting Y, R-Y and B-Y component colour video signals (Y,U,V signals) such that corresponding RGB components lie within specified limits, comprising the steps of: defining a twodimensional range of valid U,V values for values of constant luminance, comparing an input signal with the range of valid U,V values appropriate to its luminance value and if the U,V values of the input signal are outside the valid range changing them until they are within the valid range.
2. 2. A method according to Claim 1, in which the comparing step comprises comparing the input signal with maximum and minimum permitted. values of U and V respectively for that luminance.
3. 3. A method according to claim 2, in which the comparing step additionally comprises comparing the input signal with one (U,V) function appropriate to that luminance.
4. 4. A method according to Claim 1, 2 or 3, in which the changing of invalid U,V values is achieved by multiplying U and V by the same attenuating factor to retain the hue represented by the input signal.
5. 5. A method according to Claim 4, in which the attenuating factor is determined as the smallest positive values of (i) Vmax/Vin or Vmin/Vin where Vmax and Vmin are the maximum and minimum values of V for a given value of luminance which would provide valid values of R, G and B, and Vin is the value of V in the input signal, (ii) Umax/Uin or Umin/Uin where Umax and Umin are the maximum and minimum values of U for a given value of luminance which would provide valid values of R, G and B, and Uin is the value of U in the input signal, and (iii) the value of K for which components Yin, K*Uin and K*Vin would provide RGB components with G=0 or G=l.
6. 6. Apparatus for use in the method of Claim 1, comprising means defining a two-dimensional range of valid U,V values for values of constant luminance, means for comparing an input signal with the range of valid U,V values appropriate to its luminance values, and means dependent upon the result of the comparison operative if the U,V values of the input signal are outside the valid range to change them until they are within the valid range.
7. 7. Apparatus according to claim 6 in which the comparing means compares the input signal with the maximum and minimum permitted values of U and V respectively for that luminance.
8. 8. Apparatus according to claim 7 in which the comparing means additionally compares the input signal with one (U,V) junction appropriate to that luminance.
9. 9. Apparatus according to claims 6, 7 or 8 in which the means to change invalid U and V values includes means to multiply U and V by the same attentuating factor to retain the hue represented by the input signal.
10. 10. Apparatus according to claim 9 which includes means to determine the attenuating factor as the smallest possible value of: (i) Vmax/Vmin or Vmin/Vin where Vmax and Vmin are the maximum and minimum values of V for a given value of luminance which would provide valid values of R, G and B and Vin is the value of V in the input signal; (ii) Umax/Uin or Umin/Uin where Umax and Umin are the maximum and minimum values of U for a given value of luminance which would provide valid values of R, G and B, and Uin is the value of U in the input signal; and (iii) the value of K for which components Yin, K*Uin and K*Vin would provide RGB components with G=0 or G=1.
11. 11. A method of limiting Y, R-Y, and B-Y component colour video signals substantially as herein described with reference to the drawings.
12. 12. Apparatus for limiting Y, R-Y, and B-Y component colour video signals substantially as herein described with reference to the drawings.