DD292822A7 - ARRANGEMENT FOR SENSITIVITY CORRECTION OF INTEGRATED SENSOR ARRAYS - Google Patents

Abstract

The invention relates to an arrangement for the sensitivity correction of integrating sensor arrays, in particular for the PRNU and Linearitaetsfehlerkorrektur of CCD image sensors. In accordance with the present invention, the advantages of a reference table memory that is both meszadressed and sensor-indexed are exploited by including a sensor element indicator including a sensor element counter and a comparator unit for determining the characteristic endpoints of all sensor elements, and a characteristic transformer consisting of a programmable integration time sensor, a characteristic memory for storage of integration time error counters and a transcoder for grayscale matching of the ideal sensor line to the follower electronics, are present, which are coordinated by a calibration control unit in at least three calibration regimes and a Meszregime. Fig. 1 {sensitivity correction; PRNU correction; Linearitaetsfehlerkorrektur; Sensor array, integrating; CCD image sensor; Reference table memory; Sensor element indicator; Characteristic endpoints; Characteristic memory; Integration time counter, programmable; transcoder}

Description

For this 3 pages drawings

Field of application of the invention

Dif » ^r . n.unge relates to an arrangement for the sensitivity correction of integrating sensor arrays, in particular of CCD image sensors. It allows the extension of the use of image sensors on such image processing tasks, such as the optical inspection and measurement of submicron submicrometer submissions by gray image processing, which place particularly high demands on uniformity and linearity of the photosensitivity of the sensor array and thereby the physical and technological Reached limits. The invention thus finds application in all demanding image processing applications, in particular in the inspection devices for microelectronics.

Characteristic of the known state of the art

In the prior art solutions for correcting the non-uniformity of the photosensitivity of the individual sensor elements of image sensors under the name PRNU correction (Photorosponse-Non-Uniformity) are known. The basic principle is known from DE-PS 2714777 deposited by MMB. The described arrangement for digitally correcting the PRNU of each individual image sensor solves the problem such that a reference signal (eg DC level) stored at the beginning of each sample period and a dark image signal typical to the image sensor element are subtracted from the image sensor measurement and the result is stored in a correction value memory for each sensor element Gradient factor is multiplied. The slope factors are obtained on the basis of the different brightness signals of the sensor elements by a caliber measurement. The disadvantage of this solution is that only a slope factor describes the characteristic of a single image sensor element and a non-linear characteristic higher-order correction is necessary. As a result, the real-time image processing provided per se is limited by lowering the processable image signal frequency.

Further developments of this classic PRNU correction are made in the direction of the transition to the logarithmic level to simplify the required arithmetic operations in the specifications GB-PS 1526801 and DD-PS 253135. However, the basic principle and the associated disadvantages of a linear PRNU correction in the case of non-linear sensor characteristics are essentially retained.

On the other hand, there are z. In control engineering, for example, numerous solutions are used in which non-linear tanner characteristics are transformed into signals having a linear functional relationship to a particular parameter (e.g., illumination).

In DE-OS 3011771 a useful overview of such solutions * ur linearization is given. In addition to such approaches, such as the approximation of the characteristic curve with a set of straight-line segments that excrete for reasons of accuracy, and solutions such as exponential and polynomial methods that do not meet the real-time requirements, the use of so-called Sook-up Tables (LUT) or Reference tables are a viable alternative.

Thus, in DE-OS 3339349, such a reference table is used as a characteristic generator by the digital measurement signal is applied to the address input of a read-only memory and the value read from the memory from this digital address value is used for further signal processing. In the read-only memory, the values assigned via the desired characteristic curve of the analog measured variable are respectively stored in the memory locations whose address is equal to the digital signal!: *. The use of such reference tables permits a high correction speed and enables a combined PRNU and linearity error correction. In this case, the content of the table, that is the stored, corrected signal, is of particular importance.

In DE-OS 3339349 it is assumed that the output from the read-only memory characteristic is known from the outset. For a CCD image sensor, this would mean that both the end point (maximum sensor signal to be processed) and the slope (sensitivity) of the desired ideal characteristic are known. Furthermore, the PPNU and linearity errors would have to be determined and stored as a function of the signal size or illumination for all image sensor elements. In DE-OS 2549222 a digital circuit arrangement for linearization of nonlinear encoder characteristics is presented, which aims at reducing the storage capacity required a reference memory used as a reference table by not the corrected measured value variables are entered in the table, but only the difference between the output signal of the transmitter and the desired linear course. In this case, the digitized measuring signal of the encoder is converted into a count, which is used to address the read-only memory to read out the corresponding differential correction values. The differential correction values are stored in a code complementary to the counting code of the measured value counter so that the complementary code counted by the measured value counter can be taken over into a correction counter and the difference correction value in the correction counter is counted-for the counting of the measured value counter and thus for the corrected measured value counter reading at the output of the measured value counter , The disadvantage here on the one hand, that for larger nonlinear deviations, the intended memory space savings does not occur significantly and a reduction in the correction speed is recorded, so that the advantages of reference tables in terms of real-time options are almost compensated. On the other hand, the knowledge of both the real and the desired sensor characteristic is necessary for the writing of the differential correction values into the read-only memory.

Object of the invention

The aim of the invention is to find a way to qualitatively better sensitivity correction of integrating sensor arrays, which allows a non-linear correction of the measured values in real time with little time and technical effort.

Explanation of the essence of the invention

The invention has for its object to realize a possibility for sensitivity correction of integrating sensor arrays, in which a reference table with data of the desired sensor characteristic is described by a suitable calibration, the explicit knowledge of the sensor element characteristics is not required.

According to the invention, the object is achieved in an arrangement for sensitivity correction of integrating sensor arrays with a sensor array downstream ADU, a measured value addressed reference table memory and two counters, in that the output of the ADU except the connection to the address input of the reference table memory connects to both the input of a Sensorelementindikators which contains the first counter as a sensor element counter and a comparator unit for determining the characteristic curve end points of all sensor elements relative to the characteristic curve end point of the most sensitive sensor element, as well as with the address input of a characteristic transformer included in a characteristic

Characteristics memory has the fact that the data output of the sensor element indicator is also the address input of the reference table memory and the characteristic memory stored in that the characteristic transformer next to the characteristic memory, the second counter in the form of a programmable integration time counter for gradually increasing the integration time of the sensor array and a transcoder to adapt the data to the gray value resolution the signal processing downstream of the reference table memory, wherein the count output of the integration time counter is connected to the data input of the characteristic memory and the Umkodierung between the data input of the reference table memory is arranged that a means for varying the calibration input signal size in the receiving channel in front of the sensor array is arranged that a calibration control unit for writing / Read control and chip enable, to control the integration time counter and the V device Ariation the Kalibriereingangssignalgröße and for switching between the Kalibrierregimeprozessen and the measurement regime for input signal recording is present, the Kalibriersteuereinheit is temporarily switched in the calibration mode to the address line of reference table memory and characteristic memory and has a connection to the output of the comparator unit. In the sensor element indicator, the use of an additional classification memory whose data input is identical to the output of the ADU whose address input is connected to the counting output of the sensor element counter and whose data output is identical to the output of the sensor element indicator, is expedient for reducing the required memory capacity of reference table memory and characteristic memory Connections with the calibration control unit for read / write control and chip activation of the classification memory are provided.

The comparator unit is advantageously composed of a comparator and a reference value transmitter, which is set to achieve the highest level of the ADU.

The device for varying the calibration input signal variable is advantageously designed as an adjustable optical filter device.

The timing control of the integration time counter is coupled to either the read-out clock of the sensor array or to the pulses of a laser-path measurement system that captures the spatial coordinate of the input signal sample. The invention assumes that the real-time correction of PRNU and linearity error in image sensor elements is to be designed exclusively using reference table memories, especially as in image sensors, the output signal size of each sensor element not only depends on a parameter, but is a function of illuminance and integration time. In addition, it is assumed that the characteristic curves of the sensor elements are not known a priori, so that an adaptive characteristic determination is required for the subsequent correction. Taking these considerations into consideration, the concept of the invention is based on the physical relationship that the recording of the characteristic curves of integrating sensor elements by varying the input signal quantity at a constant integration time does not differ from the characteristic recording by varying the integration time with a constant input signal size. Integration time and input signal size are therefore interchangeable. This situation is taken into account in the arrangement according to the invention by the sensor element characteristics in the form of integration time counter readings with constant calibration input signal size and incremental increase of the integration time until the end of a unique characteristic curve of the most sensitive sensor element in the characteristic memory address (ie, depending on switching to the next level of measurement ADU) and converted in a transcoder into gray values adapted to the subsequent signal processing, which are then queried in the reference table memory so as to be interrogated by measurement value and sensor element.

In a preferred variant of the invention, a classification of the sensor elements into classes is undertaken in order to reduce the storage capacities of characteristic and reference store memories. The class division is based on

the characteristic curve end points of the sensor elements, based on the end point of the characteristic curve of the most sensitive sensor element.

With the arrangement according to the invention, it is possible to carry out a non-linear PRNU and linearity error correction by means of an automatic calibration of any real sensor characteristic curve (without their explicit knowledge) in a desired linear characteristic with little effort to make them in a measured value addressed and sensorelementindikativ addressed reference table memory for a demanding Signal processing task in real time to steep.

embodiment

The invention will be explained below with reference to an exemplary embodiment. The drawings show

Fig. 1: the block diagram of the inventive arrangement Fig. 2: an advantageous embodiment of the invention Figure 3: the characteristic diagram of an image sensor array.

The inventive arrangement consists in its basic structure - as shown in Fig. 1 - from an integrating sensor array 1, an ADU 2, a sensor element indicator 3, which includes a sensor element counter 31 and a comparator unit 33, a characteristic transformer 4, an integration counter 41, a characteristic memory 42 and a transcoder 43, a reference table memory 5, a calibration control unit 6 and a unit for varying the calibration input signal magnitude 7.

The arrangement will now according to FIG. 2 in a special variant, which provides for the use of a CCD line 11 as an integrating sensor array 1 and a controllable optical filter device 71 as a unit for varying the calibration input signal size 7, with the addition of a classification memory 32, the memory space reduction in Characteristic memory 42 and in the reference table memory 5 serves to be described in their function. The additionally introduced classification regime can be eliminated with a smaller number of sensor elements η, and the addressing of the characteristic curve memory 42 by the sensor element indicator 3 is tapped directly from the sensor element counter 31 (see FIG. 1).

1. Illumination calibration regime

The digital signals present at the output of the ADU 2 are supplied from the input of the sensor element indicator 3 to the comparator unit 33.

In the comparator unit 33, a comparator 331 and a reference value transmitter 332 are arranged, wherein the reference value 332 is set appropriately so that its digital output value is equal to the maximum occurring Meßwertstufe m " _lJA at the output of the ADU 2. As a result, the defined characteristic curve end point is specified for the following characteristic curve recording regime.

With this setting of the comparator unit 33, the set reference value m, _o i is compared with the measured value stage m (ni) of the ADC 2 corresponding to the output signal of each individual sensor element m for a given illuminance E within a read-out cycle of the CCD line 11. If all measured value stages m (n,) <m, _o i, the comparator 331 generates a logic signal, which it outputs to the calibration control unit 6 via a connecting line. This controls the optical filter device 71 to increase their transparency. Then, the measurement cycle described above is repeated until the reference value _re m t of at least one of the sensor elements ni, the sensitive sensor element is achieved n _MSX, ie it applies: m (n _m ") ä m _ie i. When this condition occurs, the comparator 331 changes its logic signal, and the calibration control unit 6 does not further increase the illuminance E.

2. Classification regime

For this with a small number of sensor elements ni not mandatory regime of the calibration process uses the calibration control unit 6 the last modified logic signal of the comparator 331 to inserted in the additionally inserted into the sensor element indicator 3 classification memory 32 whose data input is connected to the output of the ADU 2 and whose address input is linked to the counting output of the sensor element counter 31 to store the maximum measured value stages m _max (r0 as class numbers) .The last class division existing in the classification memory 32 when the reference value m "f is reached is stored as valid in the classification memory the change of the logical signal from the comparator 331 the storage of further measured value stages m, according to the scheme ki = m _max (n _m , x)

kj = m _{m, x} (n _m, xi)

ki = m _m , x (n _max - (l-1)),

where I is at most the number of Meßwertstufen rrij of the ADU 2 (or an integer divisor thereof), are used. The expression m _max (n _m , _x - (l-1)) may optionally be for several Sensorelemonte n _{i (} have approximately the same sensitivity (based on the maximum value of the most sensitive sensor element n _msx ) and thus in their output signal in the same Meßwertstufe mj The scheme of the class division is illustrated in Fig. 3. The calibration control unit 6 assumes in this classification regime, as in the previous illumination calibration regime, the constant setting of the integration counter 41 to those needed later Measurement integration time of the measurement regime for image acquisition.

3. characteristic recording regime

In the determination of the characteristic curves of the individual sensor elements ni (as in FIG. 1) or the sensor element classes k | required for the calculation of the corrected measured values (According to FIG. 2), it is assumed that the characteristic curves of the sensor elements η are dependent on the product of illuminance E and integration time t 1, and that these two physical variables are interchangeable. For this reason, it is provided to attenuate the calibration illumination intensity last set by the optical filter device 71 in the optical channel such that at the output of the ADU 2 for the most sensitive sensor element n _max, a predetermined minimum measurement value m is present as the initial value of the desired ideal characteristic. To record the sensor element characteristic, the integration _{clock frequency} f | is determined from the readout cycle to readout cycle, starting with the integration time t _{int of} the CCD line 11 set in the illumination _{calibration regime} and in the classification _regime by means of the programmable integration time _counter 41 _nt gradually reduced. The selection of the step size determines the measuring accuracy and is closely related to the width of the ADU 2. In the embodiment according to FIG. 2, a step size proportional to the sensor read-out frequency f 1 was selected. An unillustrated clock generator thus simultaneously supplies the counting frequency for the integration time counter 41.

At the beginning of the characteristic recording regime, the content of the characteristic memory 42 is set to zero. This memory is synchronized during each read-out cycle by the image sensor element counter 31 with the corresponding sensor element class k | from the classification memory 32 and from the ADU 2 with the corresponding measured value stage mj addressed. The current counter readings ζ of the integration time counter 41 are stored in the read-out memory locations of the characteristic memory 42 in each read-out cycle. If a switching between two adjacent measured value stages mj and mj +1 takes place within a sensor element circuit k ₍ at the output of the ADU 2), the counter reading ζ [mj (k |)] corresponding to this switching point is stored in the characteristic memory 42 at the address mj and k _t Thus, a storage of the counter values ζ of the integration counter 41 corresponding to the characteristic curves of the sensor element classes k |

The characteristic relationship measured value level m as a function of illuminance E and integration time t _{! N} , is therefore reduced to a storage of counter readings ζ.

Parallel to the storage of the integration time counter readings into the characteristic memory 42, the measured value stages m (ki) corresponding to the actual integration times t _{in the} occurrence and the individual sensor element classes kg are given to the comparator unit 33 located in the sensor element indicator 3. In contrast to the illumination calibration regime, these currently incoming measured value steps m are compared with the adaptive characteristic curve end points [m _mox (n _max ) + 1] specified by the reference value transmitter 332. If this end point is reached, the calibration control unit 6 aborts the characteristic recording regime, and in the characteristic memory 42 are for the individual sensor elements ni or

Sensor element classes Ki the interested switching points of the Meßwertstufen rrij the ADU 2 as counts ζ between the predetermined start and end point recorded.

4. Correctional Uprighting Regmo '

The Kon skturwertsberechnungsregimo can contain a variety of operations for the recoding of existing in the characteristic memory 42 counter readings ζ in useful for the subsequent signal processing gray values. Here only one possible variant is to be given by way of example for the mode of operation of the transcoder 43.

At the beginning of the calculation of the corrected measured values, the calibrating control unit 6 applies to the characteristic memory 42 the reference value m, t defined in the illumination calibration regime and the corresponding sensor element class k, (m _mix ) of the (or the) most sensitive sensor element (s) n _max ,

The content of this memory location (maximum counter count ζ of the integration time counter 42 at which the m _m , x is _transited to (m _mtx + 1) for the most sensitive sensor element n _m , _x ) is defined as the end point of the strobed linear characteristic and from the transcoder 43 taken over.

In the following, the calibration control unit 6 within each sensor element class k | the respective adjacent measured value stages ηη and m ₍ + 1 of the characteristic memory 42 addressed and the stored on the corresponding memory locations counter readings ζ Im ₁ ) and z (m ₍ +1), in which in the characteristic recording regime, the switching from the measured value stage rrij to irij +1 and πη ₊ \ was done on nrij + 2, taken in the transcoder 43.

A recoding possibility for the counter readings ζ in a gray value g, which is to be assigned to a measured value level nij to be corrected, is given by the formula

_{g (mi) =} I, j _whole | ₍

where G is a gray value resolution adapted to the subsequent signal processing.

The corrected gray value g calculated in this way or similarly in the transcoder 43 is stored in the memory table of the reference table memory 5 addressed synchronously with the sensor indication (sensor element number or class number) and the measured value step ση.

The addressing of the characteristic curve memory 42 and of the reference table memory 5 required for this correction value calculation regime is performed by the calibration control unit 6, which also passivates the classification memory 32 or the sensor element counter 31 and the ADC 2 and provides for the control of the recoder 43.

Finally, the control operations of the calibration control unit 6 and the line connections connected therewith are to be assigned once again exactly to the individual calibration regimes of the calibration process. It should first be explained that the connections of the calibration control unit 6 with the memories are double lines 61 which allow both the read / write control and the chip activation.

In the illumination calibration regime, the calibration control unit 6 receives the signals of the comparator 331 and, derived therefrom, controls the optical filter device 71. All memories are inactive.

In the classification regime, the classification memory 32 is activated and the reading of the measured value stages m (n;) is controlled by the calibration control unit 6. Characteristic memory 42 and reference table memory 5 remain inactive.

In the characteristic recording regime, the reference table memory 5 remains inactive. The classification memory 32 is read out for addressing the read characteristic curve memory 42. Further, the calibration control unit 6 controls the integration time counter 41 to gradually increase the integration time t | _nt , after the optical filter device 71 has been excited to attenuate the illuminance E.

In the correction value calculation regime, the classification memory 32 (if missing, the sensor element counter 31) and the ADU 2 are passivated. Characteristic memory 42 and reference table memory 5 are activated and are switched to read (characteristic memory 42) and write (reference table memory 5). Furthermore, the calibration control unit 6 takes over the synchronized addressing of both memories as well as the corresponding activation of the transcoder 43.

In the actual measuring regime for image recording, the characteristic memory 42 and the transcoder 43 are then passively switched by the calibration control unit 6, the integration time counter 42 is fixed to the integration time t _inl used in the illumination _calibration and classification regime, and classification memory 32 and reference table memory 5 are active in read mode connected.

Claims (6)

1. Arrangement for the sensitivity correction of integrating sensor arrays with an ADC connected downstream of the sensor array, a measured value-addressed reference table memory and two counters, characterized in that a) that the output of the ADC (2) in addition to the connection to the address input of the reference table memory (5) connects both to the input of a sensor element indicator (3), the first counter as a sensor element counter (31) and a comparator unit (33) for determining has the characteristic points of all sensor elements based on the characteristic curve end point of the most sensitive sensor element (n _m "x), as well as with the address input of a characteristic memory (42) contained in a characteristic transformer (4), b) that the data output of the sensor element indicator (3) is also supplied to the address inputs of the reference table memory (5) and the characteristic memory (42) that the characteristic transformer (4) next to the kenr.ünienspeicher (42) the second counter in the form of a programmable integration time counter ( 41) and a transcoder (43), wherein the count output of the integration time counter (41) is connected to the data input of the characteristic memory (42) and the transcoder (43) is connected between the data output of the characteristic memory (42) and the data input of the reference table memory (5). is arranged c) that a device for varying the calibration input signal size (7) is arranged in the receiving channel in front of the sensor array, and d) that a calibration control unit (6) for read / write control and chip activation, for controlling the integration time counter (41) and the means for varying the calibration input signal size (7) and for switching between the Kalibrierregimeprozessen and the measurement regime for input signal recording is provided, wherein the Kalibriersteuereinheit (6) in the calibration mode temporarily on the address lines of reference table memory and characteristic memory (42) can be switched on and has a connection to the output of the comparator unit (33). 2. Arrangement according to claim 1, characterized in that the sensor element indicator (3) additionally contains a classification memory (32) whose data input to the output of the ADC whose address input to the count output of the sensor element counter (31) is in communication and whose data output to the Output of the sensor element indicator (3) is laid, wherein connections to the calibration control unit (6) for read / write control and chip activation of the memory consist. 3. Arrangement according to claim 1, characterized in that the comparator unit (33) consists of a comparator (331) and a reference value transmitter (332), wherein the reference value transmitter (332) is set to reach the highest level of the ADU. 4. Arrangement according to claim ^, characterized in that the means for varying the calibration input signal quantity (7) is an adjustable optical filter device (71). 5. Arrangement according to claim 1, characterized in that the integration time counter (41) is linked to the readout clock of the sensor array. 6. Arrangement according to claim 1, characterized in that the integration time counter (41) is clocked by pulses of a Laserwegmeßsystems.