GB1605009A - Object detection apparatus - Google Patents

Description

(54) OBJECT DETECTION APPARATUS (71) We, MULLARD LIMITED, Abacus House, 33 Gutter Lane, London, E.C.2. a British Company, do hereby declare the invention, for which we pray that a patent may be granted to us, and the method by which it is to be performed, to be particularly described in and by the following statement: This invention relates to apparatus for processing pictures of scenes to detect small objects within such pictures. In addition to detecting small separate objects, small characteristic features of larger objects, such as corners, may also be detected. An indication that an object has been detected may be superposed on a visual version of the scene to aid human interpretation of the picture.

The objects may be detected by examination of the scene at another wavelength, for example, by emitted thermal radiation. Signals from a thermal detector scanning the scene may be converted to visual form and superposed in registration with a visual version of the objects to call the attention of a human observer to an object which is thermally emissive.

An apparatus for conducting such a thermal examination of a scene and for superposing the results on a visual version of the scene will be defined herein as a thermal pointer.

Much of the information in the picture, which may be a visual version of the scene or a visible form of a thermal version of the scene, may be irrelevant but a small proportion of the information may be very important. It is then desirable to have a method detecting objects of predetermined outline which has a high probability of object detection with an acceptably low false alarm rate.

The thermal detector provides a continuous analogue signal as it scans across the scene.

The thermal detector signal may be sampled at discrete intervals of time or of distance in the scan and so dissect the scene into picture elements. The thermal detector output may be digitized and be represented by the closest member of a predetermined set of voltage levels representing corresponding brightness levels, for example 32 levels, and a 5 bit code representing the level is then used subsequently in the signal processing equipment. A defined range of voltage levels of the thermal detector signal will be referred to herein as a temperature window. Binarized picture elements, where only '0' and 'I' are the picture element values, are not considered relevant herein.

The invention provides pictures processing apparatus for application to a picture comprising a regular two-dimensional array of picture elements each of which is the centre picture element of a plurality of radiating straight lines of contiguous picture elements aligned in radial directions and in which each picture element has a brightness value, said apparatus comprising means for determining the average brightness of one or more picture elements in such a radial direction, said average being taken over a number of picture elements small compared to the number of picture elements in the maximum dimension of the picture and excluding the said centre picture element, and means for providing an output signal when the said average brightness differs from the said centre picture element brightness by at least a predetermined threshold amount of the same sign on at least half of the said radial directions.

The said regular two-dimensional array may comprise regularly spaced rows and regularly spaced transverse columns of picture elements and the outermost picture element used for determining said average brightness in each radial direction may be contained within and be at the boundary of a rectangular array of picture elements having its sides parallel to said rows and columns respectively.

The said rectangular array of picture elements with associated signal processing apparatus will be referred to herein as an object detection operator or more simply, as an operator. It is referred to as an operator as it defines the set of picture elements involved in the operation of assessing the centre picture element brightness in relation to that of surrounding picture elements to detect an object.

The invention provides means for processing the information from the thermal detector of a thermal pointer before superposition on the visual version of the scene so as to preferentially detect those features in the scene which are man-made targets.

The purpose of using such a pointer with the invention is to find possible targets in difficult situations. Considerations of the sensitivity and range required in these limit situations are such that the image size of a detected object is only one to three times the size of the photocell. With so little information, it may then be difficult to decide on the identity of the object. This task is then left to the human observer who looks at the visible scene.

Generally, it may be said that to detect "targets' in a given signal (temporal or pictorial) is to find some characteristic sequence of events in a series of measurements on that signal.

The events not subsequently identified as targets are considered to be "noise".

The present invention although oriented towards the design of a specific system, is basically concerned with the same problems as are other target location equipments such as radar alert systems, aerialphotograph processors, star locked navigation, etc. These problems have received, and still do receive, considerable attention from scientists and engineers.

Usually a gain control is provided in the apparatus so that a similar signal is obtained from objects of large temperature difference in a highly varying scene as is obtained from objects of small temperature difference in flat scenes. Since highly varying scenes are viewed with large temperature windows it is convenient to consider the picture element brightness or temperatures in terms of quantized levels of the temperature window instead of absolute temperature units.

An example of the object detection operator provided by the invention may be considered to be built up from four segments each segment being typically five points long and oriented along one of the four principal radial directions, two orthogonal and two diagonal, in the plane of the picture as shown in Figure 2b of the accompanying drawings.

Each segment can be considered as a different filter. Only if the distribution of brightness along a segment meets a given criterion is the output of this segment "true". The overall response of the object detection operator is "true" and hence an object detected if each of the four segments gives a "true" output. The criterion for each radial direction is, as given hereinbefore, that the average brightness of the picture elements in a given radial direction, excluding the centre picture element, differs from the brightness of the centre picture element by at least a predetermined threshold amount. The difference in brightness may be of either sign. The operator may be used to detect cold objects against a warm background as well as warm objects against a cold background. The detector operates upon local differences of brightness or temperature in the scene and not on absolute brightnesses. The number of picture elements in a radial segment or direction, excluding the centre picture element, may be reduced to one in which case the average brightness reduces to the brightness of that one picture element.

Embodiments of the invention, given by way of example, will first be described in general terms with reference to Figures 1 to 4 of the accompanying drawings in which: Figure 1 shows a computer print-up of a portion of an infra-red scene.

Figure 2 shows the four segments of a 5x5 object detection operator, Figure 3 shows a schematic drawing of an asymmetrical 5 x 5 object detection operator, and Figure 4 shows the reduced corner detector (2 x 3).

Further more detailed embodiments will be described with reference to Figures 5-9 and Figures 10-12 respectively.

Referring to Figure 1, a computer print-up of a portion of an infra-red scene is shown.

The original terrain was scanned by an infra-red detector and the radiance of the scene recorded in 100 lines of 200 points or picture elements each. The radiance of each point was recorded digitally as one of 32 levels, the whole range of levels representing a predetermined temperature range or 'window' in the scene. Figure 1 shows a portion of such a scene having 21 lines of 17 points each, the digitised value of radiance from 0 to 31 being shown at each point. Thus a rectangular array of picture elements and their brightnesses is obtained.

Referring to Figure 2(a) the notation used for the picture elements surrounding a centre picture element A and comprising the operative area of the object detection operator is shown. The centre picture element is taken as a local origin of co-ordinates, positive and negative whole number ordinates and abscissae being shown. Referring to Figure 2 (b) four directions d are shown numbered and in any one direction d the positive and negative "radius" values k are shown.

An embodiment of the present invention, which can be implemented with the circuits described later with reference to Figures 10, 11 and 12, provides an object detection operator for small objects and is described in the following way.

For each direction d, two quantities are computed i.e.

p (k) being the brightness of the kth picture element in each direction and ak being a constant which depends only on the modulus of k and ak = -1 for k + 0 and ak = 2 for k = 0. and where k is a whole number function used to describe the position of the picture elements along direction d (see Figure 2), and typically will have values between +2 -2, hence defining an object detection operator which is square and has five picture elements in a side. The condition of detection, R, of a small object becomes: R = 1 if Lid and L2d 3 T for d = 1 to 4 inclusive R = 0 otherwise, T being a threshold value.

In other words, to be accepted as belonging to a small object, a point must possess a brightness which exceeds by T the mean of the two neighbours on each of the eight vectors starting from this point.

The logical conditions can be changed in order to detect hot as well as cold objects, i.e.

R = 1 if |Lld| - T # 0 for d = 1 to 4 inclusive and if IL2dI - T 3 0 R = 0 otherwise Another embodiment of the invention, which can also be implemented with the circuits described with reference to Figures 10, 11 and 12, provides an object detection operator for detecting 3 corners and small objects and can be described in a way similar to the small object detection operator. The condition of detection is: R = 1 if (Lld - T) 3 0 for d = 1 to 4 inclusive or if (L2d - T) 3 0 R = 0 otherwise In other words, for detection the centre point brightness must exceed by a threshold T the mean of that of the pairs of points along at least all those radial directions having the same sign of k. This operator can detect corners of uniform objects of any size. If the size of the object is 2 x 2 then the four corners merge into a packed quadruple.

If the condition of detection is modified so that the absolute values of Lld and L2d are considered, then the detector will detect hot as well as cold objects.

Referring to Figure 3, an asymmetrical operator is shown in which the centre picture element is a picture element adjacent, along a row, to the centre of the 5 x 5 square of picture elements. Not all the eight directions are used and the maximum radius value k along each direction is now not the same in all directions. This operator may be applied in all the four orientations in which it fits the picture element array, being rotated through 90O between applications, effectively increasing the field of view of the operator. The condition for detection of an object is that the brightness of the centre picture element must exceed by a threshold amount the average brightness of the picture elements along a direction for all of the directions used.

A reduction in size and complexity of the foregoing operators is possible by taking into account the alternating scanning motion of the pointer hereinafter described with reference to Figures 5 to 9 inclusive and as described in Patent Specification No copending Patent Application 47184/74. (Serial No. 1492179).

Provided that the vertical scans of the thermal detector remain for long enough on each horizontal bearing, the operator size can be reduced to about one half that of the operators described above. The processing of the lower half of the operator is performed during the up scan whereas the upper half of the operator is processed during the down scan. Referring to - Figure 4, the i and j coordinates of the picture elements used in each scan are given according to Figure 2 (a). In order to further simplify the required electronics, a reduction in size from 5x5 to 3 x 3 is implemented. This new detector is described as follows with reference to Figure 2.

For each direction d: Jld = p( ) - p(-1) i.e. 4 negative directions J2d = p( ) - p(l) i.e. 4 positive directions For the Up scan, R = 1 if Jld 3 T for d = 1 and 2 and 3 and J14 or J24 3 T and R = 0 otherwise.

For the Down scan, R = 1 if J2d 3 T for d = 1 and 2 and 3 and J14 or J24 3 T and R = 0 otherwise.

In other words on any one scan the brightness of the centre picture element must exceed by a predetermined threshold the brightness of each of the three picture elements in retard of the centre picture element and also the brightness of at least one of the picture elements on either side of the centre picture element.

This object detection operator offers a great simplification over the preceding ones especially in the amount of storage (or delay); only one preceding value has to be accessed on only three columns, to give the three picture elements in retard, in contrast to four values on five columns in the preceding operators.

The detection of targets either hot or cold relative to the background can be achieved by computing the absolute values of the quantities Jd introduced in the definition of the operator. Circuits for implementing this operator are described later with reference to Figures 5, 6, 7, 8 and 9.

The critical threshold, i.e. the value for which the rate of false alarms is lower than a given value (10-3 for example) is on average, dependent on the standard deviation s of the signal in the whole scene. For the 5x5 corner detector in particular, it is possible to achieve a satisfactory rate of false alarms provided that the threshold is higher than an amount given by the expression a s" with a ~ 2 1 0.5 levels of the digitised output of the thermal detectors and n = 0.5 + 0.1.

The foregoing three embodiments of the invention can be summarised as follows.

The first one uses an input data the signal from seventeen cells contained in a square 5 x 5 and, for example, it may detect and locate in the scene the presence of small objects one to three cells wide, hotter than the background by, for example, 4.5 of the temperature window of the detector. It rejects any other thermal features such as lines, large objects or edges.

The second object detection operator recognises and locates the presence of small features and also indicates corners of large objects.

Both operators can be implemented with the circuits described with reference to Figures 10, 11 and 12. They necessitate at least four stages of delay for each of five thermal detectors.

The third operator has been specifically designed to reduce the cost and size of the hardware realization. It requires only one stage of delay and a line of three photosensitive cells scanning transversely to the line of cells.

A slight modification of the operators (which can be achieved using a switch in the circuits) enables the recognition of cold as well as hot objects as compared to the local background.

A further embodiment of the invention will now be described in greater detail, by way of example, with reference to Figures 5 to 9 inclusive of the accompanying drawings in which: Figure 5 shows a schematic system block diagram of an equipment using the 3 x2 operator described with reference to Figure 4.

Figure 6 shows a practical version of a channel amplifier.

Figure 7 shows a practical version of a sample-and hold circuit.

Figure 8 shows a practical version of a set of comparators and logic circuits implementing the 3x2 operator.

Figure 9 shows a practical version of a display drive circuit.

Referring to Figure 5, a scanning device 1 scans a real image of an external natural infra-red scene 2 vertically past a horizontal row 3 of seven infra-red detectors, a, b g.

The scanning motion is oscillatory at 8 Hz, the up and down scans being of equal duration and of nearly constant speed. An apparatus for providing such a scanning motion is described in Patent Specification No. Copendmg Patent Application No.

47184/74. (Serial No. 14921/79).

Each of the detectors, a, b g has an amplifier 4, described below with reference to Figure 6. The output of each amplifier 4 is passed to direct, or "real-time" outputs al bl ......g1 g1 respectively and to "sample-and-hold" (S/H) circuits 5 described below with reference to Figure 7. The outputs a2, b2 g2 of the S/H circuits 5 provide the amplifier outputs delayed by a time corresponding to one picture element and hence provide the information for the lower row of elements of the operator of Figure 4.

Each of the five comparators 6, 7, 8, 9 and 10 implements one of the terms of the said operator. In detail, comparator 6 subtracts a2 the voltage signal of delayed picture element a and a voltage corresponding to a threshold T from bl, the real time voltage signal of picture element b. If the result Is positive the output of a high gain amplifier in comparator 6 is a voltage of defined amplitude independent of the input voltages and represents a binary digit '1'. If the result is negative, the corresponding output is a near-zero voltage and represents a binary digit '0'. Thus, the quantity: b1 - a2 - T is evaluated and converted to a binary decision. Likewise the quantities: b1 - b2 - T; b1 - c2 - T; bl - al - T and bl - cl - T are evaluated by comparators 7, 8, 9 and 10 respectively.

Logic elements 11 and 12 provided the 'AND' and 'OR' logic operations required by the operator logic described with reference to Figure 4. Thus a '1' output is only supplied along line 13 when the comparators 6, 7 and 8 all provide a '1' output and either comparator 9 or 10 provides a '1' output. These comparators and logic are described with reference to Figure 8.

A display drive circuit 14 converts the '1' output from the logic circuitry to a signal capable of driving a light emitting diode (L.E.D) at b' at a fixed brightness.

Four other corresponding sets of five comparators and a display drive circuit apply the same operator logic to the amplifier output groups b1 c1 d1 b2 c2 d2; c1 d1 e1 c2 d2 e2; c1 d1 e1 c2 d2 e2; d1 e1 f1 d2 e2 f2 and e1 f1 g1 e2 f2 g2 to provide drives for the L.E.D.'sc',d',e' and f' respectively.

The five L.E.D. visual outputs are viewed by eye through a scanner 16 driven in synchronism 17 with scanner 1. In practice scanner 1 is a mirror and scanner 16 is a small portion of the same mirror. Visual persistance reconstructs a picture having picture elements in vertical scan lines in five possible positions. By optical means not shown, this picture is superposed on a visual version of the original scene 2.

A switch 18 is actuated by the scan drive circuit 17 to switch the output of the delay drive circuits 14 from one set of L.E.D.'s b' to f' inclusive to a second set of L.E.D.'s b" to f" inclusive at the end of each scan travel of the scanner 16. Thus the reconstituted picture is displayed by L.E.D.'s b' to f' on the UP scan and by L.E.D's b" to f on the DOWN Scan.

The two sets of L.E.D.'s are separated by a distance which compensates for delays in signal processing in the equipment. The UP and DOWN displayed pictures are thus shown in registration with one another.

Referring to Figure 6, a practical version of amplifier 4 of Figure 5 is shown.

Commercially available integrated circuits C810/S1 and LA 776 are used to provide voltage gains of 300 and 6.8 respectively. A bandwidth of 1.5 Hz to 2 KHz is provided. The scan frequency of 8 Hz and the vertical picture definition combine to give 3500 picture elements per second i.e. a line pair frequency of 1750 Hz. The upper limit of bandwidth, 2 KHz, gives good pulse visibility. The overall gain of the thermal detector and amplifier and the high-frequency cut-off of all seven channels are matched to within + 2%. The photoconductive thermal detector element is connected between the terminals ain and comm.

Referring to Figure 7, a practical version of a 'sample-and-hold' circuit shown at 5 in Figure 5 is shown. The input on terminal 20 is supplied by the output of the amplifier shown in Figure 6. A switched gain control 21 passes the detector signal direct to output 22 for direct display use if reqmred. Switch 21 also feeds a buffer amplifier 23 through an A.C. coupling R1 and C1 which limits the low frequency response to 7 Hz. The matching of the 7 channels in this respect is important because a direct level comparison is later made between adjacent channels. The "black level" which enables this comparison to be made is the average signal level seen by each channel in the period determined by this coupling time constant. Due to the vertical scanning and the small horizontal subtense of the detector array, it can reasonably be assumed that all the elements of the array "see' the same mean temperature. Vertical subtense seen by each element is 20 times the angular width of the row 3 of infra-red detectors.

The output of the D.C. buffer amplifier is available in real time, 24, and is also sampled by a 'sample + hold' using a gated transconductance amplifier 25 RCA type CA 3080 A.

The sampling rate is 3.5 KHz, which is equal to the picture-element scan rate, sample pulses being applied at terminal 26 to 'enable' amplifier 25. When 'enabled', amplifier 25 charges capacitor C2 up to a voltage determined by the input 28. C2 is a low dielectric storage mica capacitor to avoid the signal due to the previous picture element having any effect on the present sample. The 'held' output is available on terminal 27 as signal a2, for example. In the following description, a suffix '1' to a channel letter indicates a real time signal and a suffix '2' to a channel letter indicates the stored signal of the previous picture element.

Referring to Figure 8 a practical version of the comparators 6, 7, 8, 9 and 10 and the logic elements 11 and 12 shown in Figure 5 is shown. Ten comparators, 30 to 39 inclusive are shown. The five even numbered comparators 30 to 38 implement the five decision terms of an operator detecting 'hot' targets against a cold background and the five odd numbered comparators 31 to 39 implement the same operator logic for 'cold' targets against a warm background. Comparator 30, for example, implements b1 - a2 - T, one of the five terms of the operator. Each comparator consists of a gated transconductance amplifier RCA type CA 3060 having two balanced voltage input terminals 40 and 41 and a single current output terminal 42. When working into a high load resistance 43, R35, the voltage gain of the amplifier is sufficiently high that signals applied to the input which are materially more than the noise level on such signals are able to drive the amplifier output to saturation giving a '1' output Qr to turn it off completely giving a '0' output. Input b1 is applied through a 10K resistor to the positive input terminal 40 and a 470K resistor connected to a bias voltage source 44. Input a2 is supplied through a 10K resistor to the negative input terminal 41 and a 470K resistor connected to the threshold voltage source T. Thus the sum of a2 and T is subtracted from b1 at the input terminals of amplifier 30. Similarly the other four amplifiers 32, 34, 36 and 38 evaluate the other four terms of the 'hot' operator. Amplifier 31 functions identically to amplifier 30 except that a2 goes to the positive input and bi goes to the negative input, thus implementing the corresponding term of the 'cold' operator. Output 46 is only a '1' when bl is less than a2 by the threshold T.

Each comparator amplifier has an enabling input: 47 for amplifier 30 and 48 for amplifier 31. An appropriate D.C. signal on 47 'enables' amplifier 30, so that the latter is able to respond to its input signals. If not so enabled, the output 42 remains at '0' regardless of input. The enabling inputs of the five 'hot' target comparators are connected together to a common input t+ and similarly for the 'cold' comparators to input t-. Thus with appropriate D.C. signals on to and t+ either one or both or neither of the 'hot' and 'cold' operators can be brought into operation.

The 'hot' and 'cold' operator outputs are combined and the subsequent logic performed by the CMOS logic circuits 49, 50, 51, 52 and 53. Circuits 49, 50 and 51 each combine 'hot' and 'cold' term outputs. For example, a '1' on either or both of outputs 42 and 46 produces a '0' on the output of circuit 49. Circuit 52 combines four inputs with the same non-exclusive inverting 'OR' logic. Circuit 53 only produces a '1' output on terminal 54 if all four inputs are '0', thus implementing the combining 'AND' logic of the operator. There are four other complete sets of ten comparators giving five processed outputs. These are needed to simultaneously process the five groups of three adjacent detector channels per group that can be obtained from the seven original channels. The threshold T may be adjustable and may be set manually by the operator or set automatically at a level giving a low false alarm rate. This may be obtained when a threshold level four times the average variation of the signal level is used.

Referring to Figure 9 a practical display drive circuit as shown at 14 in Figure 5 is shown.

A feature of this circuit is that a D-type flip-flop 60, Commercial type MC14013, is used to store the operator logic output for the duration of a picture element. The operator logic output will change between '0' and '1', in general, during a picture element as the 'real time' signal varies. To provide an acceptable display, the logic output 61 is sampled by a clock pulse 62 and stored in the D-type flip-flop 60. Output 63 remains constant at '0' or '1' for the duration of one picture element. The following logic in Figure 9 allows an alternative input to be displayed corresponding to a normal grey-scale thermal picture either alone or in combination with the operator processed picture. Suitable amplifiers and phase inverters follow to provide an output suitable for driving the light emitting diode.

The embodiment described above takes advantage of the equal speed up and down scans to break the operator into the two smaller 3 x2 operators. A single sample-and-hold circuit is required per channel. Provision is made for detecting hot or cold targets against contrasting backgrounds.

Another advantage of the scanning method is that it allows the detector bandwidth to be minimised for optimum signal-to-noise ratio and also a constant-frequency clock to be used for the operator. The performance of the equipment does not depend on where the target appears within the vertical field-of-view. Also the time delay between scanning across a target and display the presence of the target is constant (about 1 picture-element period).

This time delay can then be compensated for by a mechanical shift of the L.E.D.'s (light emitting diodes) or by using two line arrays of L.E.D.'s separated by 2 picture elements as shown in Figure 5.

A second embodiment of the invention will now be described, by way of example, with reference to Figures 10 to 12 inclusive of the accompanying drawings in which: Figure 10 shows a set of seven analogue to digital converters.

Figure 11 shows a matrix store, and Figure 12 shows vector combining logic for one channel.

The processor of the second embodiment impleme

The following sets of values and components are given by way of illustration for each of the Figures 6 to 12 inclusive.

Figure 6 Al C810/S1 A2 iiA776 V1 +5 volts V2 -5 volts C3 1 uF C4 120pF R2 100K R3 680K R4 650K R5 47K R6 47K R7 10K R8 560K Figure 7 Amplifier 23 A776 Amplifier 25 CA 3080A V3 +5 volts V4 -5 volts C5 1 F C6 120pf C7 0.1 F C8 0.1 F C9 68 F R9 6K8 R10 2K7 R11 1K R12 47K R13 100 K R14 100 K R15 100 K R16 10 K R17 560 K R18 2K2 R19 10M R20 2K2 R21 220# R22 2K7 R23 2K7 Z1 E295ZZ/01 Z2 E295ZZ/01 T1 BCY71A T2 BFR29 T3 BC109 Figure 8 Amplifier 36 1/3 CA 3060 AD Amplifier 30 1/3 CA 3060 AD Logic Circuit 52 MC 14002 Logic Circuit 49 MC 14001 Logic Circuit 50 MC 14001 Logic Circuit 51 1/4 MC 14001 Logic Circuit 53 1/4 MC 14002 R24 10K R25 470K R26 10K R27 470K R28 100K R29 lOM R30 10K R31 470K R32 10K R33 470K R34 100K R35 lOM R36 100# R37 10K R38 10K C10 0.1 uF C11 0.1 uF C12 0.1 uF Figure 9 A3 MC14010 Logic circuits 64, 65, 66 3/4 MC14011 Logic circuit 60 MC 14013 Logic circuit 67 FJH 291 D1 BAV 10 T4 BCY 71 V5 +5 volts V6 +4 volts R39 2K2 R40 10K R41 100 % Figure 10 A4 uA 776 A5 uA 776 A6 CA 3060 A7 CA 3060 L1 MC 14013 L2 MC 14013 L3 MC 14021 L4 MC 14021 Figure 11 L5 MC 14006 L6 MC 14006 L7 MC 14517 L8 MC 14517 Figure 12 L9 MC 14501 L10 MC 14002 L11 MC 14011 L12 MC 14011 L13 MC 14011

Claims (13)

WHAT WE CLAIM IS:

1. Picture processing apparatus for application to a picture comprising a regular two-dimensional array of picture elements each of which is the centre picture element of a plurality of radiating straight lines of contiguous picture elements aligned in radial directions and in which each picture element has a brightness value, said apparatus comprising means for determining the average brightness of one or more picture elements in such a radial direction, said average being taken over a number of picture elements small compared to the number of picture elements in the maximum dimension of the picture and excluding the said centre picture element, and means for providing an output signal when the said average brightness differs from the said centre picture element brightness by at least a predetermined threshold amount of the same sign on at least half of the said radial directions.

2. Apparatus as claimed in Claim 1 wherein said regular two-dimensional array comprises regular spaced rows and regular spaced transverse columns of picture elements and wherein the outermost picture element used for determining said average brightness in each radial direction is contained within and is at the boundary of a rectangular array of picture elements having its sides parallel to said rows and columns respectively.

3. Apparatus as claimed in Claim 2 wherein said rectangular array is a square having a side of five picture elements.

4. Apparatus as claimed in Claim 3 wherein the said centre picture element is at the centre of the square of picture elements.

5. Apparatus as claimed in Claim 4 wherein the output signal is obtained only when said average brightness differs from the said centre picture element brightness by at least a predetermined threshold amount of the same sign on all of the said radial directions.

6. Apparatus as claimed in Claim 4 wherein the output signal is obtained when said average brightness differs from the said centre picture element brightness by at least a predetermined threshold amount of the same sign on at least one radial direction of each of all four pairs of diametrically opposed radial directions.

7. Apparatus as claimed in Claim 3 wherein said centre picture element is adjacent, along either a row or a column, to the centre of the square of picture elements.

8. Apparatus as claimed in Claim 2 wherein said rectangular array is a square having a side of three picture elements, said centre picture element is at the centre of the square of picture elements, and said average brightness is that of one picture element adjacent to said centre picture element.

9. Apparatus as claimed in Claim 8 wherein said output signal is obtained only when said average brightness differs from the said centre picture element brightness by at least a predetermined threshold amount of the same sign on at least four contiguous radial directions.

10. Apparatus as claimed in any preceding claim comprising means for scanning the picture with a line array of radiation sensitive cells in a direction transverse to the said line of cells and means for sampling and storing the output of said cells at regular intervals so as to provide the brightness values of said regular two-dimensional array of picture elements.

11. Apparatus substantially as described with reference to Figures 1 to 4 inclusive of the accompanying drawings.

12. Apparatus substantially as described with reference to Figures 4 to 9 inclusive of the accompanying drawings.

13. Apparatus substantially described with reference to Figures 10, 11 and 12 of the accompanying drawings.