JP2001094888A - Image pickup device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To remove a control signal generation circuit to be a complicated circuit requiring much area. SOLUTION: A light receiving part 11 and an operation part 12 are two- dimensionally arranged, respective pixels 14 in the light receiving part and respective pixels 15 in the operation part are allowed to correspond at the rate of 1 to 1, each pixel in the operation part is connected to a corresponding pixel in the light receiving part and pixels adjacent to the corresponding pixel through hardware 13, optical signals from the corresponding pixel and the pixels adjacent to the corresponding pixel are inputted to the pixel concerned in the operation part through the hardware 13, the total signals inputted from the pixels in the operation part are added and the added signal is outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an imaging device,
In particular, the present invention relates to an imaging device (smart sensor) having an intelligent function, and more specifically, to an imaging device that can be suitably used for a CMOS sensor.

[0002]

2. Description of the Related Art Conventionally, a smart sensor has a three-dimensional image pickup device modeled on a human retina, as shown in IE-83-34 of the Institute of Electronics and Communication Engineers, "An Approach to Artificial Retina" or the like. No. 6-139361, Applied Physics Vol. 67, No. 4, (1998) pp424
There is an example in which only the function of the retina is replaced with a current electronic circuit, such as “development and commercialization of an artificial retinal chip”.

In this electronic circuit, a virtual amplifier is built in a pixel, and by applying a control signal in the row or column direction of the amplifier, the amplifier can invert and invert the sign of an input optical signal. . Then, for example, by adding signals of adjacent pixels to be inverted, edge detection and moving object detection can be performed by a known image processing technique. The electronic circuit and the imaging device have various other new functions, and are certainly smart sensors.

Japanese Patent Application Laid-Open No. Sho 6 (1994) discloses an example of a CMOS sensor.
JP-A-3-100879 and JP-A-9-46596.

[0005]

However, in the above-mentioned smart sensor, the pattern of the control signal added to realize the new function is the same in the row or column direction, so that the new function is limited. Join. That is, a too complicated function cannot be realized.

Also, it is necessary to prepare control signal patterns in advance by the number of new functions using wired logic, microcode, or the like. This means an increase in the sensor circuit size.

As described above, the smart sensor enables selection of a new function by changing the pattern of the control signal. However, the smart sensor has a small area to be protected, has a low function, and has a low chip cost of the sensor. It was not necessarily low.

SUMMARY OF THE INVENTION The present invention solves the above-mentioned problem, and learns a complicated function by preliminary learning and realizes it by hardware, thereby eliminating a control signal generation circuit which requires an area and requires a complicated circuit. It is an object of the present invention to provide an imaging device capable of performing such operations.

[0009]

According to the imaging apparatus of the present invention, the light receiving section and the arithmetic section are two-dimensionally arranged, and each pixel in the light receiving section and each pixel in the arithmetic section are one. And each pixel in the arithmetic unit is connected by hardware to a corresponding pixel in the light receiving unit and a pixel adjacent to the corresponding pixel, and the corresponding pixel and the corresponding pixel An optical signal from a pixel adjacent to the pixel is input to the pixel in the arithmetic unit through the hardware, the signals input in the pixels in the arithmetic unit are added, and the added signal is output from the arithmetic unit. It is characterized by that.

A preferred embodiment of the image pickup apparatus according to the present invention has at least two layers, a light receiving layer serving as a light receiving section and a calculation layer serving as a calculation section, and has a pixel array corresponding to each of them. This pixel array is generally two-dimensional, and the array is the three-dimensional I
It is realized not by L but by a general two-dimensional IL which is easy to realize a circuit. In addition, an optical signal from the corresponding light receiving layer pixel and an adjacent light receiving layer pixel is input to one pixel in the arithmetic layer. Each light signal from each of the light receiving layer pixels has one of the following: the normal direction, the sign inverted (inverted), or no influence (zero value) on the arithmetic layer pixel without changing the sign. Is transmitted in the form of The forward, reverse, and zero value conversions are the conversion coefficients,
Alternatively, the amplification factor is a value uniquely determined for a combination of an arbitrary light receiving layer pixel and a calculation layer pixel. In the learning, the value is not changed and each light receiving layer transmitted to the calculation layer pixel is not changed. The converted optical signal from the pixel is fully added in the operation layer pixel and output as a signal.

The selection of the three types of transformations in each pixel combination is determined by a known neurological technique. That is, a neural network circuit whose conversion selection is not determined is prepared, and preliminary learning is performed so as to realize the complicated function. After the function is satisfied, the selection of the conversion between the pixels is examined, and the selected group is realized by hardware for the combination of the pixels on the imaging apparatus of the present invention. In addition, the neurotechnology is described in, for example, “One Approach of Artificial Retina” Shimizu and others, IEICE ED83-
33 pp19, "Simulation of Artificial Retina", Biwaki et al., IEICE IE85-38 pp35, "MST Area Neuron Model for Relative Motion Detection", Kage et al.
The IEICE Technical Report IE94-26 pp21.

According to the present invention, there are provided a light receiving layer and a calculation layer.
Since it has a layer and can consider the correlation between pixels adjacent to each pixel, it is possible to realize more complicated functions than the conventional functions.

Further, the circuit is easily realized on the image pickup apparatus by employing the above-mentioned unique conversion value without using the analog conversion coefficient of the conventional neurotechnology. In addition, by adopting a known neurotechnology, the above functions can be realized by spending a time and a little effort of preliminary learning without having detailed know-how in realizing each function, and thus the design ability is improved. Can be greatly expected.

Although the selection of the conversion is performed by hardware on the image pickup apparatus, only one function can be realized, but an extra circuit such as the control signal generation circuit described above is not required, and accordingly, Chip area saving,
High yield and low cost are possible.

[0015]

Embodiments of the present invention will be described below in detail with reference to the drawings.

(First Embodiment) A first embodiment of the present invention is as follows.
This is an imaging device that detects the edge of the upper right straight line in an image. This imaging device is a CMOS having 64 × 48 pixels, a pixel size of 100 μm square, and a layout rule of 0.4 μm.
It is a sensor.

FIG. 1 shows an equivalent circuit diagram of a part of the pixel array section of this embodiment. In FIG. 1, reference numeral 11 denotes a light receiving layer,
Each pixel 14 is provided with a photodiode (P
D) 16 is arranged. An NMOSFET 17 for reset is connected to the PD 16, and the PD 16 and the NM
OSFET 17 is connected to output terminal 18.

The size of the light receiving layer 11 is 7 × 6.
An arithmetic layer 12 is similarly formed on a silicon chip having a square of ² mm2, and both pixels have the same pixel size.
It is 0 μm square. The pixel components in both layers are 100 μm
They are arranged in a mixture of corners. Operation layer 12
An adder 26 is arranged in the pixel 15 and the input terminal 27 to which the signal from the light receiving layer 11 is input is provided in a number corresponding to the number of corresponding pixels plus the number of adjacent pixels.
The signal input from the input terminal 27 is added by the adder 26, and the sum signal is output to the output terminal 28 in the form of a voltage.

In the present embodiment, the connection between the pixels of the light receiving layer 11 and the arithmetic layer 12 is made by the corresponding pixel (self) 14.
_{This is performed} between ₁ and the adjacent pixels 14 ₂ , 14 ₃ , 14 ₄ , and 14 ₅ of the second nearest rest at the upper left, lower left, upper right, and lower right. The connection method is determined by preliminary learning as described above.

As the hardware 13 for connecting the pixels of the light receiving layer 11 and the arithmetic layer 12, an operational amplifier 23 is prepared for the connection of the normal rotation 20 and the inversion 21, and an empty wiring (open) is prepared for the connection of the zero value 22. .

A light receiving layer 11 and an arithmetic layer 12 and a neural network connecting them are constructed on a computer by a known technique. An image including many straight lines at the upper right is given to a simulator representing an imaging device, and it is visually determined whether or not edge detection is correctly performed on an output image from the simulator. The results are fed back and learning is further advanced.

As a result, depending on the teacher image, a result is obtained in which the upper right is a normal connection and the lower left is a reverse connection.
In this embodiment, learning is performed by determining the connection with itself to a zero value.

The connection result of less than 5 × 64 × 48 = 15360 connections was realized on a silicon chip as an image pickup device using the hardware 13.

The operation of this embodiment will be described below.

The image light incident via an optical system such as a lens forms an image on a silicon chip. PD in pixel 14
Reference numeral 16 detects incident light. At this time, the PD 16 has been reset by the reset MOSFET 17 in advance. The light voltage of the PD 16 is output from the output terminal 18 in real time.

The output terminal 18 is connected to any one of the input terminals 24 of the hardware 13, for example,
In the case of 0 and the inversion 21, it is input to the operational amplifier 23. In the case of normal rotation, the optical signal voltage is (+) of the operational amplifier 23.
The output terminal 25 outputs a voltage having the same sign. In the present embodiment, the amplification gain of the operational amplifier 23 is 1. In the case of inversion, the voltage is similarly input to the (−) terminal of the operational amplifier 23, and a voltage having a different sign is output from the output terminal 25.

The converted optical signal output from the output terminal 25 is input to the input terminal 27 in the operation layer pixel.
The output voltage added in the form of current in the adder 26 and finally added is output from the output terminal 28 of the adder 26.

In this embodiment, the signal output from the output terminal 28 is finally output in the form of an image via a signal processing circuit of a known imaging device.

In the present embodiment, edge detection can be performed by selecting only the upper right straight line from the image.

FIG. 11 is a circuit diagram of a specific pixel of this embodiment. FIG. 11 shows a circuit configuration in which the light receiving layer 11, the arithmetic layer 12, and the hardware 13 are two-dimensionally developed. The hardware 13 is arranged around the PD 16 and the MOSFET 17 of one pixel of the light receiving layer, the adder 26 of the arithmetic layer 12, and the output TFT. Further, another pixel (not shown) of the light receiving layer is arranged around the hardware.

In this embodiment, the output from the PD is a voltage, but the optical signal does not need to be a voltage, but may be a charge or a current. Also, PD16
A transfer gate may be newly provided between the output terminal 18 and the output terminal 18 to output charges.

(Second Embodiment) A second embodiment of the present invention is as follows.
This is an imaging device that performs vertical and horizontal straight line edge detection in an image.

FIG. 2 shows an equivalent circuit diagram of a part of the pixel array section of this embodiment. In FIG. 2, 31 is a light receiving layer,
32 is an operation layer. In the present embodiment, the connection between the pixels of the light receiving layer 31 and the arithmetic layer 32 corresponds to the connection of the corresponding pixel (self).
And 34 _1, carried out between all adjacent pixels. Operation layer 3
2, an adder 46 is formed in the pixel 35.
Nine input terminals 47 are provided. Also, both layers 31,
The hardware for connecting the 32 pixels is the same as in the first embodiment.

In the pre-learning, similarly, an image including a large number of vertical and horizontal lines is given to an imaging device simulator including a neural network for education. As a result, depending on the teacher image, a result is obtained in which mainly the upper and right are forward connections and the lower and left are reverse connections.

[0035] Incidentally, in this embodiment performs learning decided to zero value the connection between the pixel 34 ₁ itself.

In the same manner as in the first embodiment, the connection result was realized on a silicon chip as an image pickup device using the hardware.

According to this embodiment, edge detection can be performed by selecting only vertical and horizontal straight lines from an image.

In the present invention, the light receiving layer pixels connected to the operation layer pixels are not limited to first (left and right and up and down) and second near rest (left and up and down and right and up and down) pixels. It may be a pixel.

Third Embodiment FIG. 3 shows a third embodiment of the present invention in which third nearest pixels are connected. According to this embodiment, more complicated functions can be realized. FIG.
In, 54 _{1-54 4} shows (put one pixel, the left and right and up and down) third nearest pixel. 51 is a light receiving layer, 52
Denotes a calculation layer, and 53 denotes hardware for connecting the light receiving layer 51 and the calculation layer 52, and has the same configuration as that of the first embodiment.
Although only one pixel 55 in the calculation layer 52 is shown in the figure, this is because other pixels 55 are omitted. This is the same in other embodiments. A signal from the pixel 54 up to the third nearest is input to one pixel 55 shown.

(Fourth Embodiment) A fourth embodiment of the present invention is an imaging apparatus for detecting only a specific character from the alphabet. This imaging device is a CMOS sensor having 42 × 50 pixels, a pixel size of 100 μm square, and a layout rule of 0.4 μm.

FIG. 4 shows an equivalent circuit diagram of a part of the pixel array section of this embodiment. In FIG. 4, reference numeral 71 denotes a light receiving layer;
A PD 76 is formed in each pixel 74. An NMOSFET 79 for reset is connected to PD16,
D16 and NMOSFET 79 are connected to output terminal 78. Similarly, an adder 86 is formed in each pixel 75 in the arithmetic layer 72, and an input terminal 8 of the adder 86 is provided.
7, a total of nine output terminals 78 of the corresponding pixel and eight adjacent pixels are connected to each other through the connection hardware 73 determined by the preliminary learning.

All outputs from the operation layer pixels 75 are input to an input terminal 90 of an adder 89 on the same chip. All the signals output from all the pixels 75 are added by an adder 89 to form one addition signal. The addition signal is compared with a predetermined threshold voltage by a comparator (not shown). If the addition signal is larger than the threshold voltage, it is determined that the detected alphabet is a specific character.

The preliminary learning is performed as follows.
Now, for the sake of simplicity, let the specific character be "A".
This can of course be any other character.

The character size is set to about 30 × 36 of the pixel array of the above-mentioned image pickup apparatus, and a teacher image is prepared in consideration of the displacement due to the reading (FIG. 5). The displacement here is one pixel of the pixel array pitch, ± 100 μm
It is about.

The image "A" 102 is repeatedly shown to the imaging apparatus simulator, and learning is performed until the sum of the output signals from each pixel of the simulator becomes higher than a specific threshold value and the image of another character becomes lower than the specific threshold value. Is repeated. As a result of learning,
The connection of the pixel at the position corresponding to the black portion 103 of the image “A” 102 is reversed and the connection of the pixel at the position corresponding to the white portion 104 is a normal connection.

An image "A" 10 having a size of 30.times.36
2 outside margin area 105 (42 × 50−30 × 3
6) is prepared in consideration of unexpectedly large characters and extremely large displacement. Reference numeral 101 in FIG. 5 indicates the entire teacher image.

According to the present embodiment, the signal from the light receiving layer pixel adjacent to the arithmetic layer is evaluated instead of extracting and evaluating the signal from the arithmetic layer pixel. It is possible to detect a specific character that is strong even when it is strong.

As described above, if a character has many diagonal lines, learning is performed on diagonal lines before learning of a specific character is started, thereby further improving the learning efficiency and detection accuracy. Is improved.

(Fifth Embodiment) A fifth embodiment of the present invention is as follows.
This is an imaging device that detects only a specific fingerprint from fingerprints. This imaging device is a CMOS having 512 × 512 pixels, a pixel size of 30 μm square, and a layout rule of 0.25 μm.
It is a sensor.

In this embodiment, the structures of the light receiving layer and the operation layer are the same as those of the above-described embodiment. Signals from eight pixels close to the pixel itself are input to the pixels in the operation layer via the connection hardware.

The signals from all the pixels of the operation layer are similarly fully added by the adder, and if the signal is equal to or higher than a predetermined threshold voltage, it is recognized as a specific fingerprint.

However, the situation is slightly different for the preliminary learning.

First, a specific fingerprint and an image of another fingerprint are prepared, and an image in which the characteristic ridge end point, branch point, and the like are independently extracted from the specific fingerprint is prepared.
Learning in advance for the images that become the respective parts,
After confirming that the features have been sufficiently detected, the process moves to learning of the entire fingerprint for the first time.

The detection of the end points and the branch points can be detected by a known image processing technique, and is usually performed by a detection mask of 3 × 3 pixels. According to the present embodiment, 3 × 3 = 9 signals are similarly input to the pixels of the calculation layer at the positions corresponding to the features, and the signals are analog signals instead of binary signals. Detection can be performed with high accuracy.

According to this embodiment, since a specific fingerprint can be easily detected, an inexpensive fingerprint detection system and a fingerprint tablet can be realized. Further, the present embodiment can be easily applied to the detection of other fixed and significant patterns, such as the detection of a sash of a bill.

(Sixth Embodiment) A configuration of a calculation layer of an image pickup apparatus according to a sixth embodiment of the present invention will be described. FIG. 6 shows an equivalent circuit diagram of a part of the pixel array section of the present embodiment. In the present embodiment, the light receiving layer 111 connected to the pixel 115 of the arithmetic layer 112
Pixel 114 is a total of five pixel itself 114 ₁ and second nearest pixel 114 _{2-114 5.}

In the pixel 115 in the operation layer 112, a total of five operational amplifiers 123 are already arranged in addition to the adder 126.

In the present embodiment, the connection hardware may be merely wiring. That is, the normal connection determined by learning is connected to the output terminal 11 of the light-receiving layer pixel 114.
8 may be connected to the (+) terminal 125 of the operational amplifier 123, and the ground 130 may be connected to the (-) terminal 124. In the case of the inversion connection, the connection between the (+) terminal 125 and the (-) terminal 124 is reversed. Output terminal 1 in case of zero value connection
18 is open, and both terminals 12
4, 125 may be connected to the ground 130 and short-circuited.

According to the present embodiment, since the data of the connection group determined by the preliminary learning of the imaging device simulator is merely converted into a simple routing of the wiring, furthermore, I
Design of C becomes easy. For example, the configuration is more suitable for automatic design such as a known silicon compiler.

The photoelectric conversion element included in the light-receiving layer of the present invention is not limited to a PD (photodiode), and a known photoelectric conversion element such as a photogate or a phototransistor may be used.

The connection hardware of the present invention is not limited to the one using the illustrated operational amplifier, but may be a known circuit capable of inverting, inverting, and zeroing the sign of the signal from the photoelectric conversion element. .

The imaging apparatus simulator used for the preliminary learning may be not only software realized on a personal computer but also an electric circuit realized by hardware, a simulator or the like. In other words, it is only necessary that the device has a function that can determine the connection group of the connection hardware.

The adders arranged in the operation layer are not limited to those shown in the figure, but may be any known circuits capable of performing addition.

The circuit connected to the output side of the arithmetic layer is not limited to a known image pickup device, but may be a known general circuit that does not accompany an image output as shown in FIG.

(Seventh Embodiment) A connection hardware of an image pickup apparatus according to a seventh embodiment of the present invention will be described. FIG. 7 is an equivalent circuit diagram of a part of the pixel array unit of the present embodiment. 7, reference numeral 133 denotes a pixel 134 of the light receiving layer 131 and the operation layer 13
Hardware for connecting the two pixels 135, and 14
Numeral 0 is a hardware component for connecting the pixel 135 of the operation layer 132 and the corresponding pixel, and 141 is a hardware component for connecting a pixel adjacent to the corresponding pixel. While the value of the feedback resistor attached to the operational amplifier 143 is 4R in the hardware configuration section 140, it is R in the hardware configuration section 141. For this reason, the hardware configuration unit 140
The amplification factor of the operational amplifier 143 is 1 in the hardware configuration unit 141 while the amplification factor of 3 is 4.

In this embodiment, since the optical signal from the corresponding pixel is amplified and transmitted four times and the optical signal from the adjacent pixel is transmitted without amplification, the output of the pixel in the arithmetic layer is output. Represents the optical signal of the corresponding pixel strongly reflected.

Also, in the known image processing technique, as can be seen by looking at various secondary differential masks, the weight is different between the center and the adjacent outer periphery (FIG. 1).
0). Therefore, the above-described second-order differential effect can be obtained by using this embodiment. (Eighth Embodiment) A connection hardware of an imaging apparatus according to an eighth embodiment of the present invention will be described. In the present embodiment, the amplification factor of the operational amplifier that is the connection hardware is different between the hardware component connected to the pixel itself and the hardware component connected to the first nearest and second nearest pixel, As shown in FIG. 8, the operational amplifier of the hardware component connected to the pixel itself (center in FIG. 8) is 4, and the operational amplifier of the hardware component connected to the first nearest pixel (left and right and up and down in FIG. 8)
In the operational amplifiers (upper left and lower right and upper and lower right in FIG. 8) of the hardware component connected to the second nearest pixel, the value is 1.

The connection of the operational amplifiers having different amplification factors is determined by the preliminary learning as described above. as a result,
By using a teacher image that expects the second derivative effect,
In the pixel itself and the pixel of the second nearest rest, forward connection,
Inverted connection is made for the first nearest pixel.

Even when this embodiment is used, the same as in the seventh embodiment,
The following differential effect can be obtained.

According to the present invention, a pixel having a Gaussian type spatial sensitivity characteristic which is strongly influenced by the center pixel (self)
Pixels having an antisymmetric spatial sensitivity characteristic such that the sum of the sensitivities in a certain region becomes zero can be realized by appropriately selecting a teacher image in the learning. This allows
A very wide range of image processing can be realized on the imaging device. (Ninth Embodiment) Connection hardware of an imaging apparatus according to a ninth embodiment of the present invention will be described. As shown in FIG. 9, eight operational amplifiers 183 are prepared in advance in addition to the five light receiving layer pixels 174 to be connected in advance, and the number of hardware for connecting the light receiving layer pixel 174 and the arithmetic layer pixel 175 depends on the number of hardware. The amplification factor of the combination of pixels to be connected (between pixels) can be changed.

That is, the operational amplifier 183 is connected to the light receiving layer pixel 17
By receiving the voltage input from the input terminal 4 at the input terminal 184, the input voltage which is an optical signal is converted into an output current. The adder 186 receives the output current from the operational amplifier 183, and outputs a signal obtained by fully adding the output current from the output terminal 188 in the form of a voltage. Therefore, if the number of connected hardware is increased, the number of currents output from the operational amplifier 183 increases, and the total amount of current input to the adder 186 increases. As a result, the amplification factor increases.

In this embodiment, the connection of the hardware is determined by preliminary learning.

[0073]

As described above, according to the present invention,
A smart sensor having a complicated function can be easily realized by preparing a light receiving layer and a calculation layer and connecting the two pixels with hardware. Further, since there is no extra circuit, the sensor is inexpensive and a sensor specialized for a specific application can be supplied.

Further, the sensor has good compatibility with the known neuro technology, and can provide a sensor having a future.

[Brief description of the drawings]

FIG. 1 is an equivalent circuit diagram of a part of a pixel array unit according to a first embodiment of the present invention.

FIG. 2 is an equivalent circuit diagram of a part of a pixel array unit according to a second embodiment of the present invention.

FIG. 3 is an equivalent circuit diagram of a part of a pixel array unit to which a third nearest pixel according to a third embodiment of the present invention is connected.

FIG. 4 is an equivalent circuit diagram of a part of a pixel array unit according to a fourth embodiment of the present invention.

FIG. 5 is a diagram showing a teacher image in which some positional deviation due to reading is also considered.

FIG. 6 is an equivalent circuit diagram of a part of a pixel array section according to a sixth embodiment of the present invention.

FIG. 7 is an equivalent circuit diagram of a part of a pixel array unit according to a seventh embodiment of the present invention.

FIG. 8 is a diagram illustrating connection hardware of an imaging apparatus according to an eighth embodiment of the present invention.

FIG. 9 is a diagram illustrating connection hardware of an imaging apparatus according to a ninth embodiment of the present invention.

FIG. 10 is a diagram illustrating weighting of a center and an outer periphery adjacent thereto in a second-order differential mask.

FIG. 11 is a circuit configuration diagram of a specific pixel of the first embodiment of the present invention.

[Explanation of symbols]

 Reference Signs List 11 light receiving layer 12 operation layer 13 hardware 14 and 15 pixels 16 photodiode (PD) 17 NMOSFET for reset 18 output terminal 19 reset terminal 20 forward rotation component 21 inversion component 22 zero value component 23 operational amplifier 24 input terminal 25 output Terminal 26 Adder 27 Input terminal 28 Output terminal

Claims (7)

[Claims] 1. A light-receiving unit and a calculation unit are two-dimensionally arranged, and each pixel in the light-receiving unit and each pixel in the calculation unit correspond to 1: 1. Is connected to a corresponding pixel in the light receiving unit and a pixel adjacent to the corresponding pixel by hardware, and an optical signal from the corresponding pixel and a pixel adjacent to the corresponding pixel passes through the hardware. An image pickup apparatus, wherein a signal input to a pixel in the operation unit, a signal input in the pixel in the operation unit is added, and the added signal is output from the operation unit. 2. The image pickup apparatus according to claim 1, wherein the arrangement of the pixels of the light receiving section and the pixels of the operation section is a two-dimensional arrangement. 3. The image pickup apparatus according to claim 1, wherein the hardware includes a plurality of components, and each of the components sets a code of the input optical signal in a combination of arbitrary pixels of the light receiving unit. An imaging apparatus having a conversion function of not inverting, inverting the sign of the input optical signal, or converting the input optical signal to a zero value. 4. The imaging device according to claim 3, wherein the selection of the function of the hardware is performed in advance by learning using a neurotechnology. 5. The imaging device according to claim 4, wherein each coefficient or amplification factor of the three functions for converting the optical signal is unique to a combination of the arbitrary pixels.
An imaging apparatus wherein the value is not changed by the learning. 6. The imaging apparatus according to claim 5, wherein the unique value is determined by a degree of proximity between the calculation unit pixel and the light receiving unit pixel. 7. The image pickup apparatus according to claim 3, wherein the number of connected hardware is greater than the number of pixels of the light receiving unit, thereby increasing an amplification factor of an optical signal of a pixel of the light receiving unit. An imaging device characterized by the above-mentioned.