WO2004077836A1

WO2004077836A1 - Colour decoding circuitry for a video signal

Info

Publication number: WO2004077836A1
Application number: PCT/IB2004/050112
Authority: WO
Inventors: Jeroen H. C. J. Stessen; Michel W. Nieuwenhuizen
Original assignee: Koninklijke Philips Electronics N.V.
Priority date: 2003-02-26
Filing date: 2004-02-13
Publication date: 2004-09-10

Abstract

The invention is related to a receiver for a television signal, such as a PAL signal, the receiver comprising a color information retrieving unit (8, 9, 11, 13) arranged to separate color information from a composite signal to retrieve color information. The receiver contains a filter (13) for selecting the decoded color signal from a plurality of signals, under control of a first value (b) of the color information from the central linerelative to second and third values (a, c) of the color information at corresponding position on further lines at an odd number of lines from the central line on a first and second, mutually opposite side of the central line respectively. An average is selected in which the first value (b) is weighed equally with information from a line or lines at an odd number of lines from the central line, when the first value (b) is more than a predetermined factor outside a range between the second and third value (a, c). A signal derived from the color information is selected in which at least no information (a, c) is weighed from the first or second side when the first value (b) is in the range or less than said factor outside the range and closer to the third or second value (a, c) respectively.

Description

Color decoding

The invention is related to a receiver for a video signal, such as a PAL signal, the receiver comprising a color information retrieving unit arranged to separate color information from a composite signal to retrieve color information for respective lines, and an averaging unit for forming an average signal by averaging color information for a horizontal position in a central line and a first line on a first side of the central line, spaced an odd number of lines from the central line.

A composite television signal comprises both a luminance signal and color signals. The color signals are modulated as a chroma signal onto a subcarrier and added to the luminance signal to form the composite television signal. A subcarrier is included in the television signal but is suppressed in the modulation process except during bursts. In the demodulation process a subcarrier signal is retrieved from the bursts by the color information retrieving unit using a phased lock loop, which locks onto a color burst signal. The retrieved subcarrier is used to demodulate the chroma signal. This process of recovery of the subcarrier may cause a locally drifting subcarrier frequency, which in practice also is known as demodulation angle error, leading to disturbing color artifacts in the television image.

A PAL television signal receiver is known e.g. from US 5 870 153, describing a comb filter for use in a video decoder system. Known PAL systems are arranged to support correction of these color artifacts. Because the way of modulation is alternated in different lines, demodulation errors at corresponding positions in successive lines have a different sign, thereby generating a zigzagging error profile, also known as a Venetian blind profile. Supposing a constant drift in the successive lines, the demodulation error of two successive lines is eliminated by averaging the color signals at corresponding positions in the above- mentioned lines. Therefore a PAL decoder usually comprises an averaging unit for averaging color signals at corresponding positions in two successive lines. The average is displayed at the same locations as luminance information from a particular line in the television image, usually with luminance information received with the lower of the two successive lines from which the color information is averaged. The particular line will be called the "central line" herein. Note should be taken that this is not intended to limit this central line to a central position in the image.

However, th s averaging procedure results in an undesired smoothing effect as far as the vertical direction s concerned. The averaging operation in the demodulator acts as a low pass filter, thus reduc ng the vertical sharpness of the image and loosing image information. Further, the image is shifted half a line with respect to the luminance signal. Due to shifted overlap of the television signals, this may become problematic, especially if the video image is copied a number of times, viz. via receiver to DVD and then to TV, etc.

It is an object of the invention to provide a PAL television signal receiver that mitigates color artifacts caused by demodulation angle errors, with less deterioration of the vertical sharpness of the image. The invention is defined by the independent claims. The dependent claims define advantageous embodiments.

According to the invention the decoded color signal is selected from a plurality of different possible decoded signals according to the relative values of color signals at a central lines and lines on either side of the central line. If the color value on the central line is more than a factor outside a range between the values of the lines on either side, it is assumed that a Venetian blind situation has occurred and an average is used as decode color signal in which the Venetian blind effect is weighed out. If the color value on the central line is in the range, or less than a factor outside, it is ensured that color values from the side of the central line that has a color value furthest from the color value of the central line is not weighed in the decoded color signal. Thus, edge smoothing is prevented in this case.

In an embodiment a median filter is used to implement filtering. With an appropriate choice of input signals the median filter is forced to select the decoded color signal proper for the particular color signal profile in the image, depending on the values of the input signals of the median filter. When the collection of input signals of the median filter satisfies a number of rules (termed majority rules herein), it is ensured that the median filter outputs the desired signal. The majority rules are expressed by classifying the input signals of the median filter into different sets: - a central set (including for example the signal b on the central line) upper and lower average sets (including for example averages (a+b)/2, (c+b)/2 of corresponding color signals on the central line (b) and adjacent lines (a, c)) further upper and lower sets in which the central line is not weighed (including for example the signals a, c on lines adjacent to the central line). Although the different sets are distinguished in the majority rules in terms of upper and lower sets for the sake of clarity, it should be understood that these terms serves for distinction only: they do not imply lower or upper position of video lines in the image.

The majority rules ensure that the input signals of the median filter are distributed over different sets wherein lines on different sides of a central line contribute differently, so that no undesired majorities can occur that cause the median filter to select an undesired signal. The majority rules realize a divide and rule effect, which forces the median filter to select an appropriate signal. Subject to the majority rules, the input signals of the media filter need not include signals from all of the sets. In case of a Venetian blind configuration of color signals from corresponding horizontal positions on successive lines (due to sub-carrier drift), this Machiavellistic strategy ensures that the median lies among the average signals of the upper average and lower average set. This is because the signals from the central set on one hand and signals from the further upper and further lower sets on the other hand values on opposite sides of the averages and do not form a majority. Hence, the median filter selects an average signal.

In an edge configuration, where color signals on one side of the central line differ substantially from those at the central line and on the other side of it, the selection of input signals for the median filter ensures that the signals involving the one side are never a majority, thus preventing selection of averages involving those signals. Preferably averages and signals symmetrically disposed from both sides of the central line are used. Thus directional bias effects are avoided. With five signals a, (a+b)/2, b, (b+c)/2, c, applied to the median filter (b being the signal for a horizontal position from the central line, a and c being signals from corresponding horizontal positions on adjacent lines on either side of the central line) a simple and symmetric color decoding process can be realized.

In another embodiment the third set comprises a signal that exclusively represents color information for the horizontal position on the central line.

In yet another embodiment according to the invention the second set comprises a signal that exclusively represents color information for the horizontal position on the central line. This signal is easily obtained as output from the color information retrieving unit and represents a desired value in the case when an edge value is to be preserved.

In still another embodiment according to the invention the first and the third set comprise a signal exclusively representing color information for the horizontal position in the first and the second line, respectively, next to the central line. This choice makes a relatively simple implementation possible and it reduces the required number of memor - units.

In a further embodiment according to the invention the PAL receiver comprises a five-points median filter, which is preferably implemented by means of a first and a second three-points median filter. The use of the two three-points median filters avoids the employment of the more complex five-points median filter.

The invention also relates to a method of processing PAL television signals in accordance with the above-mentioned PAL television signal receiver, a computer program product for making a programmable processor perform the method and a color decoder for applying the method.

The objects and the advantages of the invention will be illustrated using exemplary embodiments shown in the drawing. In the drawings,

Fig. 1 shows a schematic scheme of a television receiver, Fig. 2 shows a receiver with 5 input median filtering,

Fig. 3 shows signal levels of image signals of pixels on corresponding positions on three successive image lines,

Fig. 4 shows a graph of the output of a median filter as a function of signal levels of pixels on corresponding positions on three successive image lines, Fig. 5 shows a schematic scheme of a 3 -points median filter in the color decoder according to the invention,

Figs. 6 and 7 show a color decoder.

In Fig. 1 a PAL television receiver 1 is shown, comprising a receiving antenna 3 which is connected to an electrical amplifier 4. The amplifier 4 outputs the signal to a first mixer 5, which in turn is connected to both a low-pass filter 6 and a band-pass filter 7. The output signal of the band-pass filter 7 is connected to a color decoder 15, comprising second and third, coherent, mixers 9,10 being driven by a coherent oscillator 8, low-pass filters 11,12 and signal filters 13,14. The filters 7, 11 and 12 together with the mixers 9,10 form the color information retrieving unit. The outputs of the mixers 9,10 are connected to low-pass filters 11,12, and subsequently to the signal filters 13,14 which output color signals U en V, respectively. It is noted that a digital implementation of the decoder is also possible. The phase difference between the oscillations at the second mixer 9 and the third mixer 10 is 90°. In alternating lines the phases rotate 180°. The oscillator frequency and phase is preferably controlled dependent on a sub-carrier phase using a known locking circuit (not shown). The PAL television receiver 1 is arranged to receive an electromagnetic television signal 2 by means of the receiving antenna 3 converting the signal 2 to an electric signal that in turn is amplified in the amplifier 4 to form a stable electric signal for further processing. The signal is demodulated using the first mixer 5 to obtain a composite PAL television signal. The composite signal passes the low-pass filter 6 to generate a luminance signal Y, comprising the low-frequency information of the composite signal. The band-pass filter 7 is applied to the composite signal to generate a signal C comprising the color information composite television signal. Quadrature modulated color signals are retrieved using mixers 9,10. The two color signals are generated each passing the respective low pass filter 11,12 and the signal filters 13,14 to form the color signals U and V, respectively.

Fig. 2 shows an embodiment of a color decoder 15. By way of example the processing of the color signal U is described. The color signal V is treated in a similar way. The output signal of the low pass filter 11 is applied to a first adder 19, a first 3 -points median filter 21, as well as a first delay element 17. The first delay element 17 generates a signal which is also applied to the 3 -points median filter 21, the first adder 19, a second delay element 18, and a second adder 20. The output signal of the second delay element 18 is applied to the second adder 20. The output signals of the first adder 19 and the second adder 20 are applied to a second 3 -points median filter 22 via factor Vι multipliers 23,24. Also the output signal of the first 3-points median filter 21 is applied to the second 3-points median filter 22. The output of the second 3-points median filter 22 represents the color signal U.

Without affecting the function of the circuit the combination of median filters 21, 22 may be replaced by a five- input median filters as a five-points median filter, which selects its third input signal ranked according to increasing amplitude. Two three-input median filters 21, 22 can be used for this, the output of the first median filter 21 being coupled to an input of the second median filter 22, because it is known that the values of the input signals of the second median filter 22, being averages of the input signals of the first median filter 21, will always between the lowest and highest of the input signals of the first median filter 21, being averages of those signals.

The delay elements 17 and 18 retard the image signal 64 μs, which is the line period, that is the time period between color signals that represent corresponding horizontal locations in successive video lines in the PAL system. Hence, the respective retarded video signals and the actual video signal represent corresponding pixels, i.e. pixels on corresponding horizontal positions in the image. Since the color input signal of the median filter representing the central line is retarded for one line period, the corresponding luminance signal Y should also be retarded for one line period to achieve coincidence of luminance and color signals, thus avoiding mismatch or shift in the luminance and color image.

Fig. 3 will be used to illustrate the operation of the circuit of Fig. 2. Fig. 3 shows a diagram of signal levels of image signals of pixels on corresponding positions of three successive image lines. The horizontal axis 31 represents the line number of the TV image, while the vertical axis 32 represents the amplitude of the image signal of a pixel at corresponding horizontal positions. Here, I, II, and III represent the line below a central line, the central line and the line above the central line, respectively.

The signal the color image signal of the pixel at corresponding horizontal positions on the lines I, II and III, will be referred to as "a", "b" and "c" respectively. In the description of the example describing the 5-points median filter, it will be assumed that a < b. (If a > b, a similar reasoning applies.) The output of the 5-points median filter depends on value ranges c relative to a and b, more particularly, as c decreases the output of the median filter changes as c passes boundary values b, ^(a+b) , a and 2a-b (where a.=^lΛ (c+b)) successively.

Fig. 4 shows the output signal of the median filter as a function of c. In Fig. 4 the horizontal axis and the vertical axis represent the image signal c of the pixel on line III, and the output of the filter, the color signal U. In table 1 the signal value subrange of the output signal are given for decreasing values of c.

Table 1: Output signal values of the median filter It is seen that the decoder is forced to switch automatically between the input signals. The decoded color-difference signal is filtered in different ways, depending on the vertical profile of the color signal, as a result of the choice of the input signals of the median filter. If a zigzag profile, also known as a Venetian profile, (a and c close to each other on the same side of b, i.e. c<b in table I) occurs, the third or fourth row of table 1 applies and an average of the central line and a neighboring line is output. If on the other hand a sharp vertical edge occurs (a and c on opposite sides of b, or b closer to a or c than the distance between a and c) no averaging process is performed. This effectively reduces the above- described smoothing effect of the averaging process. In particular, providing the five signals a, b, c, ^(a+b) and ∑ +c) as input signals to a five-points median filter construct, which selects the third signal in ranked according to increasing amplitude, leads to an averaged signal when 2a-b < c < V2(a+b). In this case the video signal exhibits a zigzag profile in the vertical direction and renders the signal of the central line extreme. The median filter inherently never selects an extreme signal value. Signal averaging reduces the Venetian or zigzag effect that is caused by demodulation errors occurring in the mixer 9 and the low-pass filter 11. However, when the vertical profile does not match a Venetian profile, but rather resembles a sharp edge (c < 2a-b or 2(a+b) < c), the signal value averaging the edge is the second value in ranked according to increasing amplitude, and the median filter does not select this averaged signal, but the signal value of the central line or the value of a neighboring line having the least difference with the signal value of the central line, thus preserving the edge in the vertical profile. This effectively enhances the vertical resolution of the TV image.

The invention is not restricted to the described 5-points median filter embodiments.

Fig. 5 shows another embodiment of a color decoder 15. The output signal of the low pass filter 11 is applied to a first adder 19, as well as a first delay element 17. The first delay element 17 generates a signal which is applied to the 3-points median filter 21, the first adder 19, a second delay element 18, and a second adder 20. The output signal of the second delay element 18 is fed into the second adder 20. The output signals of the first 19 and the second adders 20 are via multipliers 23,24 fed into the median filter 21. The output of the 3-points median filter represents the color signal U.

The three-points median filter selects the second of its input signals ranked according to increasing amplitude. The three input signals of the three-points median filter 22 are the color signal of a horizontal position or pixel on a central line to be filtered (the output signal b of the first delay element 17), the arithmetic mean of this color signal b and the signal a on a preceding line, and the arithmetic mean of the color signal b and the signal c on a following line c. In the embodiment of Fig. 5 as well, the median filter 22 is forced to choose between input signals so that in case of a Venetian blind profile an average is output and in an edge profile selection of an average of lines on opposite sides of the edge is avoided.

The invention is not restricted to the described 3-points and 5-points median filter embodiments. Many alternative embodiments using median filters with different input signals or different numbers of input signals are possible.

In general, one should distinguish between different classes of input signals of the median filter. Signals of an upper average class are averages with the color signal b from the central line in which color signals a from lines at an odd number of lines (typically 1) from the central line on a first side of the central line are averaged with a greater weight than color signals from a second side, if any (e.g. (a+b)/2). Vice versa, signals of a lower average class are averages with the color signal b in which color signals c on the second side of the central line are averaged with a greater weight than color signals from the first side, if any (e.g. (b+c)/2). In signals of a central class signals from both sides a weighted equally, or not at all (e.g. b). In this case the sizes of the classes of signals that are applied to the median filter should be so that the signals from no class by itself should form a majority of the signals. In at least one of the signals from the central class the central line b should be weighed, or make up the entire signal. As a result in the Venetian Blind configuration the output signal of the median filter must be an average from the upper or lower average class, since the central class is extreme and cannot form a majority. Thus, filtering is applied in this configuration. In the edge situation the output signal of the median filter cannot come from the upper or lower average class if that class is on the opposite side of the edge from the central line, since the upper and lower average class is a minority in that case. Thus, output of an average from the wrong side of the edge is avoided. It will be seen that the filters of Fig. 2 and 5 meet these conditions.

More generally, a further upper and lower class of signals may optionally be applied to the median filter, which are not averages, but are dominated signals a or c from the first and second side respectively. In this case, the applied signals from the average upper class and the further upper class together should not form a majority, nor should the applied signals from the average lower class and the further lower class together form a majority. This ensures that output of a wrong average from the median filter is avoided in the edge configuration. Furthermore the applied signals from neither the central class nor the further upper and lower class together should form a majority. This ensures that the median filter will output an average from the average upper or lower class in the Venetian Blind configuration (note that this does not exclude that the average upper class or average lower class may be empty). It will be seen that the filter of Fig. 2 meets these conditions.

It is noted that the described hardware functionality may be implemented using a programmable processor that performs part or all of the operations in software. Alternatively, of course a hardware implementation may be used with median filter circuits that are known per se. Starting from the present description, these embodiments are assumed to be obvious for a person skilled in the art and are considered to fall within the scope of the following claims.

Furthermore, although the invention has been described in terms of implementation with a median filter, it will be realized that the selection of the appropriate average can be implemented in other ways.

Fig. 6 shows an alternative embodiment, in which a selection unit 100 is used, which receives signals from three adjacent lines and averages of the central line with the adjacent lines. In this embodiment, selection unit 100 serves as a multiplexer, passing a signal selected by the combination of signals from the central lines. By way of example, two averaging circuits 19+23, 20+24 are shown, but it will be understood that these may be replaced by other averaging circuits, or incorporated in selection unit 100. Various forms of selections may be used. A first form of selection can be described by the following computer code: if(a<c)

{ if(b<=2*a-c) output = (b+c)/2 if(2*a-c<b<=a) output = a if(a<b<=c) output = b if(c<b<=2*c-a) output = c if(2*c-a<b) output = (b+a)/2

} if(a>=c)

{ same as above but with a and c exchanged } Here, "<=" means smaller than or equal to, and >= means larger than or equal to. Thus, an average is used when the color signal b of the central line is more than once the difference between a and c beyond a or c (which is assumed to be due to a Venetian Blind situation). Unaveraged signals are used otherwise, either a, b, c being used, "a" is used when b is outside the range between a and c, but closer to a. "c" is used when b is outside the range but closer to c. "b" is used when b is in the range between a and c.

Of course, this implementation is but one alternative. Fig. 7 shows a more general circuit, comprising a decision unit 110 and a controllable filter 112. Decision unit 110 receives the signals a, b, c and supplies a control signal to controllable filter 112 to control, dependent on the combination of relative values of a, b and c, which type of filter operation is to be used to generate an output signal. Controllable filter 112 may contain filter elements for all possible filter operations and a multiplexer for selection the output of one of the filter elements under control of decision unit 110, or a reconfigurable filter element may be used, which is reconfigured to perform a selected filter operation under control of decision unit 110. Apart from the selections described above, other solutions are possible. For example, in a second implementation selection unit may operate consistent with the following program: if(a<c) { if(b<=2*a-c) output = (a+2*b+c)/4 if(2*a-c<b<=a) output = (a+b)/2 if(a<b<=c) output = b if(c<b<=2*c-a) output = (c+b)/2 if(2*c~a<b) output = (a+2*b+c)/4 } if(a>=c)

{ same as above but with a and c exchanged }

This implementation works with the same ranges, but uses different averages (Including a double average of the two averages input to selection unit 100 in Fig. 6). In more ranges use is made of averages as output, to avoid Venetian blind effects. Furthermore the double average (a+2*b+c) is used in the extreme situations to increase reduction of the artefacts. Both improvements may be applied separately from one another. As another example, in a second implementation selection unit may operate consistent with the following program if(a<c)

{ if(b<=(3^s!-a-c)/2) output = (a+2*b+c)/4 if((3^::a-c)/2<b<=(a+c)/2) output - (a+b)/2 if((a+c)/2<b<=(3*c-a)/2) output = (b+c)/2 if((3 *c-a)/2<b) output = (a+2*b+c)/4

} if(a>=c) { same as above but with a and c exchanged }

In this example different ranges are used. If b is more than half the distance between a and c outside the range between a and c the double average (averages of the two averages (a+b)/2 and (b+c)/2, i.e. a two sided average) is used. Otherwise, either an average of a and b is used or an average of b and c, dependent on whether b is closer to a or to c respectively. This example illustrates several possibilities: (1) always using an average, so that Venetian blinds are always avoided (2) using a smaller factor outside the range between a and c to deviate from the standard Venetian blind treatment (the double average in this case) and (3) using only four different outputs. Of course all possibilities may be applied separately.

The common element of all examples is that a Venetian blind is assumed when b is more than a factor outside the range between a and c, i.e. if b is more that f*|a-c| outside the range a-c, and that it is avoided that the same average with one of the adjacent values a, c is used for all remaining values of b. If an average is used at all, it is an average with that one of a and c that is closest to b. Thus, smearing of edges is avoided.

It will be appreciated that one implementation of selection unit 100 or decision unit 110, and controllable filter 112 may use a programmed computer to execute selection according to the examples. In this case, the averages (a+b)/2, (b+c)/2 and (a+2*b+c)/4 may be computed as needed by the computer under control of instructions. In another embodiment a hardwired solution may be used (not containing a program, or containing a program for only part of the operation). In this case the "program" segments shown above should be regarded merely as a description of the effect of the hardware. The hardwired solution may use a median filter. The first example, for example, describes one of the median filter implementations described above. It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design many alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps other than those listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the device claim enumerating several means, several of these means can be embodied by one and the same item of hardware. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

Claims

CLAIMS:

1. A color decoder for determining a decoded color signal (U, V) for display with luminance information for the central line from decoded color signals for a plurality of lines, the decoder comprising a filter (13) for selecting the decoded color signal from a plurality of signals, said selecting being performed under control of a relative position of a first value (b) of the color information from the central line relative to second and third values (a, c) of the color information at corresponding position on further lines at an odd number of lines from the central line on a first and second, mutually opposite side of the central line respectively, said selecting being arranged to result in the decoded color signal becoming - an average signal in which the first value is weighed equally with information from a line or lines at an odd number of lines from the central lines, when the first value is more than a predetermined factor outside a range between the second and third values (a, c),

- a signal derived from the color information in which at least no information is weighed from the first or second side when the first value is in the range or less than said factor outside the range and closer to the third or second value (a, c) respectively.

2. A color decoder according to Claim 1, further comprising a median filter (21, 22) for selecting a decoded color signal for the central line as a median from a plurality of signals from - an average upper set of signals, each signal representing an average with color information (b) from the central line in which the further color infoπnation (a) from one or more lines on a first side of the central line is weighed more heavily than color information (c) from a second side of the central line;

- an average lower set of signals, each representing an average with color information (b) from the central line in which the further color information (c) from one or more lines on the second side of the central line is weighed more heavily than color information (a) from the first side of the central line;

- a central set of signals (b) in which the further color information from both the first and second sides is equally weighed, if at all; - a further upper and lower set of signals with the further color information (a, c) from lines on the first and second side respectively, at least one signal from the average upper set being applied to the median filter (21, 22), as well as at least one signal (b) from the central set in which the central line at least is weighed, the number of signals from the sets that are applied to the median filter (21, 22) being so that, of signals applied to the median filter (21, 22), neither the signals from the average upper set and the further upper set together, nor the signals from the average lower set and the further lower set together, nor the signals (a, c) from the further upper and lower set together nor signals from the central set (b) form a majority.

3. A color decoder according to claim 2, wherein exactly one signal from each of the sets is applied to the median filter (21, 22).

4. A color decoder according to claim 3, wherein the color signal (b) from the central line, color signals (a, c) from neighboring lines on either side of the central line, and averages of the color signal (b) from the central line and respective ones of the color signals (a, c) from the neighboring lines on either side, are applied to the median filter (21, 22).

5. A color decoder according to claim 2, wherein the central set comprises a signal (b) applied to the median filter that exclusively represents color information for the horizontal position on the central line.

6. A color decoder according to claim 2, wherein no signals from the further upper set (a) and/or further lower set (c) are applied to the median filter (20, 21).

7. A color decoder according to claim 2, wherein no signals from either the average upper set or the average lower set are applied to the median filter (21, 22).

8. A color decoder according to claim 1, wherein color information form both the first and/or second side is weighed equally with the first value in said average signal, said further signal being an average of the first value (b) and the second value (a) when the second value (a) is closer to the first value (b) than the third value (c), the further signal being an average of the first value (b) and the third value (c) when the third value (c) is closer to the first value (b) than the second value (a).

9. A method of processing video signals, comprising the steps of separating color information from a composite signal for a plurality of lines including a central line, for determining a decoded color signal (U, V) for display with luminance information for the central line; selecting the decoded color signal (U, V) from a plurality of signals, said selecting being performed under control of a relative position of a first value (b) of the color information from the central line relative to second and third values (a, c) of the color information at corresponding position on further lines at an odd number of lines from the central line on a first and second, mutually opposite side of the central line respectively, said selecting being arranged to result in the decoded color signal becoming - an average signal in which the first value is weighed equally with information from a line or lines at an odd number of lines from the central line, when the first value (b) is more than a predetermined factor outside a range between the second and third value (a, c), - a signal derived from the color information in which at least no information is weighed from the first or second side when the first value (b) is in the range or less than said factor outside the range and closer to the third or second value (a, c) respectively.

10. A receiver for a video signal, the receiver comprising: a color information retrieving unit (8, 9, 11, 13) arranged to separate color information from a composite signal for a plurality of lines (a, b, c) including a central line (b), for determining a decoded color signal (U, V) for display with luminance information (Y) for the central line (b); and a color decoder as claimed in claim 1.