GB2191359A - A recursive method for identifying isotropic zones in a video image; a device for detecting movement and noise in a sequence of images - Google Patents

Abstract

The method enables zones to be defined in a television image, each zone consisting of a group of contiguous points in the image that have the same property, that of being in movement for example. The zones are numbered Z1, Z2....and each has a "state vector" whose components represent characteristics of the zone, e.g. the number of points contained in the zone. The assignment of a point to a zone is made as a function of the zone numbers of the neighbouring point which precedes it in the same line and of the neighbouring point in the preceding line. If these two neighbouring points had previously been assigned to two different zones, these two zones are deemed to be united, i.e. it is recognised that they are really one and the same zone. The device assigns a zone number to each point in movement. The points whose movement is due to noise are detected by the fact that they belong to zones of very small size.

Description

SPECIFICATION A recursive method for characterizing isotropic zones in a video image; a device for detecting movement and noise in a sequence of images This invention relates to a recursive method for characterizing isotropic zones in a video image; and also to a device for detecting movement and noise in a sequence of images.

An isotropic zone in a video image is a set of connected points all with a given property, that of being in movementfor example. Each zone can be characterized by a number and a state vectorcontain- ing several characteristic components whose nature is chosen as a function ofthe processing to be done on thesezones. One ofthese components can be the numberofpointscontained in the zone considered.

The value of the state vector components in each zone is determined asthe image is swept in accordance with a recursive method which is applied for each point considered.

The present invention covers a method for characterizing in reattimethe isotropic zones in a digitalized videoimageinordertoobtain inrealtimedigital processing such as a detection of real time digital processing such as a detection of movement, a reduction of noize, recognition ofan animate object, an estimation of the movement of certain zones, etc.

The invention also covers a device which applies this characterizing method for detecting movement and noise in a sequence of images.

In whatfollowstthe description ofthe method in accordance with the invention takes as an example the points in a black and white television image whose property is to be moving. The device in accordance with the invention isan example of the use of the method in accordance with the invention in the case of this property.

In prior practice there are methods for processing images which enable points in movementto be detectedandthosewhosemovementisdueto noise to be discriminated. In monochromatic television, a point is saidto be moving if, from one image to the next, its brightness changes after a movement, a rotation, or a change in the lighting ofthe object represented bythe image for example. Detection of the stateof of movement of a point is done, classically, bycomparison ofthe brightness value ofthe current pointwiththatofthe homologous point in the preceding image.

The points whose movement is due to noise are eitherisolated or surrounded by a small number of other points in movement. Methods previously known for detecting points whose movement is due to noise consist in observing the state of points contained in a windowcentred on the current point. Thesurface of this window is fixed, a square of three times three points for example. As a function of the number of points in motion in the window, the current point is determined to have a true or a noise movement. Such a detection of movement is described in French patent nO 77 1 1800 for example.

The disadvantage ofthese methods is that they considerthe points in movement individually and only deal with those in the window. They do not allow forthe fact that these points in movement may be separated or connected and form wide zones. These methods are suitable for the detection of noise disturbed points which are grouped in small zones (up to four points) but not more than that because this detection would require a window with a bigger surface and complex decision criteria. Also, it is not possible to vary the window surface when the image contains more or less large zones with noise disturbed points.

The device in accordance with the invention uses the method in the invention to regroup the points in movement in zones and discriminate those which are noise disturbed in these zones, without the disadvantages due to prior practice.

In accordance with the invention, a recursive method for characterizing isotropic zones in a video image, an isotropic zone being formed of connected points with a common property P, method which comprises the following steps: -foreach point in the image the determination of a zone number value designating a zone to which this point is assigned, this value being determined as a function of the properties and zone numbervalues, Z and Z', of the zones to which they belong, at the instant at which the current point is assigned to a zone, respectively the point beside the current point and preceding it on the same line and the neighbouring point, which is homologous to the current point in the preceding line, -the characterization of each zone by a state vector the values of whose components are functions of the points assigned to the said zone, -the updating ofthevalues ofthezone numbers and ofthe values of the components ofthe state vector of the zones as points are assigned, in particular to take account offusions of connected zones which may be made when a point is simultaneously connected with two zones and hasasamepropertyasthe points in these two zones.

In accordance with the invention, a device is also covered which detects movement and noise and applies this method to process television images in real time.

The invention will be better understood and other characteristics will appearfromt he following descrip tionand the figures attached: -figure 1 shows an example of an image in which each point is given the logic value 1 if it is moving and the logic value 0 if it is not; -figures 2to 7 show,forthis example, different steps in the regrouping of the points in movement in the zones in a first variant ofthe method in accordance with the invention as the image is swept; -figures 8 to 12 show different steps in the regrouping of points in movement for the same image example but in a second variant of the method in accordance with the invention; -figure 13 shows a general block diagram of an example of construction of the device in accordance with the invention;; -figure 14shows a more detailed block diagram of a part of this example of construction; -figure 15 shows the time diagrams ofthe clocks used in this example of construction; -figure 16 shows a block diagram ofa part ofthe example of construction shown in figure 14; -figures 17 and 18 show an example of processing done on an image by the device an example ofthe construction of which is shown in figure 19; -figure 19 shows a block diagram of a partofthe examples of construction shown in figure 13.

The following description will cover the method which is the subject of the invention in the case in which the characteristic property of the points in the isotropic zones isto be in movement. Only two properties are possible for a point: to be in movement orto be still. It is an example which in no way limits the extent of the invention.

In thefirststep, a value of 1 orO is assigned to each point in the image as a function of whether it has the property considered or not. A point is considered as being in movement if the digital value of its luminance has varied with respecttothatofthe preceding image; it then has the logic value 1, otherwise, it hasthevalue 0 (fig. 1). In the second step, the image is processed point by point, each point being assigned to a zone, to regroup in the same zone points in movement which are connected. For each pointthere may be creation, extension or disappearance of a zone. A disappearance occurs when two zones merge becausetheyare connected. When each point is processed,the number and state of the zones is updated; hence, the method is recursive.

Each zone is identified by a nu#bervalue, Z, and characterized by a state vector, E i, one of whose component is the number Nj of points contained in the zone Zi. The state vector enables various digital processes to be done. Detection of movement and noise reduction use more particularly the component Nj.

If the point considered is in movement, it is assigned to a zonethevalue of whose mumberis afunction of the zone number values ofneighbouring points.

As the zone numbervalues are constantly updated, in the description below the zone number value assigned to a pointatthe instant at which itis processed and the updated value ofthis zone number will be distinguished.

In whatfollows, the point nearto the current point and preceding iron the same line will be called the "point on the left" and the homologous point near to the current point and in the line immediately preceding will be called the "point above". These designa tions correspond to the most common case in which the image analysis is done from top to bottom and from leftto right. However, the method in accordance with the invention is not limited to this type of sweep.

Letz be the zone number value assigned to the current point, Z' the updated value ofthe zone numberforthe point above, and Zthe updated value of the zone numberforthe point on the left. The state value characterizing each zone is stored in a memory, called the state memory, which is addressed by the zone numbervalues. During the zone updating recursive method described below connected zone fusions take place. In the case in which a zone Zj absorbs a zone Zh, the resting zone is characterized by a single state vector Ej, stored in the state memory at the address Zj.

For the zone number Zh to refer to this state vector also, a memory, called the address memory, contains Zj at the address Zh Hence, the address memory enables indirect addressing of the state memory. In the same way, for any zone with numberz, which is not absorbed by another zone, the address memory contains the address zofthe state memory at its addressz.

By convention the points which do not possess the property considered, that of being in movement, all havezeroastheirzonenumbereventhoughthey cannot always be grouped in a continuous zone.

The assignment of each point to a zone and the updating ofthe state vector which is characteristic of this zone can be done in accordance with different rules. There are two main variants ofthe method in accordance with the invention, which depend on the propagation ofthezone numbers from point to point during the image sweep. The first is called "fixed propagation" and the second "adaptative prnpaga tion". Each of these main variants contains a simplified variant.

The description below considers firstthe main variant called "fixed propagation" and is given with reference to figures 2 to 7.

To assign a point in movementto a zone, four cases areto be distinguished: 1st case: The point on the left of the current point and the point above it are still. As the current point is not connected to any existing zone, it is assigned to a new zone with number Z in which Zi is the address of a free memory unit in the address memory. The number of points in the zone is Na = 1.

Z Ziand E 1

z- A The numberz= Z; is written in the address memory atthe address Zi. The number of points, Na = 1, is written in the state memory at the address 4. This case is found in the second line, fourth column in the example in figure 1:

xi y 2nd case: The point on the left of the current point is in movement and the point above it is fixed. The current point is then connected to a point belonging to a zone previously defined and it is then assigned to this zone. The number of points contained in this zone, N, is increased by one. Its updated value is: Na =N+1

F z = o = > TFl =N.1 ZZ I I Z,O N a This case is found in the fifth column, second line of the example in figure 1:

Ar 3rd case: a) if the two neighbouring points are in movement and belong to two different zones they have a common point, the current point. The zone ofthe neighbouring point on the left, Z, absorbs the connecting zone above, number Z', containing N' points, and the common point is assigned to the zone Z. At the address Z', the address memory then stores Z in the place of Z'.

Also, Z replaces Z' everywhere where this value is stored in the address memory.

Atthe address Z, in the state memory, the state vector is updated to represent the zone resulting of the fusion of the two zones.

@ zero Z absorbs Z' Island Z z#o == > z=z ~ & Na N N + I Such a case is found, for example, in the third line, fourth column ofthe example in figures 1 and 2.

n I~2 I SA= Ir A The zone ofnumber Z2 absorbs the zone of number Z1. The address memory undergoes the following change: address contents address contents Zl Zl - > Zl Z2 Z2 Z2 Z2 Z2 z2 The state memory undergoes the following change: address contents address contents Z1 10 > Z1 10 Z2 I N a = N N +1 Z2 12 ThenumberofpointsinZ1 (N' = 10) remainsinthe memory but is of n o fu rther use.

b) If the two neighbouring points are in movement and belong to the same zone, the current point is assigned to this zone.

and Z = z=Z=Z' Na=N'+I=N+1 This case is present in the third line,fifth column in figure 3. The point above was assigned to the zone Z1 when it was processed, but the zone Z1 was absorbed by the zoneZ2 when the second pointwas processed at the start of the third line. The address memory recorded the replacement of number Z1 by Z2 by storing Z2 atthe address Z1. Figure 3 represents the values of the zone numbers, such as they are at the instant at which the current point is processed.

Z'=Z=Z2 Z=z2 The address memory contents are unchanged.

Those of the state memory undergo the following change:

address / contents address | conrents Z2 1 12 Z2 z2 13 4th case: The point on the left of the current point is still and the point above it is in movement. The current point is connected to a zone already existing: hence it belongs to this zone. The number of points in the zone is increased by one.

1 and ZN:,Z)+I d)o# The case is found in the third line, ninth column of the example in figure 4.

This first zone which groups points in movement continues to be propagated from one pointto the next as long as the sweep finds points in movement connected to this zone. In the third line, fourteenth column in figure 4 a zone with numberZ3 is produced and, atthe next point, itabsorbsthezone ofmumber Z2. In the same way, zones with numbers Z4and Z5 are produced and absorbed not long after(figures5and 6). For each point processed there may be updating of the numbers of existing zones and, as a result, the characterizing of the zones is notfinished until all the points have been processed. In the example in Figure 1, the successive groupings of connected zones lead to a single zone Z7 (figure 7).

In case n 3, in the fusion of two zones, it was stated thatthe updating ofthe address memory was done by recording the value Z in all the memory units containing the value Z in all the memory units containing the value Z' of the number of the zone absorbed. In the example in figures 1 to 7, the final contents of the address memory are:

address. contents Zi z2 z7 z3 z7 4 z5 z7 Z8 Z7 z7 z7 Afterthe operation characterizing the isotropic zones in a whole image or part of an image, the state vector of the zones is used to perform some digital processingsouchasthediscriminationofnoise. If it is point by point processing, a memory called the standby memory, is provided to store the value of the zone number assigned to each point in an image (value not updated} while waiting for processing to start.To process an image point the first step consists in reading inthis memory the value zofthe zone number which was originally assigned. Then, the second step consists in reading in the address memory the value ofthe number ofthe zone which finally absorbed the zone of number z, perhaps after a series of absorptions. In the example in figure 1,fora point in movement in the second line, the value of the zone number originally assigned is Z1. As this zone Z1 was absorbed by zones Z2, Z3, Z4, ZSR 4 and Z7 in succession, the address memory Z7 correspond to Z1 and, as a result, the point considered is processed as belonging to zone Z7. The state of this zone is given by the state memory at the address Z7.

The updating ofthe address memory, like that of the state memory, is done while the points are processed. The first updating process may be the following: Atthe instant at which a knowledge ofthe updated valueZ' of the zone numberofthe point above is required, i.e. atthe instant at which the current point is to be assigned to a zone, the following operations are done: -The zone numberZ' assigned to the point above is read in the standby memory. For example, Z' = Z1.

-The contents, noted ADR (Z1), ofthe address Z1 are read in the address memory. For example,the address memory contains the following:

address contents Z1 Z2 ADR(Z1)=Z2 Z2 z3 Z3 z4 4 Zq z4 -The contents, noted ADR (Z2), of the address Z2 are read in the address memory ADR (Z2)=Z3 -The contents, noted ADR (Z3), of the address Z3 are read in the address memory.

ADR (Z3)=Z4 -The contents, noted ADR (4), of the address Z4 are read in the address memory.

ADR (Z4)=Z4 As4 Zq is recorded at the address 4, the zone with numberZ4was not absorbed by another zone. It is therefore unnecessary to continue the process.

-4 is recorded atthe address Z1 in the address memory.

After this updating, the address memory contents are:

address i contents Z7 z4 Z2 z3 Z3 z4 4 z4 The longer series of absorptions, the more complex is the updating of the address memory.

Evidently, this updating only takes into account absorptions which occur between the instant at which the point above was assigned to the zone Z and the instant at which the current point is considered. Subsequently, other updatings may be made at the address Z1. The contents ofthe addresses, Z2 Z3 and Z4, are updated later if the points above the current points considered later belong to these zones.

In practice it is not necessary to make a test to detectwhetherthe contents are equal to their address and then stop the series of readings. Experience showsthattherearehardlyevermorethantwo successive fusions. It suffices then to make three readings systematically before writing the number obtained. Also, as this series of operations is relatively long, it is necessaryto anticipate the updating by a time 3T in which T corresponds to the time to sweep a point. Furthermore, to gain time, thethree readings are not made in series in the same memory by using a buffer memory to store an address value between two readings: they are made in the address memory and in two auxiliary memories which are identical to the address memory. At the beginning, these three memories have the same contents and are updated indentically at the end.The second method for updating maybethefollowing: Let tbe the instant at which the current point is assigned to a zone. To update the address memory and the two auxiliary memories in the previous example, the operations are: -The contents, noted ADR1 (Z1), of the address Z1 are read in the first auxiliary memoryADR1 (Z1)=Z2at the instant t- 3T.

-The contents, noted ADR2 (Z1 ), of the address ADR1 (Z1) are read in the second auxiliary memory ADR2 (ADR1 (Zl))=ADR2 (Z2) atthe instant t-2T.

-The contents, noted ADR3 (ADR1 (Z1)), of the address ADR2 (ADR1 (Z1))=Z3 are read in the address memoryADR3(Z3)=Z4attheinstantt-T.

-ADR3 (ADR2 (ADR1 (Z,)))=Z4isrecorded atthe address Z1 in the three memories atthe instant t.

Thus, atthe instant t, the updated value ofthe zone number of the point above the current point is known and it is possible to process the current point. This example does not limit the scope of the invention. It is within the possibilities of the professional to make a large numlderof readings in the address memory and in the auxiliary and in the memories and distribute them over a longertime interval.

However, thisgeneral updating is not necessary for all appl ications.A simplification of the first variant of the method in accordance with the invention consists in only updating at the address of the zone absorbed and in not doing itforthe other addresses containing this value of the number.

In the example in figure 1, the final configuration of the address memoryisthen:

address contents Z2 Z2 z4 Z3 z4 Z4 Ze 4 4 Z7 Z7 z7 For example, the address Z1 refers to the zone numberZ2 instead of to zone number Z7, Zone Z2 after absorbing Z1, has 16 points whereas the addition of all the zones in Z7 contains many more. This is an unnacceptable errorfor applications concerning extensive zones.However, if the application only concerns zones extending over two lines atthe most, it is possibleto limitthe updating in this way. The use of the process is then simplified. This variant of the process in accordance with the invention is used in a device, which is the second subject of the invention and which enables discrimination between the points whose movement is due to noise and the points really in movement. As the interference zones contain statistically a small number of points and extend seldom over more than two lines, this simplified variant can be used.

The process in accordance with the invention contains a second main variant, called "adaptive propagation", which is described below with refer i ence to figures 8 to 12 and in which the rules assigning a pointto a zone are more complexforthe purpose of reducing as much as possible the series of absorptions of one zone by another and thus simplifying the updating of the address memory. In this variant, when two zones have a common point, the zone of the point above has priority in absorbing the zone of the point on the left and thus absorbs a zone produced more recently in general. However, this absorption is only authorized if the zone of the point on the left has not previously absorbed azone. If it has, it is the zone of the point on the left which absorbs the zone of the point above as in the first main variant. This alternative is given by the expression "adaptive propagation" of the zone numbers. To indicate whether a zone has already absorbed at least one zone, an absorption indicator bit is added in each state vector which is characteristic of a zone. For example,forthe zone of numberZ, the value ofthis bit is Tz = f if there has already been an absorption and Tz = 0 if there has not.

The assignment ofthe current pointto a zone is made in accordance with the same rules as those described for the fi rst va ria nt wittr "fixed propagation" in the first, second and fourth cases. In the third case, the adaptive rules are as follows:

a) E Tz=O TZEO zZ' Z absorbs Z Na=Nz+Nzg+J Tz,=l

ThezonesofnumbersZandZ'areseparateand have a common point which makes them connected.

As the zone of the point on the left has never absorbed anotherzone (Tz=0), it can be absorbed by the zone ofthe point above and its indicator bit becomes 1.

ífthe example shown in figure 1 is processed in accordance with these rules, this case is found in the third line,fourth column (figure 8) and then in the fourth line, second column (figure 9) and again in the fifth line,tenth column in figure 11.

b) 1 Zt * Z Z Zabsorbsr N I Tz=l Nzz=Z The zone of the point on the left has previously absorbed a zone (Tz= 1). Hence, it is not absorbed and this avoids a series of updatings in the address memory. The zone ofthe point on the left absorbs the zoneofthe point above no matterwhatthe state of the latter's absorption indicator. If this indicator shows thatthe zone of the point above has previously absorbed a zone, the series of updatings required in the address memory is done in the way described for the first main variant of the method in accordance with the invention.

This case is found in the fourth line, fourteenth column in figure 10. Zone Z1, which cannot be absorbed, absorbs zone Z3. In the same way, in the sixth line, sixth column in figure 11, zone Z1, which cannot be absorbed, absorbs zone Z5.

Finally, in this example, all the points in movement are grouped in zone Z1 (figure 12)

c) g r=z z - Z - lZI~Z I N =Nz+l=Nz,+l This case does not presentany special difficulty.

As can be seen in the example of application ofthis variant, there are no, orfew, series of zone absorptions. The new zones (Z2, Z3, etc...), which in fact are connected to Z1, are rapidly absorbed by it.

Thesimplificationofaddress memory managemenu has as,itscounterpart a larger number of logic operationsto be done for each image point, since the vatuesTzandTzP have to betested, and an increase in the volume of data to be stored in the state vectors, since a bit is added to them.

Forthe second variant, "adaptive propagation", as forthe first variant, "fixed propagation", ofthe method in accordance with the invention, there it is simplified variant which consists in not updating the address memory. In accordance with this simplified variant, in the third case, section b, of the description above, the zone ofthe point on the left absorbs the zone ofthe point above only if that does not result in updating oftheaddress memory, i.e. ifthezoneofthe point above has not already absorbed a zone (Tz'=0).

If it has, absorption is not allowed. The current point is assigned to the zone of the point on the left which continues to extend. The two zones are then considered as disconnected although in fact they are connected and ought to have been united. This error in zone characterizing is nottroublesome in some applications. It is then an advantage notto have to do updating in the address memory.

The method in accordance with the invention enables isotropic zones to be characterized in which the points in each zone have a common property P, without being limited to the property of movement and without being limited to a single given property.

For example, if the image is formed of points which may have 255 brightness values and a black background, these points can be regrouped in a zone of constant brightness. In this last case, the number of properties is 256 instead of 2. Each point is then characterized by the value of an 8 bit word instead of a word with a single bit and the state vectorfor each zone also contains an 8 bitword giving the property ofthe points in the zone. The assignment of a pointto a zone connected to it only occurs if the point has the same property as the points in the connected zone.

The state vector can contain other characteristic components such as the coordinates of the barycentre of all the points in a zone or the coordinates ofthe barycentre of points placed on the contours.

The simplified version ofthefirst main variant of the method in accordance with the invention is used in the device in accordance with the invention to group in zones the points in movementwhich are connected in order to discriminate those whose movement is dueto noise and thosewhose movement is true. This processing can be done in real time on a sequence of television images.

To discriminate the points really in movement and the points whose movement is probably due to noise it suffices to considerthe number of points in each zone. lfthe number is small, less than 8 for example, the zone is considered as a noise zone. To discriminate the noise, it is not necessary to delaythe transmission of data byan image, itsufficesto delay it by a number of lines p, fourforexample, to detect the zones which remain small.

In what follows, the pointwith the index non linej, which is being assigned to a zone, will be called the "current point" and the point with the index non line - p, whose movement must be determined as "true" or"duetonoise",willbecalledthe"pointtested".

Figure 1 3 shows the block diagram of an example of a device in accordance with the invention.

An inputterminal 1 receives,with a period T, the digital value ofthe brightness of a point in an image, analyzed in accordance with the classical methods of television. This value is recorded, on the one hand, in a memory 2, called the image memory, which restores itwith a delay of an image. On the other hand, it is recorded by a memory3, called a point memory, which only stores the brightness value of a point. These two memories receives a clock signal HP9 of period T.

Atthe output of memories 2 and 3, the digital values of the brightness oftwo homologous points in two successive images are supplied to a comparator 4. If the two values differ, the point is said to be in movement. To prevent a part ofthe errors due to noise, comparator4 has a threshold. It only operates for a difference in value greaterthan the usual noise values.

Control means, 15, generate clock signals, H P1 HP9, and control signals, Cl ..., C8,forall the organs of the device in accordance with the invention. These control means, 15, act as a function of three logic signals, MJ, MJ 1 and My 1 which symbolize the movement ofthe current point,the movement ofthe point on the left of the current point and the movement of the point above the current point respectively. Theirvalue is 1, for example, when there is movement.It is generated as follows: The output of comparator 4 supplies the signal MJ to the input of a memory 13 which delays the signal by a period T and supplies the signal MJ -1 at its output Amemory 14delaysthesignai MJ 1 atimeequivalent to a line and supplies the signal MJ 1 Memories 13 and 14 receive a clock signal HP1 of period The data are shifted one stage in each period.

The number of each zone isformed byan address which refers to a memory. that is why, in what follows,theterms"zonenumber"and"address"are used indifferently.

A device, 5, called the address and state generator, assigns to each point, whether in movement or not, a zone number and determines the state of each zone at each instant. In the application considered, the state vector of a zone only contains one component which is the number of points contained in the zone. Device 5 has two input terminals 27 and 28, and two output terminals, 30 and 31.

The state of each zone is stored in a memory, 6, called the state memory. As was explained in the description ofthe method, the state memory is addressed through an address memory, 7. An address input of memory 7 is connected to the output of a switch 9, a data input is connected to a terminal 31, a data output is connected to a terminal 28, a write instruction input receives a clock signal HP5,thefirst read instruction input receives a clock signal HP3 and the second read instruction input receives a clock signal HP7. Switch 9 has a first input connected to the output of a switch 11, a second input connected to the first output'of a memory 8 and a control input receiving a logic signal Cl.Switch 11 has a first input connected to terminal 31, a second input connected to the second output of memory 8 and a control input receiving a logic signal C8. Memory8 receives the data supplied by terminal 31 and returns them at the firstoutputwith a delaycorrespondingto p lines on the one hand and atthe second output with a delay corresponding to one line. Memory 8 has a control input which receives the clock signal HP1 The state memory 6 has an address input connected to the output of a switch 10 with two inputs. Its first input is connected to the data output of address memory 7 and its second input is connected to output terminal 31 of device 5. The position of switch 10 is controlled by a logic signal C7 applied to a control input.State memory 6 has a data input connected to output terminal 30 of device 5, a write instruction input which receives a clock signal HP6, a first read instruction inputwhich receives a clock signal HP4, a second read instruction input which receives a clocksignai HP8 and an output which is connected to input terminal 27 of device 5 on the one hand and to an input of a register 40 on the other. This register has a control input which receives the clock signal HP9 and an output connected to an input of a threshold comparator4l,which supplies a logicsignalto an output terminal 43 forming the output of the device in accordance with the invention through filtering means 42.

There are two phases in processing: -Thefirst consists in processing the current point.

The purpose isto generate a zone numbervalue,ADn, assigned to the current point, to store it in memory 8 with a delay of p lines, which forms a standby memory, and to update the contents of address memory7 and state memory 6.

-The second processing phase concerns the point tested. The zone value number, ADin-P, supplied by memory8 with a delay of p lines is switched, through switch 9, to the address input of memory 7. This memory addresses state memory 6 through switch 10. The number Nt of points contained in the zone of the pointtested is read in this memory. During the delay of p lines, the number of points in this zone increases if it is not a noise zone. This number of points is recorded in memory 40 thanks to validation by the clock HP9 and is applied to comparator41, which generates a logic signal MVJ P, which is equal to zero for example when the zone contains a number of points less than a fixed value h.This signal MVi P is applied to the filtering means 42, which neutralize the detection of the fixed points isolated in a zone of points in movement. The output of means 42 feeds the output terminal 43 of the device in accordance with the invention with a logic signal MVFi P,which is equal to 1, for exmple, if the point tested, after filtering, has a movement which is not due to noise.

In order notto detect the movements due to noise, parameters p and hwill be chosen with advantage such that: 1 Nzh Nz3p.

For example, h = 8 when p = 4.

Figure 1 4 shows the detailed block diagram ofthe address and state generator device 5. Its input terminal 27 receives a value read in state memory 6 and passes it to the input of a register 35. Input terminal 28 receives a value read in address memory 7 and passes it to the in put of a register 34. Registers 34 and 35 are buffer registers controlled by clock signals HP4 and HP5 respectively. Outputterminal 30 of device 5 is connected to the output D of a switch 26 and output terminal 31 is connected to an output C of switch 26. These two output terminals supply an updated value Na ofthe number of points in the zone to which the current point is assigned and the value ADA ofthe number ofthis zone respectively.

The operation of device 5 is described below by considering the operation of switch 26 and a switch 22, which arethefundamental organs of device Sin the example of production considered. Switch 26 is equivalenttoa switch with two circuits and three positions. In the first position, output C is connected to an input land output D is conected to an input o. In the second position, output C is connected to an input uand output Disconnected to an inputq. In the third position, output C is connected to an input n and output Disconnected to an input r. These three positions are selected by a control signal C5.Input n receives the value 0 permanently; input q receives the value 1 permanently; input rreceivesthevalue 0 permanently. Switch 22 is also equivalent to a switch with two circuits and three positions. In the first position, an output A is connected to an input a and an output B is connected to an input d. In the second position, output A is connected to an input band output B is connected to an input e. In the third position, outputA is connected to an input cand output B is connected to an input These three positions are selected by a control signal C4, which is supplied by control means 15, and by a logic signal suplied by a comparator 19.

Two registers, 16 and 17, have inputs which are connected to outputs C and D respectively of switch 26 and control inputs which receive clock signal HP1.

The output of register 16 supplies a signal of value ADE to inputs a and b of switch 22 and to the first input of comparator 19. The output of register 17 supplies a signal of value NEto input dofswitch 22 on the one hand and to the first input of an adder 18 on the other.

The output of register 34 supplies a signal of af va value ADE, to input coy switch 22 on the one hand and to the second input of comparator #9 and the input of a gate 20 on the other. The output of register 35 supplies a signal of value NE' to the second input of adder 18 and input fof switch 22 The output of adder 18 is connected to input e of switch 22. The output of comparator 19 is connected, as indicated above, to a control input of switch 22 and also to the first control input of gate 20. Gate 20 has a second control input which receives a control signal C2.The output of gate 20 feeds, through terminal 32, one inputofa g enerato r 250f addresses availa ble. Generator 25 has two control inputswhich receive a control signal C3 through input terminal 24and a control signal C6 through inputterminal 36 respectively. The output of generator 25 of available addresses feeds input m of switch 26 through output terminal 33. Input I of switch 26 is connected to output A of switch 22 and input o of switch 26 is connected to the output of an adder 23.

Adder 23 has an input connected to output B of switch 22. It adds one unittothevalue supplied by output B.

The description below considers the operation of the device in accordance with the invention during the processing ofthe current point and then during the processing ofthe pointtested. Then it explains the timing of the clocks which synchronize these processings.

In all cases including the one in which the current point is fixed, the following operations are done: -On instruction from clock HP1 the data are shifted in memories 8,13 and 14(figure 13)and registers 16 and 17 record the numberAD# and the number of points Na in the zone to which the preceding current point was assigned. They form the zone number and the number of points in the zone of the point on the leitforthe next current point -Control means 15 generate control signals C4, C5, Cl,C7andC8whichsetswitches22,26,9,lOandll respectively as a function ofthe case in hand and in synchronism with the clock signals.

-Clock signals HP3 and HP4cause a reading in address memory 7 and state memory 6 respectively.

-Clock signals HPS and HP6 cause a writing in address memory 7 and state memory 6 respectively to update their contents.

- Clock signals HP7 and HP8 cause a reading ofthe updated zone number and of the number of points N, in the zone of the pointtested.

-Afterthe processing of the current point and the point tested the data of memories 2 and 3 (figure 13) are shifted one level to process the next point of instruction from clock HP9.

The operation of the device in accordance with the invention is synchronized by nine clock signals HP1, and and HP9 in succession. The control signals C3, C1 and C7 only depend on the time. The control signals C4, C5, C6 and C8 depend not only on the instant considered but also on the state of the current point andtheneighbouring points.Theeightcasespossi- bleareasfollows::

M1 M1#1 M X Mi (p"oint point) the left) above Case n Operation to be carried out I I I I fuolon of two zones or extension of the zone on the left 2 1 1 O Extension of the zone on the left 3 1 0 I Extension of the zone above 4 1 0 0 creation of the new pone 5 0 0 0 assignment of the 6 0 0 1 current 7 0 1 0 point to a fictitious 8 0 I I zone (Z=O, N=O) In case n0 1, the zone of the point on the left absorbs the zone of the point above if their numbers are of different values and the current point is assigned to the zone resulting from the fusion. The current point is assigned to the zone of the point on the left if this is not so. Clock signal HP1 causes a shift of the data in memories 8, 13 and 14 and the recording in registers 1 6 and 1 7 of the zone number of the point previously processed and ofthe number of points belonging to this zone. These values form the number of zone ADE ofthe point on the left of the current point which is to be processed and the value NE of the number of points in the zone ofthe point on the left. Under the action ofclocksignal H P2, control means 15 generate a control signal C1 which places switch 9 in its first position and a control signal C8 wh ich places switch 11 in its second position. Switches 9 and 11 pass to the address input of address memory 7 a value ADn1 which the number of the zone of the point above had when the point above was assigned to it: hence, it is a value which is not updated.Clock signal HP3 causes the reading in address memory7 ofthe updated value ADE ofthezone numberofthe point above. Under the action of clock signal HP3, control means 15 generate a signal C7 which places switch lOin its the first position. Switch 10 passes the value ADE' to the address input state memory 6. Clock signal HP4 causes the reading of the value NE' ofthe number of points contained in the zone ofthe point above. The values ADE' and NE are applied to the inputs of registers 34 and 35 and are recorded in these registers undertheaction of clock signals HP4 and HPS respectively.Underthe action of clock signal HP4, control means 15 generate logic signals C4and C5 which place switch 22 in its second position and switch 26 in its first position. The values ADE and ADE are applied to the two inputs of comparator 19. The operation of the device in accordance with the invention is different as a function ofwhetherthese values are identical or different.

- If these values are different, comparator 19 supplies a logic signal which validates gate 20 to allow the reuse ofthe address ADE-, which is the number of the zone absorbed. The value ADE, is transmitted by switches 22 and 26 and forms the value AD10, which is the number of the zone to which the current point is assigned. This number istrans- mitted through terminal 31 to the data input of memory 7. Under the action of clock signal HP4, control means 15 generate logic signals C8 and C1 which place switch 11 in its second position and switch 9 in its first position respectively.The address input of address memory 7 thus receives the value AD 1 of the zone numberwhich was assigned to the point above. Clock signal HPS causes the reading of the valueADE, which forms the updated value of the zone number ofthe point above, atthe address of value ADn- Adder 18 sums the values NE and NE'. The value ofthis sum is passed through switch 22 to the input of adder 23. The value Na = NE + NE' + 1 is passed through switch 26 to outputterminal 30. This value forms the updated value ofthe number of points contained in the zone of the point on the left after absorption of the zone of the point above. Na is applied to the data input of state memory 6.Underthe action of clock signal HP5, control means 15 generate a logic signal C7 which places switch 10 in its second position. This switch then passes the updated value ADE of the zone numberofthepointabovetothe address input of state memory 6. Clock signal HP6 causes the writing ofthe updated value Na atthe address formed by the updated value ADE.

- In the case in which the values ADE and ADE are different, comparator 19 supplies a logic signal which cuts off gate 20 to prevent the use of the value ADE as an address available and forces switch 22 to its first position. The rest of the operation, in this case, is analogous to that in case n0 2 which is described below.

In case n0 2, the current point is assigned to the zone ofthe point on the left. Clock signals HP3 and HP4 cause a reading in state memory 6 and a reading in address memory 7, which is useless in this case.

Undertheactionofclocksignal HP4 control means 15 generate logic signals C1 and C8 which place switches 9 and 11 in their first position. The value ADE, which is supplied by terminal 31, is thus supplied to the address input on the one hand and to the data input on the other of address memory 7.

Underthe action of clock signal HP5, the value ADE is recorded at the address ADE, which is useless. The value NE supplied by the output of register 17 is passed to adder 23, which increases it by one unit, by switch 22. The number of points, thus updated, is passed through switch 26 and output terminal 30 to the data input of state memory 6. Under the action of clock signal Hp5, control means 15 generate a signal C7 which places switch 10 in its second position. The value ADE, which is supplied by terminal 31, is then passed to the address input of state memory 6. Under the action of clock signal HP6, the updated value Na is recorded at the address of value ADE.

In case n 3,the extension ofthe zone ofthe point above, clock signal Hp2 triggers control signals C1 and C8to place switch 9 in its first position and switch 11 in its second position respectively. Switches 11 and9passthevalueAD04 of the zone number, which was assigned to the point above, to the address input of address memory 7. Clock signal HP3 causes a reading of address memory7 atthe address of value ADi 1 and the value read, ADE, is supplied to register 34 byterminal 28.Clock signal HP4causes the recording ofAD#'in register34.AIso,thevaIueAD## is passed by switch 10 to the address input of state memory 6 as switch 10 is placed in its first position by a signal C7 triggered by clock signal HP3. Clock signal HP4 causes a reading atthe address of value ADE in state memory 6. The va lue read, NE, is passed to register 35 terminal 27 and is recorded in this register underthe action of clock signal HPS. Under the action of clock signal HP4, the control means 15 generate logic signals, C4 and C5, which place switch 22 in its third position and switch 26 in its first position.The value ADE', which is supplied by the output of register 34, is passed by switches 22 and 26 and terminal 31 to the data input of address memory7.The value AD, which is supplied bymemory8, is passed by switches 11 and 9 to the address input of address memory 7.

These switches are placed in the second and first positions respectively by signals C8 and C1, which are triggered by clock signal HP4. Clock signal HP5 causes the reading of the value ADE at the address of valueAD0'inthe address memory 7. This does not change its contents. The value NEl I which is supplied by the output of register 35, is passed by switch 22 to adder 23, which increases it by a unit, and then by switch 26 and terminal 30to the data input of state memory 6. The value ADE~, which is supplied by terminal 31, is passed to the address input of state memory 6 by switch 10, which has been placed in its second position under the instruction of a signal C7 triggered by clock signal HP5. Clock signal HP6 causesthewriting of the updated value NE' atthe address ofvalue ADE'.

In case n 4, the creation of a new zone, clock signals HP3 and HP4 cause a reading in state memory 6 and a reading in address memory 7, which is are useless in this case. Switch 26 is placed in its second position by a signal CS triggered by the clock signal HP4. The generator 25 of addresses available sup plies, at its output terminal 33, the value ADl of an address available. This value is transmitted by switch 26andterminal3l tothedatainputofaddress memory 7. Under the action of clock signal HP4, switches 11 and 9 are each placed in theirfirst position. Thus, they pass the va lue ADI to the address input of address memory 7.Clock signal HPS causes the writing of the value ADI at the address ADI in address memory 7. Output D of swich 26 supplies a value Na = 1 through outputterminal 30 to the data input of state memory 6. Switch 10, which is placed in its second position by a signal C7 triggered by clock signal HP5, passes the value AD, to the address input of state memory 6. Clock sig nal HP6 causes the writing ofthe value 1 at the address ADI.

In cases n0 5,6,7 and 8, the current point is fixed and it is assigned to a fictitious zone whose number is Oandthe numberofwhose point isO. Clock signal HP4triggers a signal C5 which places switch 26 in its third position. Clock signals HP3 and HP4cause a reading in address memory 7 and a reading in state memory 6 which are useless. Switch 11 and switch 9 are placed in theirfirst position by signals C8 and C1 respectivelyundertheaction of clock HP4. Output D of switch 26 supplies a value ADl = 0 which is passed through output terminal 31 to the data input of address memory 7 and to the address input of this memory through switches 11 and 9.Clock signal HPS causesthewriting of a value 0 at the address 0 in memory 7. That does not change its contents. Switch 10, which is placed in its second position by a signal C7 triggered by clock signal HP5, passes the value ADl = Oto the address inputof state memory 6.

Output D of switch 26 supplies a value Na = 0, which is passed by terminal 30 to the data input of state memory 6. Clock signal H P6 causes the writing of a value Oat the address 0 of this memory. That does not change its contents.

The processing ofthe point tested is independent of the processing of the current point. It is done under the action of clock signals HP7, HP8 and HP9. Under the action of clock signals HP6 and HP7, control means 15 generate signals C1 and C7 which place switch 9 in its second position and switch 10 in its first position. Memory 8 supplies at value ADn at its second output. This value is delayed by p lines and is the value, which is not updated, ofthe number of the zone to which the point tested was assigned when it was processed at the current point. The value ADi P is passed by switch 9 to the address input of address memory 7.

Clock signal HP7 causes a reading of memory 7 at the address ADn.The updated value of the zone number supplied by this reading is passed by switch 10 to the address input of state memory 6. Clock signal HP8 causes the reading in this memory of the number Nt of points contained in the zone of the point tested. This value is recorded by register 40 when it receives clock signal HP9. The value of Ntis compared with thethreshold h of comparator 41. The logic signal which the comparator supplies is filtered by filtering means 42, which filter the values corresponding to isolated fixed points. The output of means 42 feeds outputterminal 43 of the device in accordance with the invention with a logic signal which is, for example, equal to 1 for a pointwhose movement is notdueto noise.

Figure 16 shows an example of the generator 25 of addresses available. Input terminal 32 receives either an address ADE, which is available because the zone with this number had been absorbed by another one, or a zero address, because gate 20 is cut off. These data are supplied to the input of a shift memory 37, which delaysthem by p lines. The shift is controlled ; by clock signals HP5. Atthe output of this memory a comparator 38supplies a logic signal equal to one, for example, if the data are different from zero. This logic signal then validates gate 41 which enables clock signal HP6to trigger the writing ofthe data in a LIFO memory 39. In this memory, the data recorded last are those read first.The memory 37 with a delay of plines makes it possible to check that the addresses ofthe zones absorbed are really available.

In the device which is the subject of the invention the processing only covers the p lines preceding the current point. The addresses of the zones absorbed prior to the processing of these p lines are therefore usable. When there are no addresses to be reused, memory 37 recordsazerovaluewhichisthen eliminated by comparator 38 which prevents its recording in LIFO memory 39. This memory is read only when a new zone is created (case n0 4). In this case, control means 15 supply, through inputterminal 36, a control signal C6which validates a gate 45 to allow a clock pulse HP4to pass. This pulse causes the reading ofthe last data recorded in LIFO memory 39 and these data are then supplied to output terminal 33 of generator 25 of addresses available.

Atthe start of processing of an image, control means 15 generate a signal C3, which is transmitted to generator 25 by inputterminal 24 to activate LIFO memory 39 by loading a collection of addresses formed ofthe addresses in address memory 7. Also, control signal C3 resets to zero the contents of memory 37 with a delay of p lines.

Filtering means 42, which followthethreshold comparator41, are an optional improvementofthe device in accordance with the invention, which enables the logic signal MVJ P to be filtered. This signal is at level 1, for example, if the point of index non the line of index j-p is in movement. In certain applications, in thesearch forzone contours in particular, itis advantageoustoeliminatethefixed points in a zone of points in movement. Figure 17 shows such acase.

A point, 51, which is fixed and isolated, is inside a zone of points in movement, 52. Figure 18 shows zone 52 after the elimination of fixed point 51. The external countours of zone 52 are unchanged.

Figure 19 shows an example of construction of filtering device 42. In this example device 42 eliminatesfixed isolated points or pairs of fixed isolated points. Inputterminal 44 receives signal MVinP coming from comparator41. The signal is delayed a period T by a reg iste r 53 a n d then a fu rth er period T by a register 54. These two registers are control led by clock signal HP1. Signals Mv0#1 and Mn-z,which come from registers 53 and 54 respectively, feed the two inputs of an OR gate 45 whose output is connected to thefirst input of an AND gate 46 with two inputs. The output ofthe latter gate feeds the first input of an OR gate 47 whose second input is fed with the signal MVMi P3 supplied by register 54.The output of OR gate 47 is connected to a filter 48 which eliminates the transients whose length is much shorterthan T. The output of filter 48 supplies a logic signal MVFi P3, which is signal MVJ P3 filtered, to output terminal 43 of the device in accordance with the invention on the one hand to the input of a register 49 on the other.

Register 49 delays it by a time T and then passes itto the second inputAND gate 46. The clock input of register49 is fed with clock signal HP1.

The truth table below shows the operation of means 42. It is to be noted that these means do not eliminate the detection of points in movement which are isolated (fifth line of the truth table). The state of the current point being considered is in the second colu mn. The cases in which the filtering has an effect are boxed.

C t point action ni filter

VJ~P MVni 3P MV -2 MVi#P MV i-P means 42 o-4 n-3 n-2 n-I 0-3 O O O O O O 0 0 1 0 fixed points 0 o I o o not enclosed O 0 1 1 O Nofiltering O l O O O I O 1 1 Current point O I I 0 1 in movement O I I I I No filtering I 0 0 0 O Current point fixed not enclosed No filtering I O 1 i 1 I Current point I O I Ii I fixed and endorsed I O I I 1 Filtering effective l l O O l I I 0 I I Current point I I I 0 I in movement I I I 1 I No filtering On the third and fourth lines in which Mvn-4 = MVn-3 = = 0 and MVi P2= 1,thecase resembles that of a pairoof fixed enclosed points.In fact they cannot be enclosed because, if MVi P5were equal to 1, the preceding processing, concerning the filtering of MV1O#A, would have converted the value MVi P4from 0 to 1.

It is within the possibilities ofthe professional to produce filter means 42 differently. In particular, it is possible to use a ROM programmed in accordance with the truth table above.

No matterwhatthe case, even if the current point is fixed,theclocksignalsaregenerated. In certain cases,they control useless reading or writings in the memories but there is no objection to that.

Figure 15 shows the time diagram ofan example of clock signals. There are nine clocks altogetherwith a period of T, each offset by T 3 Period T is fixed by the image analysis rate. The component technology, TTL tech nology in this example, limits the maximum clock frequency. The solution adopted in this example consists in offsetting the clocks in such a way that they send three pulses during period Ton the one hand and in doing three processing operations in parallel, which a priori are sequential, on the other.During a period T, from instantt, clock pulses HP1, H P2 and HP3 trigger the first part ofthe processing of the point with index n+1 ,while clock pulses HP4,HP5 and HP6 trigger the second part of the processing of the preceding point with index n and clock pulses HP7, HP8 and HP9 trigger the third part of the processing of the point with index n- 1 and of the associated point tested. Processing of a point lasts three periods but three points are processed almost simultaneously.

The clocks are assigned in such a waythata memory cannot be triggered simultaneously bytwo clocks.

Also, control signals Cl and C8 are triggered by clock signals HP2,HP4 and HP6 which do not trigger memory 7. Control signal C7 is triggered by clock signals HP3,HPS and HP7 which do nottrigger memory 6 Thus, a switch is moved while the memory it addresses is active. Also, it is only moved Ttime units before the reading or writing in the memory it feeds. The periodT is 112 nS in the example described. Th is period corresponds to the processing of an image with 576 lines, each containing 462 useful points. The clock pulses used to process the current point with index n are filled in black in figure 15.

It can be noted'thatthe processing of the current point with index n begins atthe instanttoei and it is only at the instant t,,pthatthesignal MVF0#, which concerns the point tested with index n belonging to the line with indexj-p, is available. Also, at instant t0#1, which corresponds to the indices given on figure 14, it is the brightest ofthe the pointwith index n which isstoredin memory3(figure 13).

The invention is not limitedto the mode of construction described above. In particular, it is within the possibilities ofthe professional to use programmed logic circuits instead of wired logic circuits.

Claims (12)

1. A recursive methodforcharacterizing isotropic zones in a video image, an isotropic zone being formed of connected points with a common property P, method which comprises the following steps: -for each point in the imagethe determination of a zone numbervalue designating a zone to which this point is assigned, this value being determined as a functionofthepropertiesandzonenumbervalues,Z and Z', of the zones to which they belong, atthe instant at which the current point is assigned to a zone, respectively the point beside the current point and preceding it on the same line and the neighbouring point, which is homologous to the current point in the preceding line, -the characterization of each zone by a state vector the values of whose components are functions of the points assigned to the said zone, -the updating ofthevalues ofthezone numbers and ofthe values of the componets ofthe state vector of the zones as points are assigned, in particular to take account offusions of connected zones which may be made when a point is simultaneously connected with two zones and has a same property as the pointsinthesetwozones.

2. A method as in claim 1 which comprises the following steps also: -the storage of the values of the state vector componentsforeach zone in a m emo ry, ca I led the state memory, whose addresses form the zone numbervalues, -thestroage of the values of the zone numbers in a memory, called the state memory, whose addresses form the zone numbervalues, -thestorageofthevaluesofthezonenumbersina memory, called the address memory, which makes an updated value correspond to each value of zone number assigned to a point at any instant afterthe one at which that point was assigned to a zone, to make it possible to knowthevaluesZandZ'while taking into account the fusions which have occurred.

3. A method as in claim in which one of the state vector components of a zone is formed by the number of points assigned to that zone and which also contains the following steps: -the giving of the value 1 to the number of points in a zone when a first point is assigned to that zone, -the increasing by unity ofthe number of points in a zone when an additional point is assigned to that zone, -the giving of a value n + N' + 1 to the number of points of a zone resulting from the fusion of two zones with numbers of points N and N' respectively, the zones being connected to the current point and the said current point being assigned to the zone resulting from the fusion.

4. A method as in claim 1 in which the zone numbervalue assigned to a point is the number value: -of a new zone ifthetwo neighbouring points have not the property P, -ofthezone of the neighbouring point with the property P if only one neighbouring point has the property P, - ofthe zone to which the two neighbouring points belong if they each have the property P and updated values, Z and Z', which are identical fortheirzone numbers, - of one ofthe zones of the two neighbouring points if they each have the property P and updated values, Z and Z', which are differentfortheirzone numbers.

5. A method as in claim 4 in which, in the last case, the value, Z or Z', ofthe zone number assigned to the current point also forms the updated value of the zone number of all the points in the two zones, the two zones being united into one.

6. A method as in claim 5 in which the updated valueofthe zone number of all the points in the zone resulting from the fusion ofthetwozones is formed by the value Z ofthe zone number ofthe neighbouring point preceding the current point in the same line.

7. A method as in claim 5 in which the updated value of the zone number of all the points in the zone resulting from the fusion of the two zones is formed by the value Z' ofthezone number ofthe neighbouring point in the preceding line if the value Z of the zone number ofthe neighbouring point in the same line has never been given to a zone resulting from the fusion of two zones and is formed by the value Z ofthe zone numberofthe neighbouring point in the same line if this is not so.

8. A method as in claim 4 in which, if the current point has the property P and if its two neighbouring points have the property P and zone numbers of different values (ZqZ'), the value of the zone number assigned to the current point is: -thevalueZ'ofthezonenumberoftheneighbouring pointinthe preceding line, which also forms the value ofthe zone number resulting from the fusion of thetwo zones, if the value Z ofthe zone number ofthe neighbouring point in the same line has not previous ly been assigned to a zone resulting from the fusion of two zones.

-the values Z ofthe zone numberofthe neighbour- ing point in the same line if the preceding condition is not satisfied.The value Z also forms the value of the zone numberresultingfromthefusion ofthetwo zones is this has occurred. This only occurs if the value Z' of the zone number of the neighbouring point in the preceding line has nottoo been previously assigned to a zone resulting from the fusion of two zones.

9. A method as in claim 2 in which the value Zj of a zone number is only updated once by recording, at the address corresponding to Zj in the address memory,thevalue ofthezone number resulting from thefirstfusion of two zones containing the points to which the value Z) was assigned as zone number.

10. A method as in claim 2 in which the value Zj of a zone number is updated by recording a value Zf at the address Zj when the of number Zj is united with anotherzoneto give a resulting zone with numberZf on the one hand and is updated before being used to determine the zone number of the current point on the other by recording the value Zf in all the addresses of the address memory correspon dig to zones which have been united into the zone of numbers Zf after a fusion or several fusions in series.

11. A method as in claim 10 in which a value Zj of a zone number is updated, before being used to determine the value of the zone number of the current point, in q+1 steps by the use of q auxiliary memories 0 identical to the address memory, q being a fixed whole number greaterthan 1, by reading a value Zk+ contained atthe address Zk in the kth auxiliary memory where k=j to j+q,then by reading a value Zj+q+i atthe address of value Zj+q in the address memory and then by recording the value Zj+q+1 at the address of value in each ofthe q auxiliary memories and in the address memory.

12. Adeviceasinclaimll also containing a filtering device with an input connected to the output ofthethird means and an output to supply a logic signal when the current point has a true movement or when it is still and isolated in a zone of points with a true movement.

12. A device detecting movement and noise in a sequence of images applying the method as in claim 9 9 containing: -first means to detect the points of an image, said to be in movement, whose brightness value has changed with respectto the homologous points in the preceding image, with one output, ~second means to determine for each point in movement in succession a value of zone number and update the components of the state vector of the zone with this number,the means having afirst, a second and a third input and a first, second, third and fourth output, -a state memory with an address input, a data input and a data output coupled to the first output, second output and first input respectively of the second means, -an address memory with an address input, a data input and a data output coupled to the third output, thefourth output and the second input respectively of the second means, -a triggering device with an input connected to the output ofthe first means and an output connected to the third input of the second means to triggerthem as a function ofthe state ofthe current point, ofthe state ofthe neighbouring point which precedes it in the same line and ofthe state ofthe neighbouring point in the preceding line, the states being determined by the first means, -third means to discriminate the points with atrue movement and the points with a movement due to noise from the number of points in thezonetowhich they belong respectively, with a first input coupled to the data output ofthe address memory, a second input coupled to the data output of the state memory, a first output coupled to the address input of the address memory, a second output coupled to the address input ofthe state memory and a third output supplying a logic signal when the current point has a true movement.

13. A device as in claim 12 also containing a filtering device with an input connected to the output ofthe third means and an output to supply a logic signal when the current point has a true movement or when it is still and isolated in a zone of points with a true movement.

14. A recursive method for characterising isotropic zones in a video image substantially as herein be- fore described with reference to, and as illustrated in, the accompanying drawings.

15. A device detecting movement and noise in a sequence of images substantially as hereinbefore described with reference to, and as illustrated in, the accompanying drawings.

Amendments to the claims have been filed, and havethefollowing effect: *(a) Claims 1-15 above have been deleted or textually amended.

*(b) New ortextually amended claims have been filed asfollows: CLAIMS

1. A recursive method for characterizing isotropic zones in a video image, an isotropic zone being formed of adjoining points with a common property P and being characterized by a state vector, each stats vector having at least one component and being a function ofthe points forming the zone, consisting in:: -the storage of the values ofthe state vector components for each zone in a memory, called the state memory, whose addresses are called zone number values, -thestorageofthevaluesofthezonenumbersin a memory, called an address memory, and which stores an updated value corresponding to each value of zone number, -thestorageofafirstzonenumbervalueassigned to each point of an image in a memorycalledstand-by memory, and comprising the steps: -successively for each pointwithin each line:: the determination ofthefirstzone numbervalue designating the first zone to which this point is assigned, this value being determined as a function ofthe properties and of updated zone number values, Z and Z' respectively, of a zone containing a neighbouring point beside the current point and preceding it on the same line and of a zone containing a neighbouring pointwhich is homologoustothe current point in the preceding line, Z and Z' being read in the address memory respectively attwo addresses read in the stand-by memory at two addresses corresponding respectively to the two neighbouring points;; and the updating, in the address memory, of the zonenumbervaluesassignedtothetwoneighbour- ing points and the updating in the state memory, of the components of the state vector of the zones of the two said neighbouring points, as a function ofthese number values and of the properties of the two neighbouring points; a final zone number of each point being constituted by an updatedvalue which is stored in the address memory and which corresponds to the value ofthe firstzone numbervalue designating the first zone to which this point has been assigned, after the deter mination of a firstzone number for every point within an image.

2. A method as in claim 1 in which one ofthe state vector components of a zone is formed bythe number of points assigned to that zone and which also contains the following steps: -the giving ofthevalue 1 to the number of points in a zone when a first point is assigned to that zone, -the increasing by unity of the number of points in a zone when an additional point is assigned to that zone, -the giving of a value N + N' + 1 to the number of points of atone resulting from thefusion oftwozones wth numbers of points N and N' respectively, the zones being connected to the current point and the said current point being assigned to the zone resulting from the fusion.

3. A method as in claim 1 in which the zone I numbervalue assigned to a point is the number value: - of a new zone if the two neighbouring points have not the property P, -ofthe zone ofthe neighbouring point with the property P if only one neighbouring pointhasthe property P, - ofthe zone to which the two neighbouring points belong if they each have the property P and updated values, Z and Z', which are identical fortheirzone numbers, - of one ofthe zones ofthe two neighbouring points if they each have the property P and updated values, Z and Z', which are different fortheirzone numbers.

4. A method as in claim 3 in which, in the last case, the value, Z orZ', ofthe zone number assigned to the current point also forms the updated value of the zone number of all the points in the two zones, the two zones being united into one.

5. A method as in claim 4 in which the updated value of the zone number of all the points in the zone resulting from the fusion of the two zones is formed bythevalueZofthezonenumberoftheneighbour- ing point preceding the current point in the same line.

6. A method as in claim 4 in which the updated valueofthezone numberofallthe points inthe zone resulting from the fusion ofthe two zones is formed bythevalueZ' ofthe zone number ofthe neighbour ing point in the preceding line if the value Z of the zone number ofthe neighbouring point in the same line has never been given to a zone resulting from the fusion oftwo zones and is formed by the value Z of the zone number of the neighbouring point in the same line if this is not so.

7. A method as in claim 3 in which, if the current point hasthe property Pand if itstwo neighbouring points have the property P and zone numbers of differentvalues (Z= Z'),the value ofthe zone number assigned to the current point is: -the value Z' ofthezone number ofthe neighbour- ing pointin the preceding line,whichalsoformsthe value ofthe zone number resulting from the fusion of the two zones, if the value Z of the zone number of the neighbouring point in the same line has not previous- ly been assigned to a zone resulting from the fusion of two zones.

-the values Z of the zone number of the neighbouring point in the same line ifthe preceding condition is not satisfied. The value Z also forms the value of the zone number resulting from the fusion ofthetwo zones if this has occurred. This only occurs if the valueZ' ofthezone numberofthe neighbouring point in the preceding line has nottoo been previously assigned to a zone resulting from the fusion oftwo zones.

8. A method as in claim 1 in which the value Zj ofa zone number is only updated once by recording, at the address corresponding to Zj in the address memory,thevalueofthezonenumberresultingfr6m the firstfusion oftwo zones containing the points to which thevalue Zj was assigned as zone number.

9. A method as in claim 1 in which the value Zj of a zone number is updated a firsttime by recording a value Zf atthe address Zj of the address memory when the zone of number Zj is united with another zone to give a resulting zone with number 4 and is then updated, before being used to determine the zone number of two neighbouring points, by recording the value Zf in all the addresses of the address memory corresponding to zoneswhich have been united intothezoneofnumberzfafterafusionor several fusions in series.

10. A method as in claim 9 in which a value Zj of a zone number is updated, before being used to determine the value ofthe zone number of the current point, in qfl steps by the use of q auxiliary memories identical to the address memory, q being a fixed whole number greaterthan I, by reading a value Zk+I contained at the address Zk in the kth auxiliary memorywhereK=jtoj+q,then by reading a value Zj+q+i atthe address of value Zj+q in the address - memory and then by recording the value Zj+q+l at the address of value Zj in each ofthe q auxiliary memories and in the address memory.

11. A device detecting movement and noise in a sequence of images applying the method as in claim 9, containing: -firstmeansto detectthe points of an image, said to be in movement, whose brightness value has changed with respect to the homologous points in the preceding image, with one output; second meansto determine for each point in movement in succession a first zone numbervalue and update the zone number values and the compo nents of the state vector ofthe two zones respectively containing a neighbouring point beside the current point and preceding it on the same line and a neighbouring pointwhich is homologous to the current point in the preceding line, as a function ofthe properties and zone number values of these two neighbouring points; said means having a first, a second and control inputs and a first and a second output; -a state memory with an address input, a data input and a data output coupled to the fi rst output, second output and first input respectively of the second means, to storethevalues ofthe state vector components; -an address memory with an address input, a data input and a data output coupled to the firstoutput, the first output and the second input respectively of the second means, to store updated zone numbervalues respectively corresponding to each zone number value;; - a triggering device with an input connected to the output ofthefirst means and outputs connected to thecontrol inputs of the second meanstotrigger second means as a function of the state of the current point, of the state of the neighbouring point which precedes it in the same line and of the state of the neighbouring pointin the preceding line, the states being determined bythefirst means, -third means to discriminate between the points with atrue movementandthe points with a move menu duet noise from the number of points in the zone to which they belong respectively, with an input coupled to the data output of the state memory and an output supplying a logic signal when the current point has a true movement.