GB2184628A

GB2184628A - Video signal processing

Info

Publication number: GB2184628A
Application number: GB08702209A
Authority: GB
Inventors: Martin Weston
Original assignee: British Broadcasting Corp
Current assignee: British Broadcasting Corp
Priority date: 1983-12-13
Filing date: 1987-02-02
Publication date: 1987-06-24
Also published as: GB2184628B; GB2151431B; GB8333245D0; GB8431300D0; GB2151431A; GB8702209D0

Abstract

The formation of still video pictures from an interlaced video signal is described. Two (or more) adjacent lines of a 'first' field of the video signal are taken and an intermediate line from a 'second' field compared with them. If the amplitude of the value of the line of the second field falls between the values of the lines of the first field, then the line from the second field is used as the output. If it does not, various values may be used, e.g. (a) the value of the adjacent extremity of the range, (b) the average of the range, (c) or another value interpolated from the lines of the first field, this value tending linearly from (a) to (b) as the amount by which the range is exceeded, or (d) a value as (c) with discontinuities in slope smoothed. An improved output is obtained in which defects are minimised in dependence upon whether movement is presumed present or not. <IMAGE>

Description

SPECIFICATION Video Signal Processing This invention is concerned with the processing of video signals, for example broadcast television signals, and in particular with the forming of still video pictures from an interlaced video signal, and with the determining of whether an interlaced video signal is likely to contain movement in the scene content.

Where stationary pictures are required to be formed from moving video pictures there is a particular difficulty arising from the conventional interlace, which interlace exists in all television systems currently used for broadcasting. The existing methods of producing stationary pictures either fail to cope adequately with moving inputs or introduce unnecessary impairments even on stationary areas of the picture. The known systems are essentially of four types. These are illustrated in Figure 1 in which the vertical position of lines is plotted vertically against time horizontally. The following symbols are used: o input lines + output lines-'first' field X output lines-'second' field The terms 'first' and 'second' are merely convenient labels and may in practice be reversed.

The four known systems are: (i) Picture Repeat A picture may be frozen by repeating one pair of fields as shown in Figure 1(a). This preserves the maximum vertical definition, but any moving object is seen to flicker at the picture rate (25 or 30 Hz) between its two positions on the two fields.

(ii) Picture Average One way to remove flicker on moving objects is to average the two input fields as shown in Figure 1(b).

However, this blurs moving objects, and slightly softens vertical detail even on stationary areas.

(iii) Field Repeat All problems with moving inputs are overcome by using only one input field. This may be repeated for both output fields as shown in Figure 1(c). But this reduces vertical definition and makes the entire picture appear to hop up and down at the picture rate because the one field is displayed alternately in two positions.

(iv) Field Interpolate Picture 'hop' can be removed by generating the second field from the first by interpolation. The simplest form of interpolation is to average the two lines of the first field which lie immediateiy above and below each required output line of the second field, as shown in Figure 1(d). This removes hop but softens the picture vertically and can leave some 'vertical aliasing', which is most commonly seen as jagged sloping edges. The loss of resolution can be minimized by using more sophisticated 'higher order' interpolation, which combines signals from more than two lines. However, this has little effect on the vertical aliasing which is fundamental to the loss of information from the second field.

In all cases the lines of the selected fields are of course repeated to provide a still picture for display.

We have thus appreciated that what is best for stationary pictures is not best for moving pictures and vice versa.

Accordingly this invention provides in one aspect a method of forming still video pictures from an interlaced video signal, comprising the steps of taking two or more adjacent lines of a first field of the video signal, comparing with them an intermediate line from a second field, and determining whether the amplitude of the line from the second field lies within a range defined by the said lines of the first field, to provide a signal indicative of whether the video signal is likely to contain movement, and forming output lines repeatedly from the two fields to provide a still video signal, in which the output lines are formed differently in dependence upon whether the movement indicating signal indicates the presence of movement or not.

The invention also provides apparatus for forming still video pictures from an interlaced video signal, comprising an input for receiving the interlaced video signal, means connected to the input for defining in response to two or more adjacent lines of a first field of the input signal an amplitude range, means for comparing with the amplitude range an intermediate line from a second field and for determining whether the line from the second field lies within the amplitude range, and means for repeatedly forming output lines from the two fields of the input signal to provide a still video signal, the said forming means being operative differently in dependence upon the said determination to form the output lines differently in response thereto.

Our British Patent Specification 2,050,109A describes a high definition television receiver, in which additional video lines are interpolated either from adjacent lines in the same field or from a corresponding line in an adjacent field. At page 4 lines 31 to 37 brief mention is made of the possibility of comparing the 'adjacent field' line with the 'same field' lines, and if the 'adjacent field' value lies between the 'same field' values, then using the 'adjacent field' value as the output, but otherwise using the average of the 'same field' values as the output. These proposals are made in the context of an improved quality broadcast receiver, rather than of overcoming problems in producing still pictures from a selected picture of a continuous television signal ('freeze frame').

The invention will be described in more detail, by way of example, with reference to the drawings, in which: Figure 1 (described above) illustrates four known methods of deriving a still field from a video signal by means of diagrams showing vertical position plotted against time; Figure 2 illustrates at (a) a portion of the image where an object is moving, at (b) the resultant idealised video waveforms on different lines, and at (c) the type of waveform found in practice; Figure 3 shows inputloutput transfer functions for a still picture generator embodying the invention; Figure 4 comprises diagrams similar to Figure 1 for two arrangements embodying the invention; Figure 5 is a block circuit diagram of a simple system embodying the invention;; Figure 6 is a block circuit diagram (or alternatively a flow chart) for a more complex system embodying the invention; and Figure 7 illustrates suitable transfer functions for the non-linear circuits 78, 80 in Figure 6.

The invention will be described in the context of an electronic stills store. This replays still pictures and must be able to obtain these stills from moving inputs.

In the new method movement is detected by comparing each line of the second field with the two closest lines from the first field, i.e. those immediately above and below it. Any differences between these three lines may be due not only to movement but also to vertical detail. However, for stationary edges, the value of the signal from the middle line would normally be expected to lie somewhere between the values of the signals above and below. This is illustrated in Figure 2 which shows the waveforms which might be obtained from an objectwith one stationary and one moving edge.

Thus the picture is considered stationary provided the signal from the second field lies in the range between the values of the signals above and below from the first field. Wherever this condition is satisfied the signal from the second field is passed unchanged to the output as shown in the central region of the transfer functions of Figure 3.

Wherever the signal from the second field lies outside this range it is considered to be moving and is modified as shown in the outer regions of Figure 3.

Figure 3 illustrates four possible ways in which the second field might be modified in moving areas.

The second field is modified so that it appears to have been scanned at the same time as the first, but it still contains additional vertical detail in stationary areas.

The simplest modification is to clip the signal to the allowed range as shown in Figure 3(a). This effectively replaces lines (or parts of them) on the second field by lines (or parts of them) of the first.

This is illustrated in Figure 4(a). Moving parts of the picture thus effectively undergo a field repeat operation.

It would be betterto use a field interpolate operation on moving areas as illustrated in Figure 4(b). This could be achieved by the characteristic shown in Figure 3(b).

However, the sharp discontinuity in Figure 3(b) would distort the signals and amplify noise, so it may be better to soften the transition as shown in Figure 3(c). The characteristic of Figure 3(c) outside the 'stationary' range is a straight line from the value represented by the adjacent lines towards the average of them. The line preferably has a slope of 45 degrees. Figure 3(c) still contains discontinuities of slope which could generate visible effects on some pictures and so it may be better still to smooth the corners as shown in Figure 3(d).

The optimum characteristic would need to be determined by experiment. The characteristic of Figure 3(a) has been found to give good results on most pictures, although some test signals such as zone plates are somewhat distorted.

For regions that are detected as moving quickly the output as shown by Figure 3(b), (c) and (d) tends to the average of the adjacent pair of input lines.

More sophisticated interpolation could however be used. In one example of this, the average can be a weighted average of four lines from the adjacent field, namely +5/8 of the lines immediately above and below, and -1/8 of the lines three above and below (in the picture).

Similarly, more lines of the first field could be examined to determine the range of signals which are considered to be stationary. In general, this range would be defined as the average value interpolated from the first field plus and minus a range determined by the amount of detail in the immediate vicinity. The amount of detail would be determined by differentiation. The simplest form of differentiation is subtraction of signals from the two adjacent lines, which gives the simple system so far described. However, more sophisticated differentiation could be used and might give improved discrimination between detail and movement.

If three input fields are available, each of the outer fields may be processed using the middle field as a reference. This removes those differences from the middle field which are due to movement but hopefully leaves those differences which are due to vertical detail. The two processed fields may then be averaged to give an improved 'second' output field.

Many of the residual distortions from the processing of each of the outer field will have been cancelled out.

The techniques so far described act on a point by point basis. Unlike many other 'adaptive' systems, there is no integration of information from a finite area. It seems to work well without it, but some form of integration might bring further benefits.

The invention can be embodied in an electronic stills store using a general purpose 8-bit microprocessor to process pictures held in a semiconductor picture store. This allows the processing to be added to the final system quickly and with the minimum of additional cost. However, purpose-built hardware could be used to increase the speed and would probably be essential for many of the other applications of the new technique.

Figures 5 and 6 illustrate two possible implementations of the simplest and more sophisticated systems respectively.

Figure 5 illustrates the simplest system using only two input fields and modifying the 'second' field with the simple clipping action of Figure 3(a).

The circuit 10 has an input 12 for receiving an input video signal which is connected to a tapped delay line system 14, comprising a one-field delay 16 and a one-line delay 18 connected to the output of field delay 16. The input and output of the line delay 18 thus comprise the lines of the preceding field which are below and above (respectively) the line at the input 12. The lines of the 'preceding' field are applied to a comparator 20 which detects which is the larger signal, and uses a signal 22 dependent upon the result of this determination to control a cross-over switch assembly 24, so that the higher of the two is always applied to a terminal 26 and the lower to a terminal 28.

The higher signal is then applied to a comparator 30 which it is compared with the 'second' field signal from input 12. The output 32 from comparator 30 is used to control a switch 34 receiving the second field signal and the 'higher' signal. If the 'second' field signal is larger than the 'higher' signal, then movement is assumed to be present and the second field signal is replaced by the higher field signal.

The output of switch 34 is applied to a comparator 36 which also receives the lower signal from terminal 28. If the signal from switch 34 is smaller than the 'lower' signal then this is also most likely to be due to movement, and is thus replaced by the lower reference signal from terminal 28 by a switch 38.

The circuit output 40 is thus unchanged if it lies between the signals for the two closest lines on the preceding field, but is limited to the values for those preceding lines if it is outside that range. In other words, the second field is always constrained to have a value which lies between the values of the lines from the first reference field.

Figure 6 illustrates a more complex system 50 using three input fields in the manner described above with each 'outer' field being processed using the 'middle' field as a reference. The circuit also employs higher order interpolation than simple averaging for areas of distinct movement, and uses more than two lines of the first field to determine the range of signals which are considered to be stationary. The system can implement any of the transfer functions shown in Figure 3.

The system 50 has a video input 52 connected to a tapped delay chain 54, providing a total delay of two fields, and including three one-line delays 56, 58, 60 connected in series in between two equal delays 62, 64 each of slightly less than one field. Thus, from the middle field of three up to four consecutive lines are available, and one line from each of two outer fields are available at the input of delay 62 and the output of delay 64.

From the middle field signals an interpolator 66 derives an interpolated middle field signal which is applied to subtractors 68,70 where it is subtracted from the two outer fields. A differentiator 72 also receives the middle-field signals and a measurement of detail is obtained by differentiation. This is applied to two dividers 74,76 connected to the outputs of subtractors 68,70 to produce normalised difference signals which are applied to non-linearfunction circuits 78,80 tailored to suit any of the curves of Figure 3. The outputs of circuits 78,80 each provide a movement indication as between the middle field and the two outer fields, and these are applied to multipliers 82,84 which also receive the outputs of subtractors 68, 70. The multipliers are connected to respective adders 86, 88 which receive the outer field signals.The adder outputs are combined in a halving adder 90 to provide the circuit output 92.

It will be appreciated that the circuitry may be formed in other ways than that indicated, e.g. the adding circuits 86,88, 90 could be combined. Rather more drastically, the system could be implemented on a computer, in which case Figure 6 should be considered as more in the nature of a flow chart than as a block circuit diagram.

The combination of e.g. subtractor 68, multiplier 82 and adder 86 forms a cross-fade unit which is controlled by the multiplier 82 input. For a multiplier control input of zero the 'outer' field is selected, and for control inputs of unity the interpolator output is selected. For intermediate values complementing proportions of the two signals are taken. Figure 7 illustrates the non-linear functions which are required to implement each of the four overall transfer functions of Figure 3.

The non-linearfunction circuit 78 will normally produce an output of zero for all difference signals which are much less (in magnitude) than the detail signal, the divider output then being less than unity.

The multiplier thus produces zero output and the outer field passes straight through the following adder.

Where there is movement between the two fields, the field difference signal should be larger than the detail signal and the divider should produce an output which is larger than unity. The non-linear function then gives a finite output which allows a proportion of the field difference signal through the multiplier which modifies the output field in the following adder. For large normalised differences the non-linear function will produce an output of unity which allows the field difference signal through the multiplier at full strength. In this case the following adder will consist solely of the interpolator output. The contributions from the outer field will cancel out.

In generating a still video picture for an interlaced video signal, the lines forming the two selected output fields are of course repeated for display as a still picture, as in the known systems.

Attention is drawn to our application No. 8431300 (Serial No. 2151431)outofwhichthe present application is divided.

Claims

1. A method of forming still video pictures from an interlaced video signal, comprising the steps of taking two or more adjacent lines of a first field of the video signal, comparing with them an intermediate line from a second field, and determining whether the amplitude of the line from the second field lies within a range defined by the said lines of the first field, to provide a signal indicative of whether the video signal is likely to contain movement, and forming output lines repeatedly from the two fields to provide a still video signal, in which the output lines are formed differently in dependence upon whether the movement indicating signal indicates the presence of movement or not.

2. A method according to claim 1, in which in the presence of movement the output for the second field is constrained not to exceed the said range.

3. A method according to claim 1 or 2, in which in the presence of movement the output for the second field tends towards a value interpolated from the first field.

4. A method according to claim 3, in which as the amplitude of the line from the second field moves just outside the said range the output value is substantially equal to the value of the said line of the first field defining the extremity of the range.

5. A method according to claim 3 or 4, in which there is no discontinuity in the output value as the amplitude of the line from the second field moves outside the said range.

6. A method according to claim 3,4 or 5, in which there is no discontinuity in slope of the output value as the amplitude of the line from the second field moves outside the said range.

7. Apparatus for forming still video pictures from the two fields of an interlaced video signal, comprising an input for receiving the interlaced video signal, means connected to the input for defining in response to two or more adjacent lines of a first field of the input signal an amplitude range, means for comparing with the amplitude range an intermediate line from the second field and for determining whether the line from the second field lies within the amplitude range, and means for repeatedly forming output lines from the two fields of the input signal to provide a still video signal, the said forming means being operative differently in dependence upon the said determination to form the output lines differently in response thereto.

7. Apparatus for forming still video pictures from an interlaced video signal, comprising an input for receiving the interlaced video signal, means connected to the input for defining in response to two or mcre adjacent lines of a first field of the input signal an amplitude range, means for comparing with the amplitude range an intermediate line from a second field and for determining whether the line from the second field lies within the ampitude range, and means for repeatedly forming output lines from the two fields of the input signal to provide a still video signal, the said forming means being operative differently in dependence upon the said determination to form the output lines differently in response thereto.

8. A method of forming still video pictures from an interlaced video signal substantially as herein described with reference to Figures 3 to 8 of the drawings.

9. Apparatus for forming still video pictures from an interlaced video signal substantially as herein described with reference to Figure 5 or Figure 6 of the drawings.

Amendments to the claims have been filed, and have the following effect: (a) Claims 1-7 above have been deleted or textually amended.

(b) New or textually amended claims have been filed as follows:-

1. A method of forming still video pictures from the two fields of an interlaced video signal, comprising the steps of taking two or more adjacent lines of a first field of the video signal, comparing with them an intermediate line from the second field, and determining whether the amplitude of the line from the second field lies within a range defined by the said lines of the first field, to provide a signal indicative of whether the video signal is likely to contain movement, and forming output lines repeatedly from the two fields to provide a still video signal, in which the output lines are formed differently in dependence upon whether the movement indicating signal indicates the presence of movement or not.