Imaging System
The imaging system addresses the limitations of color filters by using a solid-state device without filters and AI-enhanced color inference, achieving clear and faithful color images in low-light scenarios.
Patent Information
- Application Number
- JP2021566374
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2019-12-27
- Filing Date
- 2020-12-14
- Publication Date
- 2024-12-09
- Estimated Expiration
- 2040-12-14
AI Technical Summary
Conventional imaging devices using color filters attenuate light, leading to reduced visibility and color inaccuracies, especially in low-light conditions, making it difficult to capture clear and faithful color images.
An imaging system that utilizes a solid-state imaging device without a color filter, combined with an AI system using a neural network to infer and adjust colors based on learned feature values, employing super-resolution processing and a reduced dataset for efficient colorization.
Enables high-visibility, vivid color images even in low-light conditions, improving image clarity and color fidelity without the need for external light sources.
Smart Images

Figure 0007596305000003 
Figure 0007596305000004 
Figure 0007596305000005
Abstract
Description
[Technical field]
[0001] One aspect of the present invention relates to a neural network and an imaging system using the same. Another aspect of the present invention relates to an electronic device using the neural network. Another aspect of the present invention relates to a vehicle using the neural network. Another aspect of the present invention relates to an imaging system that obtains a color image from a black-and-white image obtained by a solid-state imaging element using an image processing technique. Another aspect of the present invention relates to a video surveillance system, a security system, a safety information providing system, or a driving support system using the imaging system.
[0002] Note that one embodiment of the present invention is not limited to the above technical field. One embodiment of the present invention disclosed in this specification relates to an object, a method, or a manufacturing method. One embodiment of the present invention relates to a process, a machine, a manufacture, or a composition of matter. Therefore, examples of the technical field of one embodiment of the present invention disclosed in this specification include a semiconductor device, a display device, a light-emitting device, a power storage device, a memory device, an electronic device, a lighting device, an input device, an input / output device, a driving method thereof, or a manufacturing method thereof.
[0003] In this specification, the term "semiconductor device" refers to any device that can function by utilizing semiconductor characteristics, and electro-optical devices, semiconductor circuits, and electronic devices are all classified as semiconductor devices. [Background technology]
[0004] Conventionally, a technology is known in which an image captured by an image sensor is colorized by using a color filter. Image sensors are widely used as imaging components in digital cameras, video cameras, and the like. Image sensors are also used as part of security equipment such as security cameras, and therefore such equipment must capture accurate images not only in bright places during the day, but also at night or in dark places with little lighting, requiring an image sensor with a wide dynamic range.
[0005] Furthermore, there has been remarkable progress in technology using AI (Artificial Intelligence). For example, there has been active development of automatic coloring technology that uses AI to colorize black-and-white photos taken on old photographic film. Known techniques for colorization using AI include learning from large amounts of image data, generating a model, and then achieving colorization through inference using the resulting generative model. Machine learning is also a part of AI.
[0006] When it comes to colorization, learning with a large amount of image data does not necessarily result in an optimal color image; the more image data that is learned, the more likely it is to become sepia or gray, resulting in insufficient coloring.
[0007] A technology for forming a transistor using an oxide semiconductor thin film formed over a substrate has attracted attention. For example, Patent Document 1 discloses an imaging device in which a transistor that includes an oxide semiconductor and has an extremely low off-state current is used in a pixel circuit.
[0008] Moreover, a technique for adding a calculation function to an imaging device is disclosed in Patent Document 2. Moreover, a technique relating to super-resolution processing is disclosed in Patent Document 3. [Prior art documents] [Patent documents]
[0009] [Patent Document 1] JP 2011-119711 A [Patent Document 2] JP 2016-123087 A [Patent Document 3] JP 2010-262276 A Summary of the Invention [Problem to be solved by the invention]
[0010] Color images captured by conventional imaging devices such as image sensors use color filters. Image sensors are sold with color filters attached, and the image sensor is appropriately combined with lenses and other components to be mounted on electronic devices. Simply placing a color filter over the light receiving area of an image sensor reduces the amount of light that reaches the light receiving area. Therefore, attenuation of the amount of light received by the color filter is unavoidable. Various companies combine color filters with image sensors, and although optimization is attempted, they do not pinpoint the wavelength of light that passes through, but rather receive broad light in a certain wavelength range.
[0011] For example, in a dimly lit evening, at night, or in bad weather, a camera mounted on a vehicle may have difficulty recognizing distant objects due to insufficient light. Since it is difficult for a driver to recognize objects in a dark image, colorizing the objects results in an image that is emphasized, which is preferable.
[0012] For example, in the case of security cameras used in surveillance systems, there is a problem that when shooting in dim light, there is insufficient light to obtain images that are good enough for facial recognition. Security camera installers prefer color images to black-and-white images taken with infrared cameras.
[0013] For example, when taking pictures underwater, there are many places where there is not enough light, and a light source is necessary in deep water, but when trying to photograph fish, the light source may cause the fish to escape.In addition, since light does not reach the sea easily, it is difficult to photograph distant fish even with a light source.
[0014] One of the objects of the present invention is to provide an imaging method and an imaging system that can obtain an image with high visibility and faithful colors to the actual colors without using a color filter.
[0015] In particular, when capturing images in dark places with little lighting, such as in the evening or at night, there is a problem that it is difficult to capture clear images because the amount of light received is small. Therefore, images captured in dark places have poorer visibility than those captured in bright places.
[0016] Security cameras placed in light environments with narrow wavelength ranges need to accurately grasp the characteristics of the objects captured in the captured images in order to accurately understand the situation in the captured images as clues leading to incidents or accidents, etc. Therefore, for example, in the case of night vision cameras, it is important to focus on the object and capture an image with good visibility of the object, even in a dark place.
[0017] Some night vision cameras use an infrared light source and separate colors with a special color filter to obtain color images. However, because they use reflected infrared light, there is a problem that some subjects may not reflect color or may be displayed in a different color. This problem often occurs when the subject is made of a material that absorbs infrared rays. For example, human skin tends to be captured as whiter than it actually is, and warm colors such as yellow may be captured as blue.
[0018] One of the objects of the present invention is to provide an imaging method and an imaging system that have high visibility even when imaging at night or in a dark place, and can obtain images that are faithful to the colors of images captured in the presence of external light. [Means for solving the problem]
[0019] The imaging system of the present invention includes a solid-state imaging device that does not have a color filter, a storage device, and a learning device. Since no color filter is used, light attenuation can be avoided and imaging with high sensitivity can be performed even with a small amount of light.
[0020] Since there is no color filter, the obtained black and white image data (analog data) is colorized, but the coloring is performed using an AI system. The AI system, i.e., a learning device that uses teacher data stored in a storage device, uses the extracted features to perform inference and adjust the focus, making it possible to obtain a color image (colorized image data (digital data)) with high visibility even when imaging at night or in a dark place. The learning device includes at least a neural network unit, and can perform inference as well as learning and output data. The learning device may also perform inference using features that have been learned in advance. In that case, the learned features are stored in the storage device and calculations are performed, making it possible to output data at the same level as when the learned features are not used.
[0021] Furthermore, if the amount of light reaching the subject is low and part of the subject's outline becomes unclear, the boundary cannot be distinguished, and coloring of that part may be incomplete.
[0022] Therefore, it is preferable to use an image sensor without a color filter to obtain black and white image data with a wide dynamic range, and to repeat the super-resolution process multiple times, and then to distinguish and color the color boundaries. Also, the super-resolution process may be performed at least once on the teacher data. Note that the super-resolution process refers to image processing that generates a high-resolution image from a low-resolution image.
[0023] In addition, when performing image processing using multiple types of neural networks, the order in which they are performed is also important, and it is preferable to perform calculations using black-and-white image data and then perform colorization as the final image processing. Black-and-white image data has a smaller data volume than image data containing color information, which reduces the burden on the processing capacity of the calculation device.
[0024] Furthermore, if a learning model is prepared in advance, relatively bright black-and-white image data can be captured by imaging without using a flash light source in situations where there is insufficient light, and vividly colored image data can be obtained by colorizing the black-and-white image data.
[0025] The configuration of the invention disclosed in this specification is an imaging system having a solid-state imaging element without a color filter, a memory device, and a learning device, in which the solid-state imaging element acquires black and white image data, and the training data used in the learning device is a second set that has been converted to HSV color space and thinned out based on statistics of a first set that has been set to the median value of the saturation histogram, and the learning device uses the training data stored in the memory device to colorize the black and white image data and create colorized image data.
[0026] In the above configuration, the solid-state imaging element is a back-illuminated CMOS image sensor chip.
[0027] Furthermore, in the case of an imaging system mounted on a vehicle, since only a certain amount of power can be used, it is difficult to perform relatively large-scale calculations using a CPU mounted on the vehicle, etc. In particular, if the learning data is huge, it may take a lot of time or effort to learn the learning data.
[0028] In this specification, a learning algorithm known as GAN (Generative Adversarial Network) is used for colorization learning.
[0029] For example, there is a training dataset for colorization (ILSVRC2012 (ImageNet Large Scale Visual Recognition Challenge 2012)), but this dataset contains 26,646 color images, and there is a risk that training using this dataset when installed in a vehicle will take an enormous amount of time.
[0030] Furthermore, when this learning data is used to colorize an image using a neural network, the resulting image often has colors that differ from natural colors.
[0031] Therefore, when using the ILSVRC2012 dataset, rather than using all 26,646 images as training data, selected images are used to obtain images that are closer to natural coloring. The selection method uses the histogram of S (satuation) in the HSV color space. The HSV color space is expressed using three parameters: hue, saturation, and value. Since the HSV color space is a common method of expressing color, details will be omitted here.
[0032] In the HSV color space, saturation is indicated by the distance from the center point of the color wheel, and its value is indicated as 0 to 1.0. Note that when the saturation is 1.0, it is a pure color, and when the saturation is 0, it loses vividness and becomes gray overall.
[0033] When using the ILSVRC2012 dataset, image saturation is used to classify images, and by limiting the range to a statistically desirable range, it is possible to reduce the amount of training data and output vivid images with colors that are closer to natural.
[0034] Its feature is that saturation in the HSV color space is used as a criterion for selecting the training data to reduce it, and the saturation is selected to have the optimal distribution. By converting the data set based on saturation and selecting the training data using a certain criterion, and reducing the amount of data (number of photos) used for training, training can be completed in a short time, and a vivid image can be obtained when colorizing black-and-white images to color images compared to when all training data is used.
[0035] In this specification, the saturation of an image is defined as the median value when all pixels of an image are plotted in a saturation histogram. The saturation of an image is normalized to 256 levels.
[0036] For comparison, if the saturation range is limited and colorization is performed using only biased training data, for example, training data with a median saturation between 175 and 256, the output color image will also be biased, resulting in unnatural colors. It is difficult to achieve vivid colorization by simply using only highly saturated, vivid image data as training data.
[0037] Another configuration of the invention disclosed in this specification is a driving assistance system equipped with a camera in a vehicle, the camera having a solid-state imaging element without a color filter, a learning device, and a memory device, the training data captured by the solid-state imaging element and used in the learning device is a second set that has been converted into HSV color space and thinned out based on statistics of a first set that is the median of a saturation histogram, and the driving assistance system executes a software program that creates colorized image data by inference from the learning device using the training data in the memory device.
[0038] In each of the above configurations, the set of median values of the saturation histogram is classified into 256 gradations, and data of gradations of 175 or more are not used as training data. In each of the above configurations, the second set can be 1 / 250 or less of the first set.
[0039] A video monitoring system, a security system, or a safety information providing system using the above-mentioned imaging system can clearly realize imaging in a relatively dark place.
[0040] Specifically, it is a surveillance system equipped with a security camera, which has a solid-state imaging element without a color filter, a learning device, and a memory device, and while the security camera detects a person, the solid-state imaging element captures an image and executes a software program that uses the teacher data in the memory device to create colorized image data through inference by the learning device.
[0041] Also, the present invention relates to a driving assistance system using a camera mounted on a vehicle, that is, a so-called on-board camera, and the on-board camera has a solid-state image sensor without a color filter, a learning device, and a storage device, and while the on-board camera detects an object far away in the traveling direction, the solid-state image sensor captures an image, and executes a software program that creates colorized image data by inference of the learning device using pre-selected learning data. In the driving assistance system, it is preferable that the image obtained by the on-board camera is colorized more vividly than natural colors, because this increases the driver's recognition. Therefore, by performing the colorization disclosed in this specification, not only colorization but also object highlighting processing can be performed at the same time.
[0042] Furthermore, when colorizing a black-and-white image by software, the program constituting the software may be installed from a computer having a built-in hardware, or from a network or a recording medium. A program recorded on a recording medium such as a computer-readable CD-ROM (Compact Disk Read Only Memory) is installed, and the program for colorizing a black-and-white image is executed. The processing performed by the program is not limited to processing in a sequence, and does not have to be chronological, and may be performed in parallel, for example. Effect of the Invention
[0043] The imaging system disclosed herein allows for clear or vivid colorized images to be obtained even in low light or dim conditions.
[0044] Therefore, in the case of a security camera, it is relatively easy to identify a person (such as a face) or distinguish the characteristics of their clothing based on the obtained colorized image. By applying this imaging system to a security camera, it is possible to capture a person's face in a dark place using black and white image data, estimate the image, colorize it, and output it to a display device for color display.
[0045] In particular, when 8K-size input image data is obtained by a solid-state imaging element in order to improve imaging accuracy, the light receiving area of the solid-state imaging element arranged in each pixel becomes narrow, so the amount of light obtained is reduced. However, in the imaging system disclosed in this specification, since a color filter is not used in the solid-state imaging element, there is no reduction in the amount of light due to the color filter. This makes it possible to capture 8K-size image data with good sensitivity.
[0046] Furthermore, in the case of an on-board camera, the imaging system disclosed in this specification makes it possible to colorize and highlight images at the same time. [Brief description of the drawings]
[0047] FIG. 1 is a block diagram showing one embodiment of the present invention. Figure 2A shows an example of image data, Figure 2B is a graph showing saturation normalized to 256 levels on the horizontal axis and frequency on the vertical axis for a color image with a median saturation of 58, and Figure 2C is a graph showing saturation normalized to 256 levels on the horizontal axis for a color image with a median saturation of 90. FIG. 3 is a graph showing the number of sheets on the vertical axis and the number of gradations divided into classes on the horizontal axis, and a diagram showing an approximation curve. FIG. 4 is an example of a flow diagram illustrating one embodiment of the present invention. FIG. 5 is an example of a flow diagram illustrating one embodiment of the present invention. FIG. 6 is a block diagram showing an example of the configuration of the imaging unit. FIG. 7 is a diagram showing an example of the configuration of a pixel block 200 and a circuit 201. As shown in FIG. FIG. 8 is a diagram illustrating an example of a pixel configuration. 9A to 9C are diagrams showing filters. Fig. 10A is a diagram showing a configuration example of a pixel, and Figs. 10B to 10D are diagrams showing configuration examples of a photoelectric conversion device. FIG. 11 is a cross-sectional view showing an example of the configuration of an imaging device. 12A to 12C are cross-sectional views showing examples of the structure of a transistor. FIG. 13 is a cross-sectional view showing an example of the configuration of an imaging device. FIG. 14 is a cross-sectional view showing an example of the configuration of an imaging device. 15A1 to 15A3 and 15B1 to 15B3 are perspective views of a package and a module that house an imaging device. FIG. 16A is a diagram showing the external appearance of a vehicle, and FIG. 16B is a schematic diagram of the vehicle as viewed from inside the vehicle toward the front windshield. FIG. 17 shows an example of an applied product showing one embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0048] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. However, the present invention is not limited to the following description, and it will be easily understood by those skilled in the art that the modes and details of the present invention can be modified in various ways. Furthermore, the present invention is not to be interpreted as being limited to the description of the embodiments shown below.
[0049] (Embodiment 1) An example of the configuration of an imaging system 21 used in a video monitoring system, a security system, or a driving support system will be described with reference to the block diagram shown in FIG.
[0050] The data acquisition device 10 is a semiconductor chip including a solid-state imaging element 11 and a memory unit 14, and does not have a color filter. The data acquisition device 10 has an optical system such as a lens. Note that the optical system may have any configuration as long as the imaging characteristics are known, and is not particularly limited.
[0051] For example, the data acquisition device 10 may be a single semiconductor chip formed by stacking a back-illuminated CMOS image sensor chip, a DRAM chip, and a logic circuit chip. Also, a single semiconductor chip may be formed by stacking a back-illuminated CMOS image sensor chip and a logic circuit chip including an analog-to-digital conversion circuit, in which case the memory unit 14 is an SRAM. Also, the stacked chips may be stacked and electrically connected to each other using a known bonding technique.
[0052] The memory unit 14 is a circuit that stores the converted digital data, and is configured to store the data before inputting it to the neural network unit 16, but is not limited to this configuration.
[0053] The image processing device 20 is a device for estimating contours or colors corresponding to a black-and-white image obtained by the data acquisition device 10. The image processing device 20 is executed in a first stage for learning and a second stage for estimation. In this embodiment, the data acquisition device 10 and the image processing device 20 are configured as separate devices, but it is also possible to configure the data acquisition device 10 and the image processing device 20 as an integrated device. When configured as an integrated device, it is also possible to update the feature amount obtained by the neural network unit in real time.
[0054] The neural network unit 16 is realized by software calculations using a microcontroller. A microcontroller is a computer system built into a single integrated circuit (IC). When the scale of calculations or the amount of data to be handled is large, the neural network unit 16 may be configured by combining multiple ICs. The learning device includes at least these multiple ICs. Also, a microcontroller equipped with Linux (registered trademark) is preferable because it allows the use of free software, thereby reducing the total cost for configuring the neural network unit 16. Also, it is not limited to Linux (registered trademark), and other OSs (operating systems) may be used.
[0055] The learning of the neural network unit 16 shown in Fig. 1 will be described below. In this embodiment, learning is performed in advance, and processing by the neural network is performed using weights. Since the amount of teacher data for learning can be reduced, it can be stored in the storage unit 18 and learned as appropriate.
[0056] It is preferable to use a free training set (also called a training data set) for the training data, since this can reduce the total cost.
[0057] In this embodiment, it is possible to use a reduced amount of selected data from ILSVRC2012.
[0058] The data is selected using the HSV color space. The HSV color space is preferable because it allows for quantification in a state close to human vision. The median value of a histogram of all pixels in an image is taken as the saturation of the image, and the saturation of the image is normalized to 256 levels.
[0059] Here, we present an example of normalized data. Figure 2A shows a single black-and-white image. Figure 2B shows the histogram of the image of Figure 2A, which is colored under a first condition. Figure 2C shows the histogram of the image of Figure 2A, which is colored under a second condition different from the first condition.
[0060] The second condition is perceived by the human eye as a more vivid image, and the higher the median value of the histogram, the more vivid the image can be quantified.
[0061] Figure 3 shows the results of normalizing the ILSVRC2012 data using the above method and classifying it by image saturation. The median saturation is divided into 25 parts for classification. The results are also shown in Table 1.
[0062] [Table 1]
[0063] Furthermore, by carrying out the above class classification, an approximation curve can be found from these data. Figure 3 shows the approximation curve, and by selecting each range of gradations (classes) based on this approximation curve, it is possible to thin out and reduce the number of images used for learning. An example of the total number of images used for learning is shown in Table 2.
[0064] [Table 2]
[0065] As shown in Table 2, data between 175 and 256 are deleted, and the remaining classes are determined by an approximation curve. For example, if the total number of sheets is 10,000, then sorting can be done for each class. If the total number of sheets is 100 (approximately 1 / 250), then 17 sheets with gradation numbers between 0 and 25 and less than 25, 22 sheets with gradation numbers between 25 and 50, 21 sheets with gradation numbers between 50 and 75 and less than 75, 17 sheets with gradation numbers between 75 and 100, 12 sheets with gradation numbers between 100 and 125, 7 sheets with gradation numbers between 125 and 150, and 5 sheets with gradation numbers between 150 and 175. Based on the values when the total number of sheets is 100, the ratios can also be sorted out as percentages. These numbers are also standards, and the numbers should be in the range of 0.5 or 2 times that of each class.
[0066] The learning device performs inference using these training data for learning. The learning device may have any configuration as long as it can perform inference using the training data in the storage unit 18 based on black-and-white image data and output colorized image data. When using pre-trained features, the amount of data and the amount of calculation are reduced, which has the advantage that the learning device can be configured with a small configuration, for example, one or two ICs.
[0067] The colorization can use the following convolutional neural network (also called artificial neural network):
[0068] The program is created using Python under a Linux (registered trademark) operating environment. A GAN (Generative Adversarial Network) is used with a neural network as the generator.
[0069] The output data of the neural network unit 16 is linked to the time data of the time information acquisition device 17 and stored in the large-scale storage device 15. The data obtained from the start of imaging is accumulated and stored in the large-scale storage device 15. Note that although the term "large-scale storage device 15" is used, it does not refer to a large server, but rather to, for example, a security camera monitoring device or a small-sized large-capacity storage device (e.g., SSD or hard disk) that can be mounted on a vehicle. The time information acquisition device 17 may be a device that acquires time information by GPS.
[0070] The display unit 19 may be provided with an operation input unit such as a touch panel, allowing the user to select and appropriately observe data from among the data stored in the large-scale storage device 15. The imaging system 21 can be used in a video surveillance system, a security system, or a driving assistance system.
[0071] Moreover, the display unit of a mobile information terminal (such as a smartphone) of the user can also be used as the display unit 19. By accessing the large-scale storage device 15 from the display unit of the mobile information terminal, it is possible to monitor the user regardless of his or her location.
[0072] The installation of the imaging system 21 is not limited to the wall of a room, and the imaging system 21 can be mounted in whole or in part on an unmanned aerial vehicle (also called a drone) equipped with rotors to enable video surveillance from the air. In particular, imaging can be performed in environments with little light, such as in the evening or at night when street lights are lit.
[0073] In addition, in the present embodiment, a video monitoring system or a security system has been described, but the present invention is not limited thereto. By combining a camera or a radar that captures images of the vehicle periphery with an ECU (Electronic Control Unit) that performs image processing, etc., the present invention can be applied to a vehicle that can perform semi-automatic driving or fully automatic driving. A vehicle that uses an electric motor has multiple ECUs, and the ECU performs motor control, etc. The ECU includes a microcomputer. The ECU is connected to a CAN (Controller Area Network) provided in the electric vehicle. CAN is one of the serial communication standards used as an in-vehicle LAN. The ECU uses a CPU or a GPU. For example, a solid-state image sensor without a color filter may be used as one of multiple cameras (such as a camera for a drive recorder or a rear camera) mounted on the electric vehicle, and the obtained black-and-white image may be inferred by the ECU via the CAN to create a colorized image, which may be displayed on a display device in the vehicle or a display unit of a portable information terminal.
[0074] (Embodiment 2) In this embodiment, an example of a flow for colorizing a black-and-white image obtained by the solid-state imaging device 11 is shown in FIG. 4 using the block diagram shown in the first embodiment.
[0075] A solid-state imaging element having no color filter is provided in a location capable of capturing an image of the area to be monitored (such as the bonnet or ceiling of the vehicle), and imaging system 21 shown in the first embodiment is installed and started to start capturing images continuously.
[0076] First, preparation for acquiring data is started, and acquisition begins (S1).
[0077] Black and white image data is acquired using a solid-state image sensor without a color filter (S2).
[0078] Next, the black and white image data is stored in the memory unit 14 (digital memory unit) (S5a). Storing the black and white image data requires less data than storing a color image as is.
[0079] Next, the digital data (black and white image data) is converted into a signal format (such as JPEG (registered trademark)) required by the subsequent inference program (S6).
[0080] Next, the converted digital data is subjected to convolution processing using a CPU or the like to infer contours, colors, etc., and is colorized (S7). Instead of the CPU, a single IC chip integrated with a GPU (Graphics Processing Unit), PMU (Power Management Unit), etc. may be used. The colorized image data is then output (S8). The colorized image data is then saved together with time data such as the date and time (S9). The data is saved in a large-scale storage device 15, a so-called large-capacity storage device (such as a hard disk), or in a database.
[0081] The learning data for colorization is obtained by selecting the learning data set as shown in the first embodiment. The selection method shown in the first embodiment can significantly shorten the time required to learn the learning data. In addition, even with a small amount of learning data, a vivid image can be obtained by colorization, which also serves as a highlighted display. Compared to a captured image that has a narrow dynamic range and is dark when using an image sensor with a conventional color filter, an image obtained in this embodiment is displayed more highlighted than before, and can be displayed as a suitable display for a driving assistance system.
[0082] The above-mentioned acquisition of colorized image data is repeated, which makes it possible to perform colorization in real time.
[0083] The colorized image data obtained in this way is based on a black-and-white image with a wide dynamic range using an imaging element without a color filter, so that distinguishable colorized image data can be obtained even when the amount of light is too low to distinguish using a conventional imaging element with a color filter. The imaging system shown in this embodiment can realize each of the above-mentioned steps (S1 to S9) using one or more computers.
[0084] (Embodiment 3) In the present embodiment, an example will be shown that enables smoother image processing or more detailed coloring processing compared to the colorized image data obtained in the second embodiment.
[0085] A flow diagram is shown in Fig. 5. The same reference numerals are used for the same steps as those in the flow diagram shown in Fig. 4 of the second embodiment. Since steps S1 to S6 in Fig. 4 are the same as steps S8 to S9, detailed explanations will be omitted here.
[0086] As shown in FIG. 5, after step S6, a super-resolution process is performed multiple times on the converted digital data using a first learning model to infer contours (S7a).
[0087] Then, the digital data after the super-resolution process is subjected to a second learning model to infer color and the like, and is colorized (S7b). The subsequent steps are the same as those in the second embodiment.
[0088] The method described in the first embodiment reduces the amount of training data required to obtain a color image that meets a certain standard, shortens the time required for machine learning, and simplifies the configuration of the neural network unit. Note that the neural network unit is a part of the machine learning. Also, deep learning is a part of the neural network unit.
[0089] The colorized image data obtained in this embodiment has smoother edges than the image data in the second embodiment, and is optimally colorized.
[0090] This embodiment mode can be freely combined with embodiment mode 1 or 2.
[0091] (Embodiment 4) In the present embodiment, a configuration example of an imaging unit will be described below as the data acquisition device 10. Fig. 6 is a block diagram for explaining the imaging unit.
[0092] The imaging unit includes a pixel array 300, a circuit 201, a circuit 302, a circuit 303, a circuit 304, a circuit 305, and a circuit 306. Note that each of the circuit 201 and the circuits 301 to 306 is not limited to a single circuit configuration, and may be configured by a combination of multiple circuits. Alternatively, any of multiple circuits described above may be integrated. Furthermore, a circuit other than the above may be connected.
[0093] The pixel array 300 has an imaging function and a calculation function. The circuits 201 and 302 have a calculation function. The circuit 302 has a calculation function or a data conversion function. The circuits 303, 304, and 306 have a selection function. The circuit 303 is electrically connected to the pixel block 200 through a wiring 424. The circuit 304 is electrically connected to the pixel block 200 through a wiring 423. The circuit 305 has a function of supplying a potential for a product-sum calculation to the pixel. A shift register, a decoder, or the like can be used as the circuit having the selection function. The circuit 306 is electrically connected to the pixel block 200 through a wiring 413. Note that the circuits 301 and 302 may be provided outside the pixel block 200.
[0094] 7, the pixel array 300 includes a plurality of pixel blocks 200. As shown in FIG. 7, the pixel block 200 includes a plurality of pixels 400 arranged in a matrix, and each of the pixels 400 is electrically connected to a circuit 201 via a wiring 412. Note that the circuit 201 can also be provided within the pixel block 200.
[0095] In addition, the pixel 400 is electrically connected to the adjacent pixel 400 through a transistor 450 (transistors 450a to 450f). The function of the transistor 450 will be described later.
[0096] In the pixel 400, image data can be acquired and data can be generated by adding the image data and a weighting factor. In FIG. 7, the number of pixels in the pixel block 200 is 3×3 as an example, but the number is not limited to this. For example, it can be 2×2, 4×4, etc. Alternatively, the number of pixels in the horizontal direction and the vertical direction may be different. Also, some pixels may be shared by adjacent pixel blocks. In FIG. 7, an example is shown in which ten transistors 450 (transistors 450a to 450j) are provided between the pixels 400, but the number of transistors 450 may be increased. Also, some transistors may be omitted from the transistors 450g to 450j to eliminate parallel paths. Wirings 413g to 413j are connected to the transistors 450g to 450j as gates, respectively.
[0097] The pixel block 200 and the circuit 201 can be operated as a multiply-accumulate circuit.
[0098] The pixel 400 may include a photoelectric conversion device 401, a transistor 402, a transistor 403, a transistor 404, a transistor 405, a transistor 406, and a capacitor 407, as shown in FIG.
[0099] One electrode of the photoelectric conversion device 401 is electrically connected to one of the source or drain of a transistor 402. The other of the source or drain of the transistor 402 is electrically connected to one of the source or drain of a transistor 403, the gate of a transistor 404, and one electrode of a capacitor 407. One of the source or drain of the transistor 404 is electrically connected to one of the source or drain of a transistor 405. The other electrode of the capacitor 407 is electrically connected to one of the source or drain of a transistor 406.
[0100] The other electrode of the photoelectric conversion device 401 is electrically connected to a wiring 414. The other of the source and drain of the transistor 403 is electrically connected to a wiring 415. The other of the source and drain of the transistor 405 is electrically connected to a wiring 412. The other of the source and drain of the transistor 404 is electrically connected to a GND wiring or the like. The other of the source and drain of the transistor 406 is electrically connected to a wiring 411. The other electrode of the capacitor 407 is electrically connected to a wiring 417.
[0101] A gate of the transistor 402 is electrically connected to a wiring 421. A gate of the transistor 403 is electrically connected to a wiring 422. A gate of the transistor 405 is electrically connected to a wiring 423. A gate of the transistor 406 is electrically connected to a wiring 424.
[0102] Here, an electrical connection point between the other of the source or drain of the transistor 402, one of the source or drain of the transistor 403, one electrode of the capacitor 407, and the gate of the transistor 404 is defined as a node FD. Also, an electrical connection point between the other electrode of the capacitor 407 and one of the source or drain of the transistor 406 is defined as a node FDW.
[0103] The wiring 414 and the wiring 415 can function as power supply lines. For example, the wiring 414 can function as a high-potential power supply line, and the wiring 415 can function as a low-potential power supply line. The wiring 421, the wiring 422, the wiring 423, and the wiring 424 can function as signal lines for controlling the conduction of each transistor. The wiring 411 can function as a wiring for supplying a potential corresponding to a weighting coefficient to the pixel 400. The wiring 412 can function as a wiring for electrically connecting the pixel 400 and the circuit 201. The wiring 417 can function as a wiring for electrically connecting the other electrode of the capacitor 407 of the pixel to the other electrode of the capacitor 407 of another pixel through a transistor 450 (see FIG. 7).
[0104] The wiring 412 may be electrically connected to an amplifier circuit or a gain adjustment circuit.
[0105] A photodiode can be used as the photoelectric conversion device 401. Any type of photodiode can be used, including a Si photodiode having silicon in the photoelectric conversion layer, an organic photodiode having an organic photoconductive film in the photoelectric conversion layer, etc. Note that, when it is desired to increase the light detection sensitivity at low illuminance, it is preferable to use an avalanche photodiode.
[0106] The transistor 402 can have a function of controlling the potential of the node FD. The transistor 403 can have a function of initializing the potential of the node FD. The transistor 404 can have a function of controlling a current flowing from the circuit 201 in accordance with the potential of the node FD. The transistor 405 can have a function of selecting a pixel. The transistor 406 can have a function of supplying a potential corresponding to a weighting coefficient to the node FDW.
[0107] When an avalanche photodiode is used for the photoelectric conversion device 401, a high voltage may be applied, and therefore a high-voltage transistor is preferably used for the transistor connected to the photoelectric conversion device 401. For example, a transistor using a metal oxide in a channel formation region (hereinafter, referred to as an OS transistor) can be used for the high-voltage transistor. Specifically, an OS transistor is preferably used for the transistor 402.
[0108] In addition, the off-state current of the OS transistor is extremely low. By using OS transistors for the transistors 402, 403, and 406, the period during which charges can be held at the nodes FD and FDW can be extremely long. Therefore, a global shutter system in which charges are accumulated simultaneously in all pixels can be applied without complicating the circuit configuration or operation method. In addition, while image data is held in the node FD, multiple calculations can be performed using the image data.
[0109] On the other hand, it may be desirable for the transistor 404 to have excellent amplification characteristics. In addition, it may be preferable to use a transistor with high mobility that can operate at high speed as the transistor 406. Therefore, a transistor using silicon for a channel formation region (hereinafter, referred to as a Si transistor) may be used as the transistor 404 and the transistor 406.
[0110] Note that the present invention is not limited to the above, and any combination of OS transistors and Si transistors may be used. All the transistors may be OS transistors. Alternatively, all the transistors may be Si transistors. Examples of Si transistors include transistors having amorphous silicon and transistors having crystalline silicon (microcrystalline silicon, low-temperature polysilicon, and single crystal silicon).
[0111] The potential of the node FD in the pixel 400 is determined by a potential obtained by adding a reset potential supplied from the wiring 415 and a potential (image data) generated by photoelectric conversion by the photoelectric conversion device 401. Alternatively, the potential is determined by further capacitive coupling with a potential corresponding to a weighting factor supplied from the wiring 411. Therefore, a current corresponding to data obtained by adding an arbitrary weighting factor to image data can flow through the transistor 405.
[0112] It should be noted that the above is only one example of the circuit configuration of the pixel 400, and the photoelectric conversion operation can be performed using other circuit configurations.
[0113] 7, the pixels 400 are electrically connected to each other through wirings 412. The circuit 201 can perform calculations using the sum of currents flowing through the transistors 404 of the pixels 400.
[0114] The circuit 201 includes a capacitor 202 , a transistor 203 , a transistor 204 , a transistor 205 , a transistor 206 , and a resistor 207 .
[0115] One electrode of the capacitor 202 is electrically connected to one of the source or drain of the transistor 203. One of the source or drain of the transistor 203 is electrically connected to the gate of the transistor 204. One of the source or drain of the transistor 204 is electrically connected to one of the source or drain of the transistor 205. One of the source or drain of the transistor 205 is electrically connected to one of the source or drain of the transistor 206. One electrode of the resistor 207 is electrically connected to the other electrode of the capacitor 202.
[0116] The other electrode of the capacitor 202 is electrically connected to a wiring 412. The other of the source and drain of the transistor 203 is electrically connected to a wiring 218. The other of the source and drain of the transistor 204 is electrically connected to a wiring 219. The other of the source and drain of the transistor 205 is electrically connected to a reference power supply line such as a GND wiring. The other of the source and drain of the transistor 206 is electrically connected to a wiring 212. The other electrode of the resistor 207 is electrically connected to a wiring 217.
[0117] The wiring 217, the wiring 218, and the wiring 219 can function as power supply lines. For example, the wiring 218 can function as a wiring that supplies a dedicated potential for reading. The wiring 217 and the wiring 219 can function as high-potential power supply lines. The wiring 213, the wiring 215, and the wiring 216 can function as signal lines that control the conduction of each transistor. The wiring 212 is an output line and can be electrically connected to, for example, the circuit 301 shown in FIG. 6.
[0118] The transistor 203 can have a function of resetting the potential of the wiring 211 to the potential of the wiring 218. The wiring 211 is a wiring connected to one electrode of the capacitor 202, one of the source and the drain of the transistor 203, and the gate of the transistor 204. The transistors 204 and 205 can function as source follower circuits. The transistor 206 can have a function of controlling reading. Note that the circuit 201 has a function as a correlated double sampling circuit (CDS circuit) and can be replaced with a circuit having another structure having the function.
[0119] In one embodiment of the present invention, offset components other than the product of image data (X) and weighting coefficient (W) are removed to extract the target WX. WX can be calculated using data with and without imaging for the same pixel, and data when weights are applied to each of the data.
[0120] The current (I p ) is the sum of kΣ(XV th ) 2 , the current (I p ) is kΣ(W+XV th ) 2 In addition, the current (I ref ) is the sum of kΣ(0-V th ) 2 , the current (I ref ) is the sum of kΣ(WVth ) 2 Here, k is a constant, V th is the threshold voltage of transistor 405.
[0121] First, calculate the difference (data A) between the captured data and the weighted data. kΣ((XV th ) 2 -(W+XV th ) 2 )=kΣ(-W 2 -2W·X+2W·V th )
[0122] Next, the difference between the no-image data and the weighted data (data B) is calculated. kΣ((0-V th ) 2 -(WV th ) 2 )=kΣ(-W 2 +2W·V th )
[0123] Then, take the difference between data A and data B. kΣ(-W 2 -2W·X+2W·V th -(-W 2 +2W·V th ))=kΣ(-2W·X). In other words, it is possible to remove offset components other than the product of the image data (X) and the weighting coefficient (W).
[0124] The circuit 201 can read out data A and data B. Note that the difference between data A and data B can be calculated by the circuit 301, for example.
[0125] Here, the weights supplied to the entire pixel block 200 function as a filter. For example, a convolution filter of a convolutional neural network (CNN) can be used as the filter. Alternatively, an image processing filter such as an edge extraction filter can be used. Examples of edge extraction filters include a Laplacian filter shown in FIG. 9A, a Prewitt filter shown in FIG. 9B, and a Sobel filter shown in FIG. 9C.
[0126] When the pixel block 200 has 3×3 pixels 400, the elements of the edge extraction filter can be assigned as weights to each pixel 400 and supplied. As described above, data A and data B can be calculated using data with and without imaging and data when weights are added to each of the data. Here, the data with and without imaging is data with no weight added, and can be rephrased as data with a weight of 0 added to all pixels 400.
[0127] 9A to 9C are filters in which the sum (ΣΔW / N, N is the number of elements) of the filter elements (weight: ΔW) is 0. Therefore, even if an operation of newly supplying ΔW=0 from another circuit is not performed, by performing an operation of acquiring ΣΔW / N, it is possible to acquire data in which ΔW=0 has been added to all pixels 400.
[0128] This operation corresponds to turning on the transistors 450 (transistors 450a to 450f) provided between the pixels 400 (see FIG. 7). By turning on the transistors 450, the nodes FDW of the pixels 400 are all short-circuited via the wiring 417. At this time, the charge accumulated in the nodes FDW of the pixels 400 is redistributed, and when the edge extraction filters illustrated in FIGS. 9A to 9C are used, the potential (ΔW) of the nodes FDW becomes 0 or approximately 0. Therefore, data with the equivalent of ΔW=0 added can be acquired.
[0129] When rewriting the weight (ΔW) by supplying charge from a circuit outside the pixel array 300, it takes time to complete the rewriting due to the capacitance of the long-distance wiring 411. On the other hand, the pixel block 200 is a very small area, and the wiring 417 is short and has a small capacity. Therefore, in the operation of redistributing the charge accumulated in the node FDW in the pixel block 200, the weight (ΔW) can be rewritten at high speed.
[0130] 7 illustrates a configuration in which the transistors 450a to 450f are electrically connected to different gate lines (wirings 413a to 413f), respectively. In this configuration, the conduction of the transistors 450a to 450f can be controlled independently, and an operation for obtaining ΣΔW / N can be selectively performed.
[0131] 9B or 9C is used, there is a pixel to which ΔW=0 is initially supplied. On the assumption that ΣΔW / N=0, the pixel to which ΔW=0 is supplied may be excluded from the pixels to be subjected to the summation. By excluding the pixel, it is not necessary to supply a potential for operating some of the transistors 450a to 450f, and therefore power consumption can be reduced.
[0132] The data resulting from the product-sum operation output from the circuit 201 is sequentially input to the circuit 301. The circuit 301 may have various calculation functions in addition to the above-mentioned function of calculating the difference between data A and data B. For example, the circuit 301 may have the same configuration as the circuit 201. Alternatively, the function of the circuit 301 may be replaced by software processing.
[0133] The circuit 301 may also include a circuit that performs an activation function calculation. For example, a comparator circuit can be used for the circuit. The comparator circuit compares input data with a set threshold value and outputs the result as binary data. That is, the pixel block 200 and the circuit 301 can act as a part of a neural network.
[0134] Data output from the circuit 301 is sequentially input to the circuit 302. The circuit 302 can have a configuration including, for example, a latch circuit and a shift register. With this configuration, parallel-serial conversion can be performed, and data input in parallel can be output to the wiring 311 as serial data.
[0135] [Pixel configuration example] 10A is a diagram showing an example of the configuration of a pixel 400. The pixel 400 can have a laminated structure of a layer 561 and a layer 563.
[0136] The layer 561 includes the photoelectric conversion device 401. The photoelectric conversion device 401 can include a layer 565a and a layer 565b as shown in Fig. 10B. Note that the layer may be referred to as a region in some cases.
[0137] 10B is a pn junction photodiode, and may use, for example, a p-type semiconductor for the layer 565a and an n-type semiconductor for the layer 565b. Alternatively, the layer 565a may be an n-type semiconductor, and the layer 565b may be a p-type semiconductor.
[0138] The pn junction photodiode can be typically formed using single crystal silicon.
[0139] Furthermore, the photoelectric conversion device 401 included in the layer 561 may be a laminate of layers 566a, 566b, 566c, and 566d, as shown in Fig. 10C. The photoelectric conversion device 401 shown in Fig. 10C is an example of an avalanche photodiode, in which the layers 566a and 566d correspond to electrodes, and the layers 566b and 566c correspond to a photoelectric conversion unit.
[0140] Layer 566a is preferably a low resistance metal layer, etc. For example, aluminum, titanium, tungsten, tantalum, silver, or a laminate of these may be used.
[0141] The layer 566d is preferably a conductive layer having a high light-transmitting property to visible light. For example, indium oxide, tin oxide, zinc oxide, indium-tin oxide, gallium-zinc oxide, indium-gallium-zinc oxide, graphene, or the like can be used. Note that the layer 566d may be omitted.
[0142] The layers 566b and 566c of the photoelectric conversion unit may be configured as a pn junction photodiode with a photoelectric conversion layer made of, for example, a selenium-based material. It is preferable that the layer 566b is made of a selenium-based material that is a p-type semiconductor, and the layer 566c is made of gallium oxide or the like that is an n-type semiconductor.
[0143] Photoelectric conversion devices using selenium-based materials have the characteristic of high external quantum efficiency for visible light. In such photoelectric conversion devices, avalanche multiplication can be used to increase the amplification of electrons relative to the amount of incident light. In addition, selenium-based materials have a high optical absorption coefficient, which offers the advantage in terms of production, such as the ability to fabricate a thin photoelectric conversion layer. Thin films of selenium-based materials can be formed using vacuum deposition, sputtering, or the like.
[0144] As the selenium-based material, crystalline selenium such as single crystal selenium or polycrystalline selenium, amorphous selenium, a compound of copper, indium, and selenium (CIS), or a compound of copper, indium, gallium, and selenium (CIGS), or the like can be used.
[0145] The n-type semiconductor is preferably made of a material that has a wide band gap and is transparent to visible light. For example, zinc oxide, gallium oxide, indium oxide, tin oxide, or a mixture of these oxides can be used. These materials also function as a hole injection blocking layer, and can reduce dark current.
[0146] Furthermore, the photoelectric conversion device 401 included in the layer 561 may be a laminate of layers 567a, 567b, 567c, 567d, and 567e as shown in Fig. 10D. The photoelectric conversion device 401 shown in Fig. 10D is an example of an organic photoconductive film, in which the layer 567a is a lower electrode, the layer 567e is a light-transmitting upper electrode, and the layers 567b, 567c, and 567d correspond to photoelectric conversion units.
[0147] In the photoelectric conversion portion, one of the layer 567b and the layer 567d can be a hole-transporting layer, the other of the layer 567b and the layer 567d can be an electron-transporting layer, and the layer 567c can be a photoelectric conversion layer.
[0148] The hole transport layer may be made of, for example, molybdenum oxide. The electron transport layer may be made of, for example, C 60 , C 70 or derivatives thereof can be used.
[0149] As the photoelectric conversion layer, a mixed layer (bulk heterojunction structure) of an n-type organic semiconductor and a p-type organic semiconductor can be used.
[0150] 10A includes, for example, a silicon substrate. Si transistors and the like are provided over the silicon substrate. The pixel 400 can be formed using the Si transistors. In addition, the circuit 201 and the circuits 301 to 306 shown in FIG. 6 can be formed.
[0151] Next, the laminated structure of the imaging device will be described using a cross-sectional view. Note that the elements such as the insulating layer and the conductive layer shown below are examples, and other elements may be further included. Alternatively, some of the elements shown below may be omitted. Also, the laminated structure shown below can be formed using a bonding process, a polishing process, etc., as necessary.
[0152] 11 includes layers 560, 561, and 563. In FIG. 11, the transistors 402 and 403 are shown as elements provided in the layer 563; however, other elements such as transistors 404 to 406 may also be provided in the layer 563.
[0153] The layer 563 includes a silicon substrate 632, an insulating layer 633, an insulating layer 634, an insulating layer 635, and an insulating layer 637. Also, a conductive layer 636 is provided.
[0154] The insulating layers 634, 635, and 637 function as an interlayer insulating film and a planarizing film. The insulating layer 633 functions as a protective film. The conductive layer 636 is electrically connected to the wiring 414 shown in FIG.
[0155] The interlayer insulating film and the planarizing film may be, for example, an inorganic insulating film such as a silicon oxide film, or an organic insulating film such as an acrylic resin or a polyimide resin. The protective film may be, for example, a silicon nitride film, a silicon oxide film, an aluminum oxide film, or the like.
[0156] The conductive layer may be made of a metal element selected from aluminum, chromium, copper, silver, gold, platinum, tantalum, nickel, titanium, molybdenum, tungsten, hafnium, vanadium, niobium, manganese, magnesium, zirconium, beryllium, indium, ruthenium, iridium, strontium, lanthanum, etc., or an alloy containing the above-mentioned metal element as a component, or an alloy combining the above-mentioned metal elements, etc. The conductor is not limited to a single layer, and may be a plurality of layers made of different materials.
[0157] The Si transistor shown in Fig. 11 is a fin type having a channel formation region in a silicon substrate. A cross section in the channel width direction (a cross section of A1-A2 shown in layer 563 in Fig. 11) is shown in Fig. 12A. The Si transistor may be a planar type as shown in Fig. 12B.
[0158] 12C, it may be a transistor having a silicon thin-film semiconductor layer 545. The semiconductor layer 545 may be, for example, single crystal silicon (SOI: Silicon on Insulator) formed on an insulating layer 546 on a silicon substrate 632.
[0159] The layer 561 is provided with a photoelectric conversion device 401. The photoelectric conversion device 401 can be formed on the layer 563. Fig. 11 shows a configuration in which the organic photoconductive film shown in Fig. 10D is used as the photoelectric conversion layer for the photoelectric conversion device 401. Note that here, the layer 567a is a cathode, and the layer 567e is an anode.
[0160] The layer 561 includes an insulating layer 651, an insulating layer 652, an insulating layer 653, an insulating layer 654, and a conductive layer 655.
[0161] The insulating layers 651, 653, and 654 function as an interlayer insulating film and a planarizing film. The insulating layer 654 is provided to cover an end of the photoelectric conversion device 401 and also functions to prevent a short circuit between the layer 567e and the layer 567a. The insulating layer 652 functions as an element isolation layer. An organic insulating film or the like is preferably used as the element isolation layer.
[0162] The layer 567a corresponding to the cathode of the photoelectric conversion device 401 is electrically connected to one of the source and the drain of the transistor 402 included in the layer 563. The layer 567e corresponding to the anode of the photoelectric conversion device 401 is electrically connected to the conductive layer 636 provided in the layer 563 through the conductive layer 655.
[0163] The layer 560 is formed on the layer 561. The layer 560 includes a light-shielding layer 671 and a microlens array 673.
[0164] The light-shielding layer 671 can suppress the inflow of light into adjacent pixels. A metal layer such as aluminum or tungsten can be used for the light-shielding layer 671. In addition, the metal layer and a dielectric film having a function as an anti-reflection film may be laminated.
[0165] A microlens array 673 is provided on the photoelectric conversion device 401. Light passing through each lens of the microlens array 673 is irradiated onto the photoelectric conversion device 401 directly below. By providing the microlens array 673, concentrated light can be incident on the photoelectric conversion device 401, so that photoelectric conversion can be performed efficiently. The microlens array 673 is preferably formed of a resin, glass, or the like that is highly translucent to light of the wavelength of the object of imaging.
[0166] 13 is a modification of the stacked structure shown in FIG 11, and is different in the configuration of the photoelectric conversion device 401 included in the layer 561 and in a portion of the configuration of the layer 563. In the configuration shown in FIG 13, a bonding surface is provided between the layer 561 and the layer 563.
[0167] Layer 561 includes photovoltaic device 401 , insulating layer 661 , insulating layer 662 , insulating layer 664 , insulating layer 665 , conductive layer 685 , and conductive layer 686 .
[0168] The photoelectric conversion device 401 is a pn junction type photodiode formed on a silicon substrate, and has a layer 565b corresponding to a p-type region and a layer 565a corresponding to an n-type region. The photoelectric conversion device 401 is a buried type photodiode, and a thin p-type region (part of the layer 565b) provided on the surface side (current extraction side) of the layer 565a can suppress dark current and reduce noise.
[0169] The insulating layer 661, the conductive layer 685, and the conductive layer 686 function as bonding layers. The insulating layer 662 functions as an interlayer insulating film and a planarizing film. The insulating layer 664 functions as an element isolation layer. The insulating layer 665 functions to suppress the outflow of carriers.
[0170] The silicon substrate is provided with grooves for separating pixels, and the insulating layer 665 is provided on the upper surface of the silicon substrate and in the grooves. By providing the insulating layer 665, it is possible to prevent carriers generated in the photoelectric conversion device 401 from flowing out to adjacent pixels. The insulating layer 665 also has a function of preventing the intrusion of stray light. Therefore, the insulating layer 665 can prevent color mixing. An anti-reflection film may be provided between the upper surface of the silicon substrate and the insulating layer 665.
[0171] The element isolation layer can be formed by using a LOCOS (LOCal Oxidation of Silicon) method. Alternatively, it may be formed by using a STI (Shallow Trench Isolation) method or the like. For example, an inorganic insulating film such as silicon oxide or silicon nitride, or an organic insulating film such as polyimide resin or acrylic resin can be used as the insulating layer 665. The insulating layer 665 may have a multi-layer structure. It is also possible to adopt a structure in which no element isolation layer is provided.
[0172] A layer 565a (n-type region, corresponding to a cathode) of the photoelectric conversion device 401 is electrically connected to a conductive layer 685. A layer 565b (p-type region, corresponding to an anode) is electrically connected to a conductive layer 686. The conductive layers 685 and 686 have regions embedded in an insulating layer 661. The surfaces of the insulating layer 661, and the conductive layers 685 and 686 are flattened so that they are at the same height.
[0173] In the layer 563, an insulating layer 638 is formed over the insulating layer 637. In addition, a conductive layer 683 electrically connected to one of the source and the drain of the transistor 402, and a conductive layer 684 electrically connected to the conductive layer 636 are formed.
[0174] The insulating layer 638, and the conductive layers 683 and 684 function as bonding layers. The conductive layers 683 and 684 have regions buried in the insulating layer 638. The surfaces of the insulating layer 638, the conductive layers 683, and the conductive layers 684 are flattened to be at the same height.
[0175] Here, it is preferable that the conductive layers 683 and 685 are mainly composed of the same metal element, and it is preferable that the conductive layers 684 and 686 are mainly composed of the same metal element. It is also preferable that the insulating layers 638 and 661 are mainly composed of the same metal element.
[0176] For example, Cu, Al, Sn, Zn, W, Ag, Pt, Au, or the like can be used for the conductive layers 683 to 686. In particular, Cu, Al, W, or Au is preferably used because of ease of bonding. Silicon oxide, silicon oxynitride, silicon nitride oxide, silicon nitride, titanium nitride, or the like can be used for the insulating layer 638 and the insulating layer 661.
[0177] That is, the same metal material as described above is preferably used for each of the conductive layers 683 to 686. The same insulating material as described above is preferably used for each of the insulating layers 638 and 661. With this structure, the layer 563 and the layer 561 can be bonded to each other at a boundary therebetween.
[0178] Note that the conductive layers 683 to 686 may each have a multilayer structure including a plurality of layers, in which case the surface layers (bonding surfaces) may be made of the same metal material. The insulating layer 638 and the insulating layer 661 may also have a multilayer structure including a plurality of layers, in which case the surface layers (bonding surfaces) may be made of the same insulating material.
[0179] By this bonding, the conductive layer 683 and the conductive layer 685 can be electrically connected to each other, and the conductive layer 684 and the conductive layer 686 can be electrically connected to each other. In addition, the insulating layer 661 and the insulating layer 638 can be connected to each other with sufficient mechanical strength.
[0180] To bond metal layers together, a surface activation bonding method can be used, in which oxide films and adsorbed layers of impurities on the surfaces are removed by a sputtering process or the like, and cleaned and activated surfaces are brought into contact with each other to bond them. Alternatively, a diffusion bonding method can be used, in which surfaces are bonded together using a combination of temperature and pressure. In either case, bonding occurs at the atomic level, so a bond that is excellent not only electrically but also mechanically can be obtained.
[0181] In addition, for bonding insulating layers, a hydrophilic bonding method can be used in which high flatness is obtained by polishing, etc., and then surfaces that have been hydrophilically treated with oxygen plasma, etc. are brought into contact with each other to form a temporary bond, and then the final bond is performed by dehydrating them through heat treatment. Hydrophilic bonding also produces bonds at the atomic level, so mechanically excellent bonds can be obtained.
[0182] When the layer 563 and the layer 561 are bonded to each other, a metal layer and an insulating layer are mixed on each bonding surface, so that, for example, a surface activated bonding method and a hydrophilic bonding method may be combined.
[0183] For example, a method of cleaning the surface after polishing, subjecting the surface of the metal layer to an anti-oxidation treatment, and then subjecting it to a hydrophilic treatment and bonding can be used. The surface of the metal layer may be made of a resistant metal such as Au and then subjected to a hydrophilic treatment. Note that bonding methods other than the above-mentioned methods may also be used.
[0184] By the above-mentioned bonding, the elements included in the layer 563 and the elements included in the layer 561 can be electrically connected to each other.
[0185] FIG. 14 shows a modification of the laminated structure shown in FIG. 13, in which the configurations of a layer 561 and a layer 563 are partially different.
[0186] In this modified example, the transistor 402 included in the pixel 400 is provided in a layer 561. In the layer 561, the transistor 402 covered with an insulating layer 663 is formed as a Si transistor. One of the source and the drain of the transistor 402 is directly connected to one electrode of the photoelectric conversion device 401. The other of the source and the drain of the transistor 402 is electrically connected to a node FD.
[0187] 14, the layer 563 includes transistors other than at least the transistor 402 among the transistors included in the imaging device. In FIG. 14, the transistors 404 and 405 are shown as elements provided in the layer 563, but other elements such as the transistor 403 and the transistor 406 can also be provided in the layer 563. In addition, an insulating layer 647 is provided between the insulating layer 635 and the insulating layer 637 in the layer 563 of the imaging device illustrated in FIG. 14. The insulating layer 647 functions as an interlayer insulating film and a planarization film.
[0188] (Embodiment 5) In this embodiment, a package that houses an imaging unit, that is, an image sensor chip, will be described below.
[0189] Fig. 15A1 is a perspective view showing the appearance of the upper surface of a package containing an image sensor chip. The package includes a package substrate 410 for fixing an image sensor chip 452 (see Fig. 15A3), a cover glass 420, and an adhesive 430 for bonding the two together.
[0190] 15A2 is an external perspective view of the bottom surface of the package. The bottom surface of the package has a BGA (Ball grid array) with solder balls as bumps 440. Note that the package is not limited to a BGA, and may have an LGA (Land grid array) or a PGA (Pin Grid Array), etc.
[0191] 15A3 is a perspective view of the package with a portion of the cover glass 420 and the adhesive 430 omitted. Electrode pads 460 are formed on the package substrate 410, and the electrode pads 460 and the bumps 440 are electrically connected via through holes. The electrode pads 460 are electrically connected to the image sensor chip 452 by wires 470.
[0192] 15B1 is a perspective view of the appearance of the upper surface side of a camera module in which an image sensor chip is housed in a lens-integrated package. The camera module has a package substrate 431 for fixing an image sensor chip 451 (FIG. 15B3), a lens cover 432, a lens 435, and the like. Also, an IC chip 490 (FIG. 15B3) having functions such as a drive circuit for an imaging device and a signal conversion circuit is provided between the package substrate 431 and the image sensor chip 451, and the camera module has a configuration as a SiP (System in Package).
[0193] 15B2 is a perspective view of the appearance of the bottom surface side of the camera module. The bottom surface and side surfaces of the package substrate 431 have a QFN (Quad flat no-lead package) configuration with mounting lands 441 provided thereon. Note that this configuration is one example, and a QFP (Quad flat package) or the above-mentioned BGA may also be provided.
[0194] 15B3 is a perspective view of the module in which a portion of lens cover 432 and lens 435 is omitted. Land 441 is electrically connected to electrode pad 461, and electrode pad 461 is electrically connected to image sensor chip 451 or IC chip 490 by wire 471.
[0195] By housing the image sensor chip in a package of the above-mentioned type, mounting on a printed circuit board or the like becomes easy, and the image sensor chip can be incorporated into various semiconductor devices and electronic devices.
[0196] This embodiment mode can be appropriately combined with the descriptions of other embodiment modes.
[0197] (Embodiment 6) By using the imaging device according to the above-described embodiment, an imaging device suitable for a vehicle performing semi-automatic driving is provided.
[0198] In Japan, the automation levels of driving systems for automobiles and other vehicles are defined as four levels, from level 1 to level 4. Level 1 refers to the automation of either acceleration, steering, or braking, and is called a safe driving support system. Level 2 automates multiple operations out of acceleration, steering, and braking at the same time, and is called a quasi-automated driving system (also called semi-automated driving). Level 3 is when acceleration, steering, and braking are all automated, with the driver only taking action in emergencies, and is also called a quasi-automated driving system (also called semi-automated driving). Level 4 is when acceleration, steering, and braking are all automated, and is called fully automated driving, with little to no driver involvement.
[0199] In this specification, we propose a new configuration or a new system, mainly assuming semi-automated driving at level 2 or level 3.
[0200] In order to display a warning to the driver of danger based on the situation obtained from various cameras or sensors, a display area corresponding to the number of cameras or sensors is required.
[0201] Fig. 16A shows an external view of vehicle 120. Fig. 16A shows an example of the locations of forward image sensor 114a and left side image sensor 114L. Fig. 16B is a schematic diagram showing the driver's field of vision from inside the vehicle. The upper part of the driver's field of vision is windshield 110, and the lower part of the field of vision is where display device 111 with a display screen is installed.
[0202] The upper part of the driver's field of vision is the windshield 110, which is sandwiched between pillars 112. In Fig. 16A, an example is shown in which the forward image sensor 114a is installed in a position close to the driver's line of sight, but this is not particularly limited, and the forward image sensor 114a may be installed in the front grille or the front bumper. In addition, although a right-hand drive vehicle is shown as an example in this embodiment, this is not particularly limited, and if the vehicle is a left-hand drive vehicle, the forward image sensor 114a may be installed according to the position of the driver.
[0203] It is preferable to use the image sensor chip shown in the fifth embodiment as at least one of these image sensors.
[0204] The driver accelerates, steers, and brakes while mainly looking at the display device 111, and checks the outside of the vehicle from the windshield as an auxiliary. The display device 111 may be any one of a liquid crystal display device, an EL (Electro Luminescence) display device, and a micro LED (Light Emitting Diode) display device. Here, an LED chip with a side dimension of more than 1 mm is called a macro LED, an LED chip with a side dimension of more than 100 μm and 1 mm or less is called a mini LED, and an LED chip with a side dimension of 100 μm or less is called a micro LED. It is particularly preferable to use a micro LED as an LED element applied to a pixel. By using a micro LED, a very high-definition display device can be realized. The higher the resolution of the display device 111, the more preferable it is. The pixel density of the display device 111 can be 100 ppi or more and 5000 ppi or less, preferably 200 ppi or more and 2000 ppi or less.
[0205] For example, the central portion 111a of the display screen of the display device displays an image acquired from an imaging device installed in front of the vehicle. Also, parts 111b and 111c of the display screen display meters such as speed, predicted driving distance, and abnormality warning display. Also, the lower left portion 111L of the display screen displays an image of the left side outside the vehicle, and the lower right portion 111R of the display screen displays an image of the right side outside the vehicle.
[0206] The left lower portion 111L and the right lower portion 111R of the display screen can be electronic side mirrors (also called door mirrors), eliminating the side mirror protrusions that protrude significantly outside the vehicle.
[0207] The display screen of the display device 111 may be configured to be touch-input operable to enlarge or reduce a part of an image, change the display position, or enlarge the area of the display region.
[0208] The image on the display screen of the display device 111 is created by combining data from a plurality of image capture devices or sensors, and is therefore created using an image signal processing device such as a GPU.
[0209] By using the imaging system shown in the first embodiment, it is possible to acquire black-and-white image data with a wide dynamic range, and perform inference using an image signal processing device or the like to colorize the image and output the highlighted image to the display device 111. For example, even in a tunnel, it is possible to add color to dark human figures and output an image in which the human figure is highlighted.
[0210] By using AI appropriately, the driver can operate the vehicle mainly by looking at the image displayed on the display device, i.e., the image using the image sensor and AI, and only look at the front of the windshield as an auxiliary. Operating the vehicle by looking at the image using AI can be a safer way to drive than driving with the driver's eyes alone. In addition, the driver can operate the vehicle with a sense of security.
[0211] The display device 111 can be applied to the driver's seat (also called the cockpit) of various types of vehicles, including large, medium, and small vehicles, and can also be applied to the driver's seat of vehicles such as aircraft and ships.
[0212] In addition, in this embodiment, an example is shown in which the forward image sensor 114a is installed below the windshield, but this is not particularly limited, and the imaging camera shown in FIG. 17 may be installed on the hood or around the rearview mirror inside the vehicle.
[0213] 17 can also be called a drive recorder, and includes a housing 961, a lens 962, and a support portion 963. By attaching double-sided tape or the like to the support portion 963, the imaging camera can be installed on the windshield, the hood, or a rearview mirror support portion.
[0214] An image sensor is disposed in the imaging camera of Fig. 17, and a driving image can be recorded and stored in a storage device inside the imaging camera or mounted on the vehicle. Furthermore, when a color filter is not used, an image with an expanded dynamic range based on black and white image data can be stored in the storage device, and the capacity of the storage device can be reduced compared to when a color image is captured. Furthermore, the black and white image data stored in the storage device may be colorized and displayed by the method shown in the first embodiment only when the user wants to see the image in color after driving, not in real time.
[0215] This embodiment mode can be freely combined with other embodiment modes. [Explanation of symbols]
[0216] 10: data acquisition device, 11: solid-state imaging element, 14: memory section, 15: large-scale storage device, 16: neural network section, 17: time information acquisition device, 18: memory section, 19: display section, 20: image processing device, 21: imaging system, 110: windshield, 111: display device, 111a: center section, 111b: part, 111c: part, 111L: lower left section, 111R: lower right section, 112: pillar, 114a: front image sensor, 114L: left side image sensor, 120: vehicle, 200: pixel block, 201: circuit, 202: capacitor, 203: transistor 204: transistor, 205: transistor, 206: transistor, 207: resistor, 211: wiring, 212: wiring, 213: wiring, 215: wiring, 216: wiring, 217: wiring, 218: wiring, 219: wiring, 300: pixel array, 301: circuit, 302: circuit, 303: circuit, 304: circuit, 305: circuit, 306: circuit, 311: wiring, 400: pixel, 401: photoelectric conversion device, 402: transistor, 403: transistor, 404: transistor, 405: transistor, 406: transistor, 407: capacitor, 410: pad Cage substrate, 411: wiring, 412: wiring, 413: wiring, 413a to 413f: wiring, 413g to 413j: wiring, 414: wiring, 415: wiring, 417: wiring, 420: cover glass, 421: wiring, 422: wiring, 423: wiring, 424: wiring, 430: adhesive, 431: package substrate, 432: lens cover, 435: lens, 440: bump, 441: land, 450: transistor, 450a to 450g: transistor, 450f to 450j: transistor, 451: image sensor chip, 452: image sensor chip, 460: electrode pad pad, 461: electrode pad, 470: wire, 471: wire, 490: IC chip, 545: semiconductor layer, 546: insulating layer, 560: layer, 561: layer, 563: layer, 565a: layer, 565b: layer, 566a: layer, 566b: layer, 566c: layer, 566d: layer, 567a: layer, 567b: layer, 567c: layer, 567d: layer, 567e: layer, 632: silicon substrate, 633: insulating layer, 634: insulating layer, 635: insulating layer, 636: conductive layer, 637: insulating layer, 638: insulating layer, 647: insulating layer, 651: insulating layer, 652: insulating layer, 653: insulating layer, 654: insulating layer,655: conductive layer, 661: insulating layer, 662: insulating layer, 664: insulating layer, 665: insulating layer, 671: light shielding layer, 673: microlens array, 683: conductive layer, 684: conductive layer, 685: conductive layer, 686: conductive layer, 961: housing, 962: lens, 963: support portion,
Claims
1. A solid-state imaging element having a function of acquiring black-and-white image data; A first storage device; A second storage device having a function of storing teacher data; A learning device; An imaging system having: a first step of acquiring the black-and-white image data in the solid-state imaging device; a second step of storing the black-and-white image data acquired by the solid-state imaging device in the first storage device; a third step of converting the black and white image data from the first storage device into a signal format required by an inference program; In the learning device, a step of inferring a contour of the black-and-white image data converted into the signal format by performing a super-resolution process a plurality of times on the black-and-white image data converted into the signal format; A step of inferring colors from the black-and-white image data after the super-resolution processing is performed a plurality of times using the training data, and generating colorized image data; having The training data is a second set that is converted into an HSV color space and thinned out based on statistics of a first set that is a median value of a saturation histogram. Imaging system.
2. In claim 1, The solid-state imaging element is a back-illuminated CMOS image sensor chip. Imaging system.
3. In claim 1 or 2, A set of median values of the saturation histogram is classified into 256 gradations, and data of 175 gradations or more are not used as the training data. Imaging system.
4. In any one of claims 1 to 3, the second set is 250 times smaller than the first set; Imaging system.
Citation Information
Patent Citations
Image processor, electronic camera, and image processing program
JP2006080746A
Image processing apparatus,monitoring center, monitoring system, image processing method, and image processing program
JP2007208481A
Method for driving liquid crystal display device
JP2010262276A
Semiconductor device
JP2011119711A
Semiconductor device and electronic apparatus
JP2016123087A
