FR2484111A1 - Optical document reading circuit and associated filter - includes bar of photocells scanning line by line with filtering circuit comparing successive output signals to minimise errors - Google Patents

Abstract

The document device includes a lighted window (6) illuminating one line of a document (7). An optical system forms an image of the illuminated line on a bar (1) of photosensitive cells. A parallel - series storage register (2) is connected to the outputs from the bars of photocells (1) to provide signals (e j) relating to each of the cells in the bar. A filter (3) receives the signal (ej) and provides output signals (sj) which are transmitted to utilisation circuit. The circuits (1,2) are charge coupled devices (CCD) the former comprising a bar of 1728 photocells. The register includes two parallel-series registers whose outputs are switched alternately. The filtering circuit compares signals from successive cells in order to provide error-free output signals.

Description

 The present invention relates to a device for optical analysis of lines of n points (n integer greater than 1) of a document, comprising n photosensitive elements performing respectively, line after line, the reading of the n points of each line, a storage register coupled to the n elements and supplying n information signals relating respectively to these n points and a circuit for filtering the information signals.

It is known that the information obtained by reading using a strip of photosensitive cells is tainted with errors, that is to say that the level of the signal supplied by a photosensitive cell is not solely a function of information ~ located to the point that this photosensitive cell is intended to read; indeed different faults can disturb the reading
- imperfect focus of the light beam illuminating the bar,
- curvature of the objective field forming the image of a line to be read on the array of photosensitive cells,
- energy loss during transfer by shift registers of the energy contained in the photosensitive cells,
- possibly electrical coupling between photosensitive cells when these cells are placed side by side,
- in the event that one wishes to carry out a continuous reading of the document, the signal for reading a point may, as will appear in the description below, come from the total reading of this point and from the reading from the bottom of the previous point and from the top of the next point, this for reasons of playing time.

 It is known to avoid the first two of these reading faults by using very good quality objectives to give the image of the line to be read on the array of photosensitive elements; but such lenses are very expensive.

 It is also known to correct these reading faults by means of threshold circuits which must be controlled by variations in the brightness of the light source illuminating the document insofar as significant variations are to be feared, but these threshold circuits do not allow not correct information correctly, and significant loss of information may result.

 The present invention aims to provide a reading device made so as to compensate optimally all or part of the reading faults mentioned above.

This is obtained by taking account, to determine the corrected information relating to a point of the document, not only information relating to the reading of the point considered but also information relating to the reading of points neighboring the point considered.

According to the invention, an analysis device of the type indicated at the beginning of this text is characterized in that the filtering circuit is constituted by means.

calculation providing, as a function of the information signals relating to a given point and to p-1 points neighboring the given point (p integer at least equal to 2), a corrected signal relating to this given point.

The present invention will be better understood and other characteristics will appear from the following description and the related drawings which show
FIGS. 1, 2, 5 and 7 of the signals relating to optical document analysis devices,
FIG. 3, an optical document analysis device according to the invention,
- Figures 4, 6 8 and 9 partial diagrams of optical document analysis devices according to the invention.

 FIG. 1 relates to a strip of photosensitive cells associated with storage registers for reading, line by line, a document; this document is lit by a light slit and an image of the line is formed, using a lens, on the bar. A diagram showing these different elements will be given with the help of Figure 3.

Figure 1 shows the output voltages delivered by the storage registers. These output voltages correspond respectively to the charges accumulated in three of the cells (i-2, ii, i) of the strip when reading one of the lines of the document. This line of the document presented, at the time of its reading
- a black dot under cells i-3 and i-2
- a white point under the cell iI
- a black dot under cells i and i + 1.

it appears that the cell i-1, placed in front of a white point1 is sensitized and provides, at the output of the storage registers, a voltage to which, for simplicity, is assigned the value 100. it also appears that, in the concrete case in the example described, cells i-2 eti, although placed in front of black dots, supply voltages at the output of the storage registers of approximately 20% of the value of the voltage relative to cell iI (value 20 on the ordinate); this is mainly due to the imperfection of the focusing of the light beam on the bar. This voltage of 20% which is not due to the point placed under the cell, will be called in the following error voltage. This error voltage of 20% is not very serious insofar as the points at to read are very distinct black dots and white dots; indeed in this case a system with thresholds would make it possible to find the information given that to a white point would correspond a tension of 100 to 140% (100% + twice 20% for two neighboring cells placed in front of white points) and that a black point would correspond to a voltage of 0 to 40% (0% + twice 20% for two neighboring cells placed in front of white points). However, this error tension easily leads to interpretation errors when, in addition to black and white dots, gray dots of different gray levels must be considered.

 The ideal response to the illumination of cell ii with a white point while cells i-2 and i are placed in front of black dots, is given in Figure 2: no voltages relative to cells i-2 and i, level 100 voltage relative to cell iI.

The idea of the invention was to make a correction filter which transforms the response according to FIG. 1 into a response according to FIG. 2. Since the bars, in the area of their characteristics where they are generally used, have an almost linear response, that is to say that the response to two excitations is equal to the sum of the responses to these excitations, the correction filter can play its role whatever the configuration of points on the line to be analyzed and even with several gray levels in the same line of the document to be analyzed; this is confirmed by practice as will appear below.

 By calling - li + 1 the output voltages of the registers, relating to cells i-3 i-2 ..., i + 1, (ie the input voltages of the correction filter) and if- 2, si-1, if the output voltages of the correction filter, the filter must carry out the transformation below to pass from the response according to FIG. 1 to the response according to FIG. 2.

e. k1 + ei k2 + ei + 1 k3 = if
ei-2 k1 + e. k2 + e. k3 = if 1
ei-3 k1 + ei-2 k2 + ei-1 k3 = si-2 which gives, replacing the voltages with their values
1.k1 + 0.2 k2 + 0.k3 = 0
0.22.k1 + 1 k2 + 0.22k3 = 1
0.k1 + 0.2 k2 + 1.k3 = 0 the resolution of these equations gives
k1 = k3 = -0.22 and k2 = 1.08
FIG. 3 schematically shows how the optical document analysis device containing the filter which has just been calculated is made. A light slit, 6, illuminates a line of a document, 7; in reality, the slot 6 is very close to the document 7. An optic gives the illuminated line an image on a strip, 1, of photosensitive cells; this optic has been schematized by a simple objective 5. A storage register, 2, of the parallel-series type, associated with the bar 1 delivers the signals e. relating to each of the 3 cells of the bar and which was discussed further (ei-3 ... ei + 1). A filter, 3, receives the signals e. and outputs the s signals. which was also discussed further (si-2, si-1, si); these latter signals are sent to usage circuits 4.

The circuits 1 and 2 in Figure 3 are constituted by a C.C.D. (initials of the English words "charge coupled device" called "circuit à charge coupled" in French) produced by the company Fairchild under the reference- 121H and comprising a strip of 1728 cells and the associated storage register; this register is in fact composed of two parallel-series registers whose outputs are alternately switched.

In FIG. 3, the clock circuits necessary for the operation of the analysis device have not been shown. The operation of the storage register, 2, is ensured in a conventional manner; in the case of the example described, the reading by the cells of the bar is done at the rate of a line every 20 mS and the transfer of a line into the storage register lasts 1 PS; the transfer of the contents of the register is done at the rate of 250kHz and therefore lasts, for a line, 1728 x 250000 # 7 S.

FIG. 4 is a detailed diagram of the filter 3 of FIG. 3 to which has been added a clock circuit, 30, receiving the signal at 250 kHz used for the operation of the storage register 2 and a reference signal at 1 MHz used to produce the signal at 250kHz.

These two signals as well as the signals, H1, H2, H3 produced by the filter clock circuit will be described using the diagram according to FIG. 5.

The circuit according to FIG. 4 comprises a level transposition circuit, 31, which receives the signals coming from the storage register 2 (FIG. 3); the role of this circuit is to bring back to OV the level of the signals supplied by the storage register and corresponding to the reading of a black point on a document; this level transposition circuit is produced using an operational amplifier manufactured by the company Texas
Instrument under reference TL084.

 The output signals from the level transposition circuit, 31, are sent to the input of an impedance matching circuit, 32, also produced using an operational amplifier TL084; this operational amplifier mounted as a unit gain amplifier has an input impedance of 1M # and an output impedance of 100 #.

The output signals of the impedance matching circuit 32 are sent on the one hand to the first input of an adder-subtractor circuit, 37, on the other hand to the signal input of a blocker sampler circuit 33 of which the clock input receives a signal H delivered by the clock circuit 30. The sample-and-hold circuit 33 is produced using an analog switch of the MC 14016 type manufactured by the company Motorola, and comprises, between the output from this analog switch and ground, an nF capacitor; this sample-and-hold circuit 33 plays the role of memory without controlling the signal H1
The output signals of the blocking sampler circuit 33 are sent to the input of an impedance matching circuit, 34, identical to circuit 32; the output of circuit 34 is connected on the one hand to the second input of the adder-subtractor circuit 37 and on the other hand to the signal input of a sampler-blocker circuit 35, identical to circuit 33 and which receives on its clock input the signal H2 delivered by the clock circuit 30.

 An impedance matching circuit 36, identical to circuit 32 couples the output of the sample-and-hold circuit to the third input of the adder-subtractor circuit 37.

 The adder-subtractor circuit 37 is produced using an operational amplifier TL084 whose feedback resistances are determined to assign a multiplicative coefficient equal to 0.22 at the input - and a multiplicative coefficient equal to 1, 08 at the "+" input. As for the three inputs of the adder-subtractor circuit, they are coupled by resistors, for the first and the third, to the input - "of the operational amplifier and, for the second, to the input "+" of the operational amplifier; thus the adder-subtractor circuit comprises three input multiplier circuits d respectively coupled to its three inputs and whose multiplying coefficients are respectively kt = 0.22, k2 = 1, 08 and k'3 = 0.22.

 The adder-subtractor circuit 37 is followed by a sampler-blocker circuit 38r itself followed by an impedance matching circuit 39; the circuit 38 is identical to the circuit 33 and receives on its clock input the signal H3 delivered by the clock circuit 30, as for the circuit 39 it is identical to the circuit 32.

The operation of the filter according to Figure 4 can be explained as follows. Due to the information transfers, the process of which will be described below, at a given instant the sample-and-hold circuits 33 and 35 have stored, by their output capacitor, the signals ei t and ei-2, respectively; ; they therefore respectively apply these signals to the second and third inputs of the adder-subtractor circuit 37. The adder-subtractor circuit 37 continuously performs the calculation for which it is intended, that is to say to provide an output signal of equal level with the difference between the level dt signal on its second input and the sum of the levels of the signals on its first and third inputs. The output signal of the adder-subtractor circuit 37 must therefore only be taken into account when, for example, the signals ei ei-1 ei-2 are on its three inputs, that is to say outside the instants of the transfers carried out by the sample-and-hold circuits 33, 35 controlled respectively by the pulses H1 and H2 and when a new signal is appeared on its first entry; these conditions are fulfilled, as shown in the time diagram of FIG. 5, during the times of the pulses H3 which control the taking into account of the output signals of the adder-subtractor circuit 37 by the sampler-blocker circuit 38.

FIG. 5 shows the signals at 1 MHz and 250 kHz applied to the input of the clock circuit 30. The signals at 250 kHz give the instants of appearance, at the input of the filter, of the signals ei-2,, e ei 1 ei
FIG. 5 also shows the output signals H1, H2 w H3 of the clock circuit 30; these signals consist of pulses of IS of duration repeating at the frequency of 250 kHz: the pulse of the signal H3 appears a microsecond after the signal ei appeared at time t on the input of the filter then at times t + 2 Set t + 3 S the H2 and H1 pulses appear respectively.

Thus between the time to + 1 S and to + 2 S, while the signals e1 ei-1 and ei-2 are respectively present on the three inputs of the adder-subtractor circuit 37; the sample-and-hold circuit 38 takes into account the output signal of circuit 37, that is to say that it receives the signal
si-1 = k1 ei-2 + K2 ei-1 + k3 ei and delivers it as output signal, for 4 while being blocked from t + 2 / uS to t + 5 S During this time when the output signal from adder-subtractor circuit 37 can no longer be taken into account, the transfer of the signal ei-1 1 in the blocking sampler circuit 35 can be carried out during the time interval ranging from t + 2 S to t + 3 S (under the control of l pulse of signal H2); then, still while the circuit 38 is blocked and while the signal e. is always present on the input of the filter, the transfer of the signal ei in the sampler-blocker circuit 33 can be carried out during the time interval ranging from to + 3 / uS to t + 4 S (under the control of the signal pulse H1).

At time t + 5 S t while a new signal, ei + 2 Z has already appeared from 1 / uS on the filter input, the cycle can resume, that is to say: taking into account the signal of output of the adder-subtractor circuit 37 which is then
si = k1 ei-1 + k2 ei + k3 ei + 1 then transfer from e. in circuit 35, then transfer from e. in circuit 33 and so on so as to successively obtain if + 1 'if + 2 etc ...

A simplified explanation of the overall operation of the filter according to Figure 4 can be given using Figure 6. To obtain output signals of the form
si-1 = - 0.22 ei-2 + 1.08 e. - 0.22 ei which was discussed further in the description, the filter is essentially composed, as shown in the block diagram of FIG. 6, by: two delay circuits 40, 41, three multiplication circuits 42, 43, 44 and an adder circuit 45. The delay circuits are constituted by the sample-and-hold circuits 33, 35 of FIG. 4, the sampling control of which is constituted by the signals H1 and H2; these circuits keep on their output capacitor of 1nF and during the time separating two sampling pulses, i.e. for 4 S, the information they receive on their input. The signals ej follow three channels between the input and the output of the filter, namely:
- a first non-delayed track, comprising the
multiplication circuit 42 whose coeffi
multiplication cient is - 0.22 and of which
the output is connected to a first input of the
adder circuit 45,
- a second channel causing a delay of 4 S,
comprising the delay circuit 40 followed by the cir
cooked multiplier 43 whose coefficient of
multiplication is + 1.08 and whose output
is connected to a second circuit input
adder 45,
- a third channel causing a delay of
4 + 4 = 8/15, including the delay circuit
40 followed by delay circuit 41, itself followed
of the multiplier circuit 44 whose coefficient
multiplication is - 0.22 and whose sor
tie is connected to a third circuit input
adder 45.

So at the moment when the signal "- 0.22 e." appears on the first input of the adder circuit 45, the signals "+ 1.08 ei-1" and "- 0.22 ei-2" appear respectively on the second and third inputs of circuit 45, causing the appearance of the signal if -1 on the output of the adder circuit 45
As an indication are shown, with the aid of FIG. 7, the results obtained with the filter which was used for the description above in the case where the information to be analyzed is considered relatively to a five-level scale: black, N, and white levels, B, and three intermediate levels of gray. These three intermediate levels, hereinafter called G @ G2 G3, enter respectively, with the notations already used on the occasion of FIGS. 1 and 2, the amplitudes 75, 50 and 25 on the output signals relating to the cells placed opposite these points and the amplitudes 15, 10 and 5 on the two neighboring cells.

The input signal ej of the filter has been drawn in strong lines and the output signal, sj 'in broken lines (see Figure 7). Signal s. it clearly appears
3 that the points analyzed are points respectively
N, B, N, Ne B, G1, B, N what the control from the document verifies perfectly. On the other hand, a threshold system using the signal e. (signal in strong lines) as input signal, n could have determined, at the same time, that the second point is a white point, B, and the seventh a point G1 since the level of the signal ej is lower for the second point (.100) to that of the seventh point (115); this would therefore have resulted in at least one interpretation error: second point interpreted as a G1 point or seventh point interpreted as a white point.

The filter which has been described above is a filter working with analog signals but it is also possible to produce a filter according to the invention working with digital signals: an analog-digital converter must then be inserted between the storage register associated with photosensitive cells and the filter; as for the filter, it will include delay circuits, for example shift registers, and means for calculating, in numerical calculation, the value
si-1 = - 0.22 ei-1 + 1.08 ei-1 - 0.22 ei these calculation means being either specially designed for the filter or obtained by connection to a digital computer programmed for this purpose.

 it should be noted that to calculate si-1, only e i-I and e were taken into account. that is, it was considered that the optical coupling did not extend beyond the two adjacent cells of the illuminated cell; in fact this is not entirely accurate but the quality of the results obtained in this way with three coefficients is such that it is not necessary, in the majority of cases, to have recourse to five coefficients si-1 = k1 ei-3 + k2 ei-2 + k3 ei-1 + k4 ei + k5 ei + 1 or with even more coefficients.

it should also be noted that it is possible to vary the coefficients, as a function of the cells concerned, in order to take account of the fact that, the objective (5, FIG. 3) not being perfect, the image is of better quality on the cells placed in the center of the strip than for the cells placed at the ends of the strip.

Likewise, it may be useful to vary the coefficients, depending on the cells concerned, to take account of the energy loss which occurs, in the storage register (2, FIG. 3) at each shift so that, for a line the first signal (e) has proportionally much less energy loss than the last (in et728)
A third example of a case where it may be necessary to modify the coefficients is the case where the optical analysis device has several different interchangeable objectives, making it possible to read documents of different sizes. In the first two examples the coefficients must be adjusted during the analysis of a line while-in this third case the coefficients are modified when the objective is changed.

 In the case of analog filters the variation of the coefficients is given by a variation of amplification gains which can be obtained, for example, by means of resistors switched according to the cells considered; it is also possible to use field effect transistors which allows a continuous variation of the coefficient to be obtained: the field effect transistor itself acts as a resistance and the value of this resistance is a function of the value of its grid tension.

 In the foregoing, only parasitic illuminations due to neighboring points on the line being analyzed have been taken into account for the information relating to a point. For a conventional analysis of the document, with a flag on each line to be analyzed, there are practically no other parasitic illuminations which should be taken into account; it is different in the case of a continuous document analysis. Indeed during a continuous document analysis, so that when the light received and therefore the exposure time are sufficient, it is most often necessary to start the reading of a line while the upper part of the cells of the bar is still on the previous line and to stop the reading while the lower part of the cells is already on the following line: it as a result it is then necessary to take into account, for the information relating to a point, the parasitic illuminations due to the neighboring points on the line being analyzed, on the preceding line and on the following line.

 FIGS. 8 and 9 are two diagrams showing how account can be taken of the parasitic illuminations due to the line preceding and to the line following the line being analyzed.

FIG. 8 is a partial diagram of a device for continuous analysis of a document. In this device, a circuit "cell strip + storage register" 50, similar to the elements t and 2 of FIG. 3, delivers the signals e. To be processed. These signals are processed in a first filter 51 similar to the filter according to the figure. 6 and which therefore does not take into account, in this processing, the stray light due to the line preceding and to the line following the line being analyzed; this first filter provides output signals of the form
5i-1 = k1 ei-2 + k2 ei ,? + k3 ei
A second filter 54 receives, on a first input, the output signals of the filter 51, on a second input the output signals of the filter 51 delayed by a first shift register, 52, and, on a third input, the filter output signals 51 delayed by the first shift register, 52, then by a second shift register, 53. These two shift registers each provide a delay equal to the time interval between the start of the onset, on the output of the Sir filter of the analysis signals of two successive lines of the document.

The filter 54 is composed of three multiplier circuits, 55, 56 and 57, connected respectively to its three inputs and of an output adder circuit, 58, so as to deliver an output signal of the form:
Si-1 = K1 s'i-1 + K2 si-1 + K3 s "i-1 where K1, K2 t K3 are coefficients determined experimentally like the coefficients k1 e k2 r k3 which have been discussed about the filter according to FIG. 6, and where if 1 being the signal supplied by the filter 51 relative to a given point on a given line, s'i 1 and 511i-1 are the signals supplied by the filter 51 relative to the points placed respectively just above and just below the given point. The filtered signal S. is therefore determined from signals supplied by the photosensitive cells relating to nine points of the document: e'i-2 e'i-1 and e'i for the line preceding the given line, ei-2 ei-1 and ei for the given line and e "i-2 e" i-1 and e "i for the line following the given line.

FIG. 9 is a partial diagram of another device for continuous analysis of a document. In this device a circuit "cell strip + storage register", 60, similar to elements 1 and 2 of FIG. 3; delivers signals e. These are sent to the nine inputs of a filter 69 via nine different channels:
- a first direct route
- a second channel comprising a delay circuit
61 whose delay is equal to the interval of
time between the start of two successful points
sifs of the same line,
- a third track comprising, in series, the cir
delay cook 61 and delay circuit 62 of
same delay time as circuit 61,
- a fourth channel comprising a dice register
timing 63 which brings a transmission delay
equal to the time interval between the start
of the appearance, on the output of circuit 60
analysis signals of two successive lines
ves of the document,
- a fifth channel comprising, in series, the re
shift register 63 and a reta-rd circuit 64
same delay time as circuit 61,
a sixth channel comprising, in series, the regis
be offset 63, delay circuit 64 and a
delay circuit 65 with the same delay time as
circuit 61,
- a seventh channel comprising, in series, the regis
be shift 63 and shift register 66
which brings the same delay time as the re
shift register 63,
- an eighth channel comprising, in series, the re
shift registers 63 and 66 and a circuit to re
late 67 at the same time as the circuit
61,
- and a ninth channel comprising, in series, the
shift registers 63 and 66, the circuit to re
late 67 and a delay circuit 68 at the same time
delay than circuit 61.

Let e'i-2, e'i-1, e'i ei-2, ei-1, ei, e "i-2, e" i-1 and e "i be the same signals provided by the photosensitive cells as in the case of the description of FIG. 8; that is to say the signal ei-1 of a given point it of a given line, the signals ei-2 2 and ei of the points surrounding this given point on the given line and the signals relating to the points just above and just below these three points of the given line. At the instant when the signal e "i appears on the first input of the filter 69, the signals e "i i-1, e" i-2, ei, ei-1, ei-2, e'i, e'i-1 and e'i-2 will appear respectively on the second to ninth entries of filter 69 because delays from the second to ninth routes.

The filter 69 is composed of nine multiplier circuits, 71- to 79, the inputs of which constitute the nine inputs of this filter respectively, and of an output adder circuit, 70. The multiplication coefficients K '@ to to K'g of the nine multiplier circuits are determined experimentally as the coefficients k1, k2, k3 relative to the filter according to FIG. 6. At the output of the filter 69 therefore appears a signal
If i-1 = K1 e "i + K'2 e" i-1 + K'3 e "i-2 +
K'4 ei + K'5 ei-1 + K'6 ei-2 +
K'7 e'i + K'8 e'i-1 + K'9 e'i-2 which corresponds to the signal ei-1 @ relative to the given point of the given line which was discussed above but corrected by function of the signals from the photosensitive cells relating to the eight points of the document surrounding the given point.

As in the case of a stop analysis on each line it is possible, for a continuous analysis of a document, to take into account on the line being analyzed as well as on the lines preceding and following the line in during analysis, more than three signals provided by the photosensitive cells to determine the corrected signal relating to a given point.

It is also possible, in the case of continuous reading, to vary the multiplicative coefficients making it possible to obtain the corrected signal and this under the same conditions as during an analysis with stop on each line.

Claims (5)

 1. Device for optical analysis of lines of n points (n integer greater than 1) of a document (7), comprising n photosensitive elements (t) performing respectively, line after line, the reading of the n points of each line, a storage register (2) coupled to the n elements and supplying n information signals relating respectively to these n points and a filtering circuit (3) of the information signals, characterized in that the filtering circuit (3) consists of calculation means supplying, as a function of the information signals relating to a given point and to pt points adjacent to the given point (p integer at least equal to 2), a corrected signal relating to this given point.  2. Analysis device according to claim I in which n is greater than 3 and p greater than 2, characterized in that the calculation means comprise an addition circuit (45) with m inputs (m integer at least equal to 3) and a direct channel (42) and m-2 delayed channels (40-43, 40-41-44) coupling the storage register respectively to the m inputs, in that the m channels comprise respectively m multiplier circuits ( 42, 43, 44) of coefficients k. , where j varies from 3 1 to m, and in that the ml delayed channels respectively comprise ml delay circuits (40, 40-41), of delay time equal to gd, where g varies from 1 to mi and where d is the time interval separating the appearance, at the exit of the storage register, of two information signals (ei, ei) relating to two consecutive points of the same analyzed line.  3. Analysis device according to claim 2 intended for continuous analysis of a document and in which p is greater than 8, characterized in that the calculation means also include a circuit of caS cul (54) with 3 inputs, and 3 connections to couple the added circuit respectively to the 3 inputs of the calculation circuit, in that one of the connections is a direct connection and the other two (52, 52-53) respectively introduce delay times equal to D and 2D, where D is the time interval separating the appearance, at the output of the storage register, of the information signals relating to two consecutive lines of the document and in that the calculation circuit (54) comprises 3 circuits - input multipliers followed by an output adder circuit. 4. Analysis device according to claim 1, intended for the continuous analysis of a document in which p is greater than 4, characterized in that the calculation means comprise a calculation circuit (69) with p inputs, and -p channels coupling the storage register (60) respectively to the p inputs, the p channels introducing delays determined so that at their outputs the information signals relating to the given point and to the p-1 neighboring points of the given point are concomitant and of matter that the p points are points of three consecutive lines and in that the calculation circuit (69) comprises p input multiplier circuits followed by an output adder circuit. 5. Filtering circuit (6) of information signals, for an optical analysis device (1 to 7) of lines of n points (n integer greater than 1) of a document (7), this device comprising reading and transmission means (1, 2 5r 6) performing respectively, line after line, the reading of n points of each line, and delivering n information signals relating respectively to these n points, characterized in that it is constituted by calculation means (3) providing, as a function of the information signals relating to a given point and to p-1 points neighboring the given point (p integer at least equal to 2), a corrected signal relating to this given point .