FR2710433A1 - Method for recognizing objects, moved in successive images in time belonging to a sequence - Google Patents

Abstract

a) Method for recognizing objects, moved in successive images in time belonging to a sequence. b) Method characterized in that upon recognition of a displaced object, the image signal is recorded as a new image signal of the reference image and is used as a reference image for the following comparison.

Description

"Method for recognizing objects, moved in

  successive images in time belonging to a

  The present invention relates to a method for recognizing an object displaced in successive images in time belonging to a sequence, an image being subdivided into image areas, and for each image area the image signal is compared. in the reference image and / or from the comparison a displacement vector is determined and the variation of the image signal and / or the magnitude of the displacement vector is compared to a threshold, and from these comparisons it is recognized a moved object

  and, if so, a signal is emitted.

  It is already known according to J. Eliss et al., "A know-

  Ledge-based approach to automatic alarm interpretation using computer vision, on image sequences ", ICCST, 1989,

  page 215 and following, a method for recognizing ob-

  jets moved in images that follow each other in time; according to this method, the image signal is compared with a fixed reference signal to detect a displaced object. Moreover, the

  displacement vectors obtained are compared with

  predetermined displacement sensors for recognizing an ob-

jet moved.

  The present invention relates to a method corresponding to

  the type defined above characterized in that the recognition of a displaced object, the signal is recorded

  image as a new image signal of the reference image.

  and is used as a reference image for the com-

following parison.

  The method according to the invention offers the advantage that compared with the reference signal used for

  each zone of an image, according to the result of the ex-

  operation, adaptive updating, automatic

  the image analysis that is taking place. As a result, it is not necessary to provide different comparison thresholds for different scenes to be monitored, or

  to indicate different comparison motion vectors

  rent. The method according to the invention makes it possible, thanks

  image lysis used, to automatically adjust the comparison thresholds to different image areas. It leads to

  to greater security for the detection of objects de-

  placed, regardless of the environment observed and the

climatic influences.

  It is particularly advantageous to compare the

  image area of the image present in the corresponding image area.

  of the reference image as to the texture characteristics. This increases the security of detection of a moved object. To determine the texture characteristics, it is particularly advantageous to use two points

  adjacent images. This gives a sufficient amount of

  training and more the amount of data to be processed is

  low and allows fast processing of information.

  The method offers the advantage that for the detection of an object, the measurement quantities such as variation of the image signal, offset and texture determined for an image pair (reference image and present image), are not compared to a fixed threshold valid for the entire image, but these thresholds are obtained from a measurement

  separately for each image area and correspond to

  the statistical distribution over time of these measured variables (described by mean value and variance).

  The speed and security of the detection of ob-

  The displaced jet is made optimal by first using the variation of the image signal to recognize a displaced object and, in the case where the object has been moved, using texture features to recognize the object. displaced and, in the case where this comparison makes it possible to conclude that a displaced object is used, the displacement vector is used to recognize the moved object and only if this comparison also indicates a displaced object that is considered to be reliably recognized, and if a signal

  Alarm display of the object is given. Despite the

  important security linked to the detection of displaced objects,

  arrives at an extremely short processing time.

  The comparison of the quantities of measurement of the

  of the image signal, the displacement vector and the texture characteristic is simply and rapidly performed if the instantaneous measurement value of the measurement quantity of an image area, minus the average value over time for this quantity. measurement, is compared with the

  product of the standard deviation in the time of the

  measurement factor weighted by a weighting constant.

  This allows a quick and easy comparison of the quantities

  measurement for the detection of the object.

  The evaluation of the average value and the

  measurement magnitudes, signal vector image of de-

  placement and texture characteristics, is advantageously performed in a first-order recursive filter,

  i.e. a low-pass filter "Infinite Impuls Res-

  This allows a simple assessment of the value

mean and variance.

drawings

  An exemplary embodiment of the invention is

  presented in the drawings and will be described in more detail below. Thus: Figure 1 shows a mounting for the detection of a moved object.

  Figure 2 shows an image unit and dou-

  overripe. Figure 3 shows a distribution function of

  density for the exploitation of a measurement quantity.

  Figure 4 shows schematically the unfolding

of the process program.

  Description of the exemplary embodiment

  Figure 1 shows a mounting for detecting an object moved in images that follow each other in time. A shooting unit 1 is connected to a calculation unit 2 by a data line 7. The calculation unit 2 is connected by a data line 7 to an input unit 3, by another data line. 7 to a memory 4, by another data line 7 to an image screen 6 and by a

  other data line 7 to a signal display 5.

  Figure 2 shows an image unit 31 comprising

  For example, 5x5 image points 32 belonging to an image 30 are used. To determine the texture characteristics, the so-called doublet characteristics are used in the process described here. A doublet includes a set of two image points adjacent to each other locally. As measurement quantity for the texture characteristics, the sum of the squares of the differences of the image signals of the pairs of image points fixed by doublet and belonging to an image unit is used. As doublet we use three different types of doublet: a horizontal doublet 33, a diagonal doublet 34

and a vertical doublet 35.

  Figure 3 shows a distribution function of

  density P (vi) 40 of the measured variables vi for a situation

  scenic feature. The average value is represented in the time mvi of the measurement variable vi. The "characteristic" area of the measured variable is in the range of the mean value in time mvi +

  (Cvix svi) with Cvi corresponding to a weight constant

  fixed ration, and svi the standard deviation in time of the measured variable vi. The range of values 41 in which there are "atypical" measurement quantities is represented by the hatching in FIG. 3. An atypical measurement quantity is a sign of a displaced object. The range of values 42 in which there is a measurement quantity

  typical is not hatched. Memory 4 contains for each

  that image range a distribution density function for each measurement quantity (variation of the image signal,

  texture characteristics and offset vector).

  Figure 4 schematically shows the progress of the method for detecting a displaced object. In step 11 of the program, the shooting unit 1 takes an image and transmits it to a calculation unit 2. Then, the calculation unit 2 stores in memory in step 12 of the program, a signal image for each point of the image. The picture 30

  is subdivided into 31 units. In this example of achievement

  the image unit consists of an image area

presenting 5x5 points image 32.

  During the next program step 13, the computing unit 2 compares the variation of the image signal between the taken image and a stored reference image.

  in the memory 4. The variation of the image signal is

  trimmed according to a distribution function valid for this

  image unit as to the variation of the image signal de-

  written by the mean value and the variance, which represents a characteristic scenic situation. The image units whose variations of the image signal are atypical, are characterized by the calculation unit 2 by a first bit mask and are stored in the memory 4. A mask

  binary consists of a memory field that contains the

  memory placement for each image unit. This location

  memory is occupied by a signal 0 or 1 which characterizes it. In step 14, the computing unit 2 determines for each image unit and according to a predetermined method.

  which is stored in memory 4, the characteristics

  texture ticks; these are stored in the memory 4. In step 20, the computing unit 2 determines whether the texture characteristics determined for the image units are atypical texture characteristics. The image units whose texture characteristics are atypical are characterized by the calculation unit 2 by a second bit mask and the mask is recorded in the

memory 4.

  Then the calculation unit 2 determines in the step, starting from the comparison of the present image with

  the reference image entered in the memory 4, for each

  unit of image, the displacement vector of the image unit and written in the memory 4. The calculation unit marks in a third bit mask the image units having an atypical displacement factor and records this

  third mask in the memory 4. In step 21,

  if the image displacement vectors differ

  of a displacement vector typical of the image unit cor-

sponding.

  If yes, go to step 22. In step 22, the computing unit 2 recognizes in the image, by

  comparison of the binary masks, that there is a moving object

  for the image unit whose variation of the image signal, the texture and the displacement vectors are

  atypical; this is displayed by the signal display 5.

  Then we go to step 26 of the program.

  In step 26, the reference image is processed.

  In the case of a displaced, recognized object, the calculation unit 2 replaces the image signal of each image unit of the reference image with the image signal of the image unit of the image. image present. Then we go to program step 11; we take the next image and we go through

the program again.

  When the interrogation in step 21 indicates that all motion vectors of all image units are typical, step 18 is passed to the vectors.

  of displacement, in step 17 to the characteristics of tex-

  and at step 16 to the variations of the image signal, each

  than times using the calculation unit 2 with a first-order low-pass filter according to formula 2 taking into account

  measurement quantities present (variations of the

  image, texture and displacement), for a new mean value over time and a new variance over time of the measured values according to formulas 3 and 4

  for the present description of the distribution function

  measurement quantities of each image unit of the image.

  In addition, in step 16 of the program, the calculation unit 2 replaces the image signal of the image units of the reference image with the image signal and the image unit of the image.

  the present image, whose displacement vector is identical.

  absolutely nil. The image signal of image units whose

  displacement vector is different from zero remains unchanged

gen.

  At step 19 of the program, we record the

  their average and the variance of the measured values

  (variation of the image signal, texture and vector of

  cement), for each image unit in the memory 4; the average values and variances recorded until then

  for the variation of the image signal, the texture and the

  movement are erased.

  In this particular process, for each point

  image of a unit of image the quantities of

  sure (variation of the image signal, texture and displacement vector). We optimize the calculation time by determining

  the measured variables for each image unit.

  Particularly fast processing of the image occurs if after step 13 of the program, in step 14, the computational unit 2 determines only the texture characteristics of the image units whose variation of the image signal is atypical. At step 15,

  calculation unit 2 only determines the vectors of

  of image units whose signal variation

  image and texture characteristics are atypical.

  An exemplary embodiment will be described below.

  particular with the aid of the schematic flow of the

  FIG. 4 and FIGS. 1, 2, and 3. In step 11, the shooting unit 11, which in this particular case is a CCD camera, takes an image in the form of a video signal. analog. As a characteristic resolution of the image we have 756x288 pixels; for each image point we take an image signal. The analog video signal is first filtered by a low-pass filter in the shooting unit 1; it is digitized and transmitted to the calculation unit 2. In step 12, the calculation unit 2 entered

  the image present in the memory 4; then the recorded image

  is detected by calculation unit 2; in this example

  for example, the horizontal subdetection coefficient is equal to 4 and the vertical under-detection coefficient is equal to 2.

  2x4 points image in a point image. The sub-image

  detection is entered by the computing unit into the

  4. The whole of the detection of the object is done on

  the under-detected image. This procedure considerably reduces

  the means to be used for the operation without significantly influencing the detection characteristic. In the case where a displaced object has been recognized,

  we immediately have the image and all its resolu-

  tion; for example, it can be exploited for

  height knowledge or alarm control.

  In the case of implementation in the dark or at dusk, by a local accumulation of the image points in the video signal detection frame as well as by an accumulation of several images which follow one another,

  artificially prolongs the exposure time on the

  range of 1 / 50th of a second, in order to increase

  ter the sensitivity of the camera. For this purpose, the present statistic of the image signal is measured and, from this statistic, the contrast and the average brightness of the image of the camera are determined. When the lighting conditions are insufficient, the contrast of the local accumulation and in time of the image points of the video signal is increased; the average brightness is enslaved appropriately. These measurements lead, under unfavorable lighting conditions, to an improvement of the image signal and thus also to the security of the detection of

  the object in the dark or at dusk.

  Then the calculation unit 2 determines for each image signal of each point of the under-detected image, the difference with respect to the image signals of the image points of a reference image written in the memory 4. The image signals from the reference image are occupied by

  fixed initial values, transmitted to the calibration unit

  2 by the input unit 3. After a few images, the image signals of the reference image are occupied by

  values which correspond to a typical scenic situation

  teristic. In this exemplary embodiment, to detect changes in the time of image signals for each image unit, the quadratic differences of the image points between the undetected image present and the reference image are accumulated. This sum of the squares serves as a measure to describe the variation of the image signal over time in a

unit of image.

  During an object detection, that is to say during the recognition of a moved object, it is checked whether the

  determined measuring quantity, present (signal of

  images, texture signal, displacement vector) describes a typical or atypical scenic situation. This

  exploitation is based on a statistical analysis of the

  measurement units to determine for each unit

  of image a statistical distribution over time of

  measuring instruments. When operating a measurement quantity

  it is checked whether the measured value of the distribution

  a characteristic measurement quantity is sufficient or not. Since the measurement of the distribution function is complicated, the mean value and the variance are evaluated to describe the distribution. This evaluation is always done when there is no detection of an object alarm and

  that the unit of image corresponds to a scenic situation

  ractéristique. In this way, the mean value and the variance describe the distribution of the measured variables in characteristic scenic situations. When a present measurement value differs significantly from a

  magnitude of this description, this event

  gives an atypical scenic situation, that is to say

responding to an object moved.

  The decision whether a measurement variable differs significantly or not from a distribution

  statistics, is taken on the basis of a decision

  a threshold value. For this we use the following formula: (vi (x, y) - mvi (xy)) 2 / (cvi x Svi (Xy)) 2> 1 (1) in this one

  cvi: weighting constant [O <cvi <50 in relation to

  measurement with measured variables] vi (x, y): present measurement value of the measured variable vi at the (x, y) location Il mvi (x, y): average value over time of the quantity measure vi at the location (x, y) Svi (X, y): standard deviation in time of the

  measurement variable vi at the location (x, y).

  The weighting constant cvi also depends on the measurement quantity examined respectively. This is a measure that can deviate from multiple multiples of the

  standard deviation from the pre-measured value

  relative to the estimated average value of the

  measure. More this weighting constant is

  large and the present measurement value differs greatly-

  the assessed average value to trigger a

object alarm.

  Figure 3 graphically shows the inequality

  1. The distribution density function P (vi) of a measurement variable vi of coordinates (x, y) of the camera image is described by the two characteristic variables mvi (mean value) and svi (deviation standard). When the measurement variable vi is in the range [-tvi + tvi] (for example vi (1)), the probability density p (vi (1)) of the measurement value vi (1) is greater than a

  requested threshold Pmin- This threshold Pmin is the probability density

  bility of the measurement value vi to which the following condition applies: (vi-mvi) 2 = (ci x si) 2 When the measurement value is present vi is outside the range [-tvi, + tvi] (for example

  vi (2)), the object alarm is triggered. The probability of

  the residual probability, that is, the probability of

  clude to an atypical scenic situation by mistake, although the present measurement value of the distribution is sufficient, is characterized by the integral of all the possible values of measurement outside the interval [-tvi, + tvi]

and hatching in Figure 3.

  To make a threshold value decision, it is necessary to evaluate the statistical characteristic variables over time necessary, namely the average value and the

  variance for all image units and all

  measuring instruments by proceeding separately. This evaluation is done in a low-pass filter

  of type IIR- (Infinit Impulse Respond) of the first order se-

  according to the following formula: mvi (x, y, t): r * vi (x, y) + (1-r) * mvi (x, y, t-1) (2) or Pvi (x, y, t) ): r * vi2 (x, y)) + (1-r) * pvi (x, y, t-1) (3) and Svi2 (x, y, t): Pvi (Xyt) - mvi2 (x, y, t), (4) with mvi (X, y, t), mvi (x, y, tl): mean value in time of the measurement variable vi at time t or tl, Pvi (X, y, t), Pvi (x, y, t-1): power over time of the

measurement value vi at the in-

  as t or tl, vi2 (x, y, t): time variance of the measurement variable vi at time t, vi (x, y): value of the quantity vi at time t at location (x, y), r: coefficient of recurrence of

low-pass filter IIR.

  This filter is characterized in that the evaluation of the respective characteristic quantity per detection cell requires only a few calculation steps (operations) and can be done with only one memory

  additional measurement quantity or characteristic

  teristic. According to the coefficient of recurrence r fixed,

  the average oscillation time of the pass filter is changed

  bottom IIR (number of steps in the unit step at the input of the filter until the output of the filter where an amplitude is collected equal to 0, 9 times that of the input). This transient duration is a measure of the minimum number of images that must be processed until

  statistical characteristics sufficiently precise

  his. The principle is as follows: the higher the coefficient of

  recurrence is small and longer will be the transitional

  tory. The coefficient of recurrence determines whether the characteristic quantities obtained take the statistics

  long-term measurement quantity (coefficient of

  very small currency) or if we also take into account medium or short-term variations of the measured values in the statistics (such a large recurrence coefficient

  as possible). In addition, the description of the distribution

  measurement quantities continue in time by the recurrent evaluation for the measured variables

  adapting to scenic situations.

  At step 13 of the program, by detecting a

  variation, variations in the image signal are detected.

  be the reference image and the present image. The beaches

  in which there are atypical image

  are characterized in the first bit mask. The small variations of the image signal that correspond to a typical scenic situation, that is to say that have been generated by the noise of the video sensor, are suppressed for the detection of the object. In the embodiment described here, in addition to the difference of the image signal, account is taken in the decision of the image or the range

  image that precedes. Next, during the previous

  variation, an image range has been recognized

  as modified or unmodified, two different thresholds are used.

  to exploit the present variation of the image signal. Calculation unit 2 checks whether the squared differences of the image points of the image unit between the reference image and the present image differ more from the average value mvi of the differences squared of the points picture

  of the corresponding image unit, in relation to the

  time-course svi differences in squared image points multiplied by weighting constants

  cvi entered in Brief 4; in other words, we

  if the inequality represented by the formula (1) is

  tisfaite. The computing unit marks the image units that satisfy the formula (1), as atypical in the first

binary mask.

  In step 14 of the program, the calculation unit 2 examines whether the atypical variations of the image signal have

  caused by a change in the texture of the image.

  A texture is described by doublet characteristics.

  In the exemplary embodiment described here, two neighboring image points directly diagonally are used as doublet, as shown in FIG. 2. By measuring the neighboring image points in the diagonal direction, local correlations are taken into account. diagonal direction. By the analysis of the texture, we can detect the following scenic situations and thus eliminate an alarm of object: diffuse movements (for example rustling of leaves, rain, snow) global variations of clarity (day, night, projection of shadow in supermarkets) and small movements

of the camera (shake, shake).

  The calculation unit 2 determines the texture for

  each image unit. For that, a doublet describes the dis-

  local position of two adjacent image points. In this embodiment, the doublet is used in the form

  diagonal doublet 34. As a measure of the size of the

  texture characteristic, we use the sum of the difference squares of pairs of image points fixed by doublet in the image unit. This measurement quantity depends on the type of doublet used as well as the local correlation of the image signals within a unit of image. The

  The texture characteristics thus determined are

  in memory 4 by computing unit 2. As other

  doublets we have the horizontal doublet 33 and the vertical doublet

  cal 35 which can be used to determine the texture.

  According to the step 20 of the program, the measurement quantities of the texture characteristics are verified using the calculation unit 2 according to the formula (1). Image units whose texture characteristics are atypical, ie those which satisfy the formula

  (1), are marked in the second bit mask.

  At the program step 15 the calculation unit 2 de-

  ends a displacement vector to describe a piece of information

* Motion movement associated with each image unit. By evaluating the displacement vector, it is checked whether a unit of image of the present image is in the environment

  local from the previous image. For this research, we define

  Nite appropriate degrees of analogy, for example the absolute sum or the sum squared of the differences of the image points of the image area fixed by the unit of image.

  The local position of the image unit in

  the previous image (that is, for example, the one that cor-

  responds to the least sum of the squares of the image point differences) is associated with a displacement vector

  to the current image unit. The displacement vector obtains

  naked is written in memory 4. This describes the difference in the position of a unit of image in two images that

  follow in time. To evaluate the vectors of displacement

  In the process presented here, an algebra is used.

  "Three Step Blockmatching" described by T. Kogy, K. Linuma, A. in Hirano, T. Iijiama and T. Ishiguro in

  "motion-compensated interframe coding for video conferen-

  This "Three Step Blockmatching" algorithm is characterized by low complexity and reduced computation.

move on to program step 21.

  In step 21, for each image unit of the present image, it is asked whether the displacement vector corresponds to an atypical scenic situation, that is to say

  if the measuring variable of the displacement vector satisfies

made to inequality (1).

  If the measurement quantity of the vector of

  of all image units does not satisfy the unequal

  lite (1) we have displacement vectors corresponding to a typical scenic situation and we go to the stage of

program 18.

  In the program step 18, the calculation unit 2 determines for the motion vectors of the units

  image, using the formulas (2), (3), (4), a new

  a mean value in time, a new power over time and a new variance over time for the vectors of displacement of the image units. Then, in the program step 17, in the calculation unit 2, the texture characteristics determined for each image unit are used to obtain, according to the formulas (2), (3), (4), a new value average over time, a new power over time and a new variance in the

  time for the texture characteristics of each element.

image.

  In step 16, the computing unit uses the dif-

  squared off image points, obtained by applying formulas (2), (3), (4), to determine a new mean value over time, a new power over time and a new variance over time for differences squared image points of each image unit. The calculation unit 2 then replaces the image signals of the image units of the reference image, whose displacement vectors are zero, by the image signals corresponding to the image units of the present image. In step 19, the calculation unit 2 replaces the current average values, the power

  mean in time and the variance over time of the

  displacement of texture characteristics and differences of image points, by the new average values over time, the new power over time

  and the new variance over time calculated by the uni-

  calculation using formulas (2), (3), (4).

  When the displacement vector of a unit of image of the present image satisfies the inequality (1) at the program step, that is, if, according to the procedure described for the detection of the quantities of measurement, all the measured variables (modification of the image signal,

  texture and displacement vector) of the image unit envi-

  describe a non-typical scenic situation, we move to program step 22. At step 22, the

  calculation 2 recognizes a displaced object and, using the

  by signal 5, it emits a signal. At the same time, the present image is represented with all its resolution

  on screen 6. In step 26, the calibration unit

  cul 2 replaces the reference image with the present image.

  We thus arrive at what the current scene, with

  the displaced object, describes the typical scenic situation

  read in the following. When the typical scenic situation

  change, the reference image is adapted to the new

  scenic killing by the object alarm. This adaptation by a new initialization is repeated until

  the object is stopped or leaves the image.

  Then we go to program step 11 and we

goes through the program again.

  The variation of the image signals of the image

  is decided according to the importance of the

  movers. We distinguish between the following cases: the displacement vector is equal to zero; it means that there is no directed movement in the image. In the image range envisaged, the variations of the signal

  picture can all be described by a movement.

  Causes: many different movements in a beach

  image, local variations of clarity.

  The displacement vector is different from zero: for the case of a significant movement, the object alarm is triggered. To be able to also detect displaced objects that move slowly and in a directed way,

  the test of statistical significance of the description of the

  motion is used in addition to the detection of the object for further processing of the reference image. In all the image ranges in which there is a vector

  of displacement other than zero and which corresponds

  time to a characteristic situation, the reference image is no longer treated, that is to say that the reference image

  remains unchanged for these image areas.

  This procedure leads to small movements

  Directed elements are accumulated over time during an evaluation of the motion between the present image of the camera and the reference image. This makes it possible to distinguish between a

  intruder who moves slowly through the image and an

  alternatively moving in the wind, because in the first case, the movement accumulates and its importance

  increase in time, whereas in the second case the movement

  has no mean value since the tree is

anchored to the ground.

R E V E N D I CATIONS

  1) Method for recognizing an object moved in

  successive images in time belonging to a series of

  quence, an image being subdivided into image areas, and for each image area the image signal is compared to the reference image and / or from the comparison

  determines a displacement vector and compares the

  the image signal and / or the importance of the vector of

  placement at a threshold, and from these comparisons is recognized a displaced object and, if so, it emits a signal, characterized in that at the recognition of a displaced object, the image signal is recorded as

  new image signal of the reference image and it is

  read as a reference image for the next comparison.

  2) Method according to claim 1, characterized in that the image area of the present image is compared with the corresponding image area of the reference image for

  check the texture characteristics.

  3) Process according to claim 1 and 2, characterized

  in that, for the threshold test of the quantities of

  measurement, we use the statistical distribution of the

  used (as described by the mean value

and variance).

  4) Method according to claim 3, characterized in that the mean value and the variance for each image area are separately determined, and they are used for

the comparison.

  Method according to claim 3 and 4, characterized in that for determining the characteristics of

  texture, we use the difference of two points image lo-

calmly neighbors.

  6) Method according to one of claims 3 to 5,

  characterized by comparing the image signal of the unit

  the image signal to the corresponding reference image signal and if the variation of the image signal is at a fixed level above or below a determined image signal threshold, the variation of texture is compared unity

  of the image present at the corresponding image unit

  of the reference image with respect to a fixed threshold and that if the variation in texture differs from

  of a value determined in relation to the fixed threshold, we de-

  ends the vector moving the image range by

  relative to the corresponding image range of the image

  then a moving object is recognized and a signal is emitted when the displacement vector is greater than one

given threshold.

  7) Method according to one of claims 3 to 6,

  characterized in that, as a threshold comparison for the measured variables, an image signal and / or displacement vector and / or texture variation is used according to the following formula: (vi (x, y) mvi ( xy)) 2 / (cvi * svi (x, y)) 2> l, with cvi: weighting constant vi (x, y): present value of the variable vi at the (x, y) location, mvi (x, y): mean value in time of measurement variable vi at (x, y), svi (x, y): standard deviation in time of magnitude

measure vi at the point (x, y).

  8) Method according to claim 7, characterized in that the average value and the variance of the measured variables are evaluated: image signal, displacement vector and texture, for each image area using a filter low pass of the first order, according to the following formulas: mvi (x, y, t) = r * vi + (1-r) * mvi (x, y, t-1l) or Pvi (x, y, t) = r * vi2 (x, y) + (1-r) * pvi (x, y, t-1) and Svi (x, y, t) = Pvi (x, y, t) - mvi2 (x, y , t), in which: mvi (x, y, t), mvi (x, y, t-1) represent the mean value over time of the measurement variable vi at time t or tl, Pvi (x , y, t), Pvi (x, y, t-1) represent the power over time of the measurement variable vi at time t or t-1,

  Svi2 (x, y, t): variance over time of the magnitude of

  sure vi at time t vi (x, y): value of measurement variable vi at time t at location (x, y),

  r: coefficient of recurrence for the low-pass filter.