KR20180063869A - Method, apparatus and computer program stored in computer readable medium for state decision of image data - Google Patents

Abstract

A method for determining the state of image data according to an embodiment of the present invention is disclosed. The method for determining the state of image data includes the steps of: obtaining first output data by a network function based on the image data; obtaining second output data by an algorithm with an effect which is different from the network function, based on the image data; and determining the state information of the image data based on the first output data and the second output data. Accordingly, the present invention can provide the state information of the image data based on the inputted image data.

Description

[0001] METHOD, APPARATUS AND COMPUTER PROGRAM STORED IN COMPUTER READABLE MEDIUM FOR STATE DECISION OF IMAGE DATA [0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to status determination of image data, and more specifically, to status determination of image data using a network function.

Pattern recognition is a field of machine learning, which means a discipline that recognizes the regularity of patterns and data. Pattern recognition techniques include supervised learning and unsupervised learning methods. Teacher learning method means a method in which an algorithm learns pattern recognition using data ("training" data) in which the result of pattern recognition is already determined. Where each training data may be referred to as labeled data. The comparator learning method means a method in which an algorithm finds a previously unknown pattern without labeled data.

A neural network may be used to implement the pattern recognition technique. The neural network consists of at least two nodes and links connecting these nodes. Each link may be weighted and the weight assigned to the link may be variable. The weight assigned to the link may be modified to be suitable for performing the pattern recognition that the neural network is intended to perform.

U.S. Patent No. 7,698,239 shows an example of such a neural network.

The present invention has been devised in response to the background art described above, and it is an object of the present invention to provide a method of determining a state of data.

An object of the present invention is to provide state information of image data through a method of determining the state of image data.

A method for determining the state of image data using a network function learned for at least one pattern according to an embodiment of the present disclosure for realizing the above-described problems is disclosed. Wherein the method comprises: obtaining first output data by the network function based on the image data; determining, based on the image data, a second output by an algorithm having an effect different from the network function And determining status information of the image data based on the first output data and the second output data.

Alternatively, the network function may include a plurality of links having a plurality of nodes and weights, and may include at least one input node and at least one output node.

Alternatively, the network function comprises a set of one or more nodes including a plurality of nodes and a plurality of links having weights, and the set of nodes consisting of a number of nodes equal to or less than the number of nodes of the set of input nodes . ≪ / RTI >

Alternatively, the method may further comprise determining at least one of a node, a weight, and a connection state of the network function based on preset image data before the step of acquiring the first output data.

Alternatively, the method may further include repeating the step of determining at least one of a node, a weight, and a connection state of the network function, wherein the preset image data may be one or more normal image data without a defect.

Alternatively, the algorithm may be an algorithm that loses or transforms at least a portion of the image data.

Alternatively, the algorithm may be an algorithm that divides the image data into images of a predetermined size, and then obtains second output data based on the average data value of the separated images.

Alternatively, the predetermined size may be a size set based on a data loss rate of the network function.

Alternatively, the second output data may be image data that is modified such that at least one of the first output data and the image resolution, size, color mode, and aspect ratio are the same.

Alternatively, the step of determining the status information may be a step of determining status information of the image data based on the identity of the first output data and the second output data.

Alternatively, the step of determining the state information may be a step of determining the state information of the image data based on the difference image of the first output data and the second output data.

Also disclosed is a method for determining the state of image data using a network function learned for at least one pattern according to another embodiment of the present disclosure. The method comprising: obtaining third output data by a first network function based on the image data; determining, based on the third output data, a second network function different from the first network function, And determining the status information of the image data based on the image data and the fourth output data.

Alternatively, the second network function may be a network function that restores at least a portion of the data of the lost third output data.

An apparatus for determining the state of image data using a network function learned for at least one pattern according to an embodiment of the present disclosure is disclosed. Wherein the device for determining the state of the image data includes a camera unit for photographing a state determination object, a control unit for controlling the device and using the network function, and the control unit A computation module for obtaining first output data and for obtaining second output data by an algorithm having an effect different from the network function based on the image data and a second module for generating second output data based on the first output data and the second output data And a status determination module for determining status information of the image data.

An apparatus for determining the state of image data using a network function learned for at least one pattern according to another embodiment of the present disclosure is disclosed. Wherein the apparatus for determining the state of the image data includes a camera section for photographing a state determination object, and a control section for controlling the apparatus and using the network function, wherein the control section controls the first network function A calculation module for obtaining fourth output data by a second network function different from the first network function based on the third output data, And a status determination module for determining status information of the image data based on the status information.

A computer program stored on a computer readable storage medium is disclosed that includes a plurality of instructions executed by one or more processors of the apparatus for determining the state of image data in accordance with one embodiment of the present disclosure. The computer program comprising instructions for causing a computation module to obtain first output data by a network function based on the image data, instructions for causing the computation module to perform an algorithm based on the image data, Instructions for obtaining second output data, and instructions for the status determination module to determine status information of the image data based on the first output data and the second output data.

A computer program stored on a computer-readable storage medium is disclosed that includes a plurality of instructions executed by one or more processors of the apparatus for determining the state of image data in accordance with another embodiment of the present disclosure. The computer program comprising instructions for obtaining third output data by a first network function based on the image data, instructions for obtaining a third output data by a second network function different from the first network function based on the third output data, 4 output data, and instructions for determining the status information of the image data based on the image data and the fourth output data.

The present invention can provide status information of image data based on input image data.

The present invention can inspect defects in the image data shooting object through the status information of the image data.

In order that the features of the above-mentioned subject matter of the present invention will be understood in detail and with reference to the following embodiments in more detail, some of the embodiments are shown in the accompanying drawings. In addition, like reference numerals in the drawings are intended to refer to the same or similar functions throughout the several views. It should be understood, however, that the appended drawings illustrate only typical exemplary embodiments of the present invention and are not to be considered limiting of its scope, and that other embodiments having the same effect may be fully recognized Please note.
1 is a conceptual diagram illustrating an example of a neural network according to an embodiment of the present disclosure.
2A is a block diagram illustrating a configuration of an apparatus for determining the state of image data according to an embodiment of the present disclosure.
FIG. 2B is a block diagram illustrating a configuration of a control unit according to an embodiment of the present invention in more detail.
3 is a flowchart illustrating a method of determining the state of image data using a network function learned for at least one pattern according to an embodiment of the present disclosure.
4 is a diagram schematically illustrating the structure of an auto-encoder among network functions according to an embodiment of the present disclosure.
FIG. 5A is a graphical illustration of exemplary input data of a method for determining the state of image data according to an embodiment of the present disclosure.
FIG. 5B is a diagram showing, in a method for determining the state of image data according to an embodiment of the present disclosure, a visual representation of exemplary data output by a network function based on the input data shown in FIG. 5A. FIG.
6A is a diagram illustrating another exemplary input data of a method for determining the state of image data according to an embodiment of the present disclosure.
FIG. 6B is a diagram showing a visual representation of exemplary data output by the network function based on the input data shown in FIG. 6A, in the method of determining the state of image data according to an embodiment of the present disclosure.
7A is a diagram illustrating another exemplary input data of a method for determining the state of image data according to an embodiment of the present disclosure.
FIG. 7B is a diagram showing a visual representation of exemplary data output by the network function based on the input data shown in FIG. 6A, in the method for determining the state of image data according to an embodiment of the present disclosure.
FIG. 8A is a diagram illustrating exemplary data obtained by obtaining a difference image based on the images of FIGS. 5A and 5B in the method of determining the state of image data according to an embodiment of the present disclosure.
FIG. 8B is a diagram showing an example of data obtained by obtaining a difference image based on the images of FIGS. 6A and 6B in the method of determining the state of image data according to an embodiment of the present disclosure.
FIG. 8C is a diagram showing an example of data obtained by obtaining a difference image based on the images of FIGS. 7A and 7B in the method of determining the state of image data according to an embodiment of the present disclosure; FIG.
FIG. 9 is a conceptual diagram for explaining a state determination structure of image data according to an embodiment of the present disclosure; FIG.
10 is a flowchart of a method for determining the state of image data according to an embodiment of the present disclosure.
11 is a block diagram illustrating a state determination structure of image data according to another embodiment of the present disclosure;
12 is a conceptual diagram for explaining a status determination structure of image data according to another embodiment of the present disclosure.
13 is a conceptual diagram illustrating a network function to which two or more network functions are connected according to another embodiment of the present disclosure;

Various embodiments are now described with reference to the drawings. In this specification, various explanations are given in order to provide an understanding of the present disclosure. It will be apparent, however, that such embodiments may be practiced without these specific details.

In addition, the term "or" is intended to mean " exclusive or " That is, it is intended to mean one of the natural inclusive substitutions "X uses A or B ", unless otherwise specified or unclear in context. That is, X uses A; X uses B; Or when X uses both A and B, "X uses A or B" can be applied to either of these cases. It should also be understood that the term "and / or" as used herein refers to and includes all possible combinations of one or more of the listed related items.

It is also to be understood that the term " comprises "and / or" comprising " means that the feature and / or component is present, but does not exclude the presence or addition of one or more other features, components and / It should be understood that it does not. Also, unless the context clearly dictates otherwise or to the contrary, the singular forms in this specification and claims should generally be construed to mean "one or more. &Quot;

The description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features presented herein.

1 is a conceptual diagram illustrating an example of a neural network according to an embodiment of the present disclosure.

Throughout this specification, a neural network, a network function, and a neural network may be used interchangeably. A neural network may consist of a set of interconnected computational units, which may be referred to generally as " nodes ". These " nodes " may also be referred to as " neurons. &Quot; The neural network comprises at least two nodes. The nodes (or neurons) that make up the neural network may be interconnected by one or more " links ".

Within a neural network, two or more nodes connected through a link may form a relationship of input and output nodes relatively. The concepts of the input node and the output node are relative, and any node in the output node relationship with respect to one node can be in the input node relationship with the other node, and vice versa. As described above, the input node-to-output node relationship can be generated around the link. One or more output nodes may be connected to one input node via a link, and vice versa.

In an input node and an output node relationship connected through one link, the output node can be determined based on data input to the input node. Here, a node interconnecting an input node and an output node may have a weight. The weights may be variable and may be varied by the user or algorithm to perform the desired function of the neural network. For example, if one or more input nodes to one output node are interconnected by respective links, then the output node is set to the values input to the input nodes associated with the output node and to the corresponding links to the respective input nodes The output node value can be determined based on the weight.

As described above, the neural network is interconnected by two or more nodes over one or more links to form an input node and an output node relationship within the neural network. The characteristics of the neural network can be determined according to the number of nodes and links in the neural network, the association between the nodes and links, and the value of the weight assigned to each of the links. For example, if there are the same number of nodes and links, and there are two neural network networks with different weight values between the links, then the two neural network networks may be recognized as being different from each other.

As shown in Figure 1, a neural network may comprise two or more nodes. Some of the nodes that make up the neural network may configure one layer based on distances from the original input node. For example, a set of nodes with a distance n from the first input node may be , n layers can be constructed. The distance from the original input node may be defined by the minimum number of links that must go through to reach that node from the original input node. However, the definition of this layer is arbitrary for explanation, and the degree of the layer in the neural network can be defined in a manner different from the above. For example, the layer of nodes may be defined by distance from the final output node.

The first input node may refer to one or more nodes in the neural network network to which data is directly input without going through a link in relation to other nodes. Or may refer to nodes in a neural network network that do not have other input nodes connected by a link in the relationship between the nodes based on the link. Similarly, the final output node may refer to one or more nodes that do not have an output node, in relation to other ones of the nodes in the neural network. Also, the hidden node may refer to nodes constituting a neural network other than the first input node and the last output node.

In the exemplary embodiment shown in FIG. 1, the neural network may have three layers. The three layers may include an initial input layer 110, a hidden layer 120, and a final output layer 130, in accordance with the above definition. However, the exemplary neural network shown in FIG. 1 is simplified for explanation, and the number of layers and the number of nodes constituting the neural network can be variously set according to the purpose and complexity of the neural network. In other words, there may be a neural network that includes two or more hidden layers (not shown). In addition, the number of nodes included in each layer may be more or less than that shown in FIG.

In the initial input layer 110, information about an object to be determined by the neural network can be input to each of the nodes 111, 112 and 113 constituting the initial input layer 110. [ The number of input nodes of the initial input layer 110 may be different depending on the type of the neural network. In one example, if the object to be determined by the neural network is image data, the input data extracted from the image data may be input to the initial input layer 110 of the neural network. In a different neural network having the same purpose of determining image data, the number of input nodes (not shown) of the neural network may be determined differently depending on the purpose, design, accuracy, etc. of the neural network.

For example, a neural network network may include a corresponding number of input nodes (not shown) in the original input layer 110 of the number of optical information for each of the pixels making up the image data. For example, in the case of a neural network for determining square image data having a size of 128 x 128 in the horizontal and vertical directions, information about the R, G, B numerical values for each pixel is stored in the first input layer 110 < / RTI > Thus, the original input layer may include 128 X 128 X 3 input nodes. In another example, a neural network network for determining image data may include fewer nodes in the initial input layer 110 than the example described above. For example, numerals (e.g., the number of consecutive pixels on the image data, the number of inflection points, etc.) indicative of the characteristics of the image data to be determined may be input to the initial input layer.

The neural network networks used in the embodiments disclosed herein are not limited to neural network networks for determining image data, and neural network networks may include various data (e.g., nodes on a computer network) Data included in a computer file to be compressed, data indicating a physical state of a mechanical component in robotics engineering, financial analysis data, etc.) can be processed as input data.

In the neural network 100, data input to one n layer can be output to the (n + 1) th layer. Nodes included in the n-th layer can be linked with nodes included in the n + 1-th layer. One node included in the n + 1 layer may include as input values a value computed based on weights assigned to one or more nodes and one or more links included in an n-layer connected via a link with the node have. In other words, k nodes (Nnode ₁ , Nnode ₂ , Nnode ₃ , ... Nnode _k ) included in the n layer and one node (N + 1) node ₁ included in the n + (L ₁ , ..., L _k ), the following formula can be established.

[Formula 1]

Value ((N + 1) node 1) = f (weight (L 1), weight (L 2), ..., weight (L k), value (Nnode 1), value (Nnode 2), ..., value (Nnode _k ))

Here, Value (X) may mean a node value input to a node X existing in a neural network. Weight (L) may refer to the weight of link L existing in the neural network. y = f (x) may mean that the y value has a functional relationship determined based on the x factor.

Other nodes included in the (n + 1) -th layer may also have input values as values calculated based on the values of the nodes included in the n-th layer and the weights of the links. In other words, the node values to be input to the nodes included in the (n + 1) -th layer are the values of the nodes included in the n-th layer and the connection relations of the links connecting the nodes of the n- And their weights.

As described above, the initial input layer 110 may be defined as a zero layer. Therefore, the initial input layer 110 and the hidden layer 120 of FIG. 1 are a 0-layer and a 1-layer, and are included in the 0-layer (the initial input layer 110) according to the computation method in the above- The value of the nodes 121, 122, 123, 124 included in one layer (hidden layer 120) is calculated based on the weight of the data and links 115 input to the nodes 111, 112, Can be determined. Similarly, the values of the nodes 131 and 132 included in the two layers (the final output layer 130) are the same as those of the nodes 121, 122, 123, and 124 included in the first layer 120, 125). ≪ / RTI >

It will be appreciated by those skilled in the art that the neural network 100 shown in FIG. 1 is only an example of a neural network and that there may be various types of neural network. A neural network may be referred to as a network function in the sense that input and output values are present. In other words, in this specification, the network function may be understood to mean a neural network.

In this specification, a network function, a neural network, can be used interchangeably. Also, the network function and the neural network may be understood to include an auto encoder, a diabolo network, and the like. When two or more neural network networks exist, if more than one neural network output nodes are input nodes of another neural network, two or more neural network networks may be referred to as a serial connection. When two or more neural network networks are connected in series, two or more serially connected neural network networks can constitute a neural network network.

On the other hand, through the neural network, objects of the inspection target in the process are imaged and inspected, and the target patterns are found. For example, an imaged textile fabric may be inspected during fabrication to find scratches, and various objects may be imaged and inspected to find defects in the material's holes or semiconductor circuits. There is a limitation in detecting and analyzing image images by a person, and a method for determining the state of image data with high accuracy is required in the art.

2A is a block diagram showing the configuration of an apparatus for determining the state of image data according to an embodiment of the present disclosure.

The method of determining image data according to an embodiment of the present invention may be executed in devices having computation capability. For example, the method of determining image data may be performed in a personal computer (PC), a work station, a server computing device, and the like. Or may be performed by a separate device for performing the method of determining image data according to an embodiment of the present invention. A method according to an embodiment of the present invention may be performed in one or more computing devices. For example, at least one or more steps of the image data determination method of the present invention may be performed in a client device, and the other steps may be performed in a server device. In this case, the client device and the server device can be connected to each other via a network to transmit and receive the operation result. Alternatively, the method of determining image data according to an embodiment of the present invention may be performed by distributed computing technology.

According to an aspect of the present disclosure, an apparatus 200 for determining the state of image data includes a camera unit 210, a display 230, a network connection unit 240, an input unit 250, a memory 260, and a control unit 270 can do. The above-described configurations of FIG. 2 are exemplary, and the apparatus may further include a user input unit, a lighting unit, and the like. The apparatus may also be constructed solely of the components of the components described above. The device may be used in communication with an external server or may itself comprise a server.

According to one embodiment of the present disclosure, the camera unit 210 can capture an image of a determination object and acquire an image. The photographed object to be judged may include various photographs such as cloth, leather, skin, part of human body, hydrate, animal or plant. The photographing is merely an example, and the camera unit 210 can take photographs without limitation on the type of photographing, and the present disclosure is not limited thereto. The images photographed by the camera unit can be stored in the memory 260 or transmitted to the outside through the network connection unit 240. [ The camera unit 210 may be composed of one camera or two or more cameras depending on the use environment.

In one embodiment, the camera unit 210 may be integrally connected to the device 200 for determining the state of image data according to an embodiment of the present invention, or may be configured as a detachable module. The camera unit 210 may include various hardware configurations to acquire image data to be subjected to status determination.

As an example, the camera unit 210 can operate in conjunction with photographing means (not shown) such as, for example, a conveyor belt, in order to acquire data at high speed on a photograph. (Not shown) may be a component of the device 200 according to one embodiment of the present invention or may be a component of a separate external device. For example, the camera unit 210 may be configured as a line camera for scanning an object to be photographed moving along the conveyor belt. However, the configuration of the camera unit 210 according to an embodiment of the present invention is not limited to the above-described configuration, and any combination for efficiently acquiring image data can be used in an embodiment of the present invention.

In addition, the camera unit 210 can capture various types of images. The photographing result of the camera unit 210 may include a 3D image, a monochrome image, a GIF image stored in accordance with the passage of time, an image image, an infrared image, an electronic image, and the like. The camera unit 210 may include a film camera, a digital camera, a microscope, a magnifying glass, an infrared camera, an ultraviolet (UV) camera, an X-ray, a magnetic resonance imaging apparatus and an arbitrary image acquiring apparatus. The configuration of the camera unit 210 may be changed according to the photographed object to be photographed.

According to one embodiment of the present disclosure, the display 230 may display an image captured by the camera unit 210. [ The display 230 may display an image currently being photographed by the camera unit 210 or may display an image stored in the memory 260. Display 230 may be capable of sound output. It is also possible to output a warning sound in accordance with the result of the state determination or not. The display 230 may output a status determination result for the photographed object. It may also receive and output an image of an external device, or may output an image status determination result of the external device. The display 230 can output a photographing mode that can be photographed. For example, you can output 200dip, 300dip, and 600dip image resolution. In addition, a selection screen can be displayed so that an image size, an image shooting mode, a photographing apparatus, and the like can be selected. The display 230 may comprise any visual output device such as an LCD, LED, CRT, or the like. The display examples and components described above are merely illustrative, and the present disclosure is not limited thereto.

According to an embodiment of the present disclosure, the network connection unit 240 may transmit the image photographed by the camera unit 210 to another device or a server. The network connection unit 240 may include one or more modules that enable wireless communication between the two devices when the image capturing device and the image determining device are separately configured. The network connection unit 240 may include a transmitting unit and a receiving unit. The network connection unit 240 may include a wired / wireless Internet module for network connection. WLAN (Wi-Fi), Wibro (Wireless broadband), Wimax (World Interoperability for Microwave Access), HSDPA (High Speed Downlink Packet Access) and the like can be used as wireless Internet technologies. Wired Internet technologies include XDSL (Digital Subscriber Line), FTTH (Fiber to the home), and PLC (Power Line Communication).

According to one embodiment of the present disclosure, the network connection unit 240 may receive an image from an external device or a server. The received image can be judged or classified by the controller 270 by judging the status of the image data. The controller 270, which extracts image data from the image and determines the state of the image data, may be configured as a separate device. The received image may be a digital image, a digital image, a compressed image Data, a RAW image, a data file, and the like, and the data that can be converted into an image can be received without limitation. The image received from the outside via the network connection unit 160 may be used for determining the state of the image data in the control unit 270 or may be used to determine the state of the state determination photograph by determining the state of the image.

In addition, the network connection unit 240 may transmit data to and receive data from an electronic device including a short-range communication module, which is located in a relatively short distance from the apparatus for determining the state of image data. Bluetooth, Radio Frequency Identification (RFID), infrared data association (IrDA), Ultra Wideband (UWB), ZigBee, and the like can be used as a short range communication technology. In one embodiment of the present disclosure, the network connection unit 160 can detect the connection state of the network and the transmission / reception speed of the network. Data received via the network connection 160 may be output via the display 230, stored via the memory 260, or transmitted to other electronic devices in the vicinity via the local communication module.

According to one embodiment of the present disclosure, the input unit 250 may connect an external device to use the device, or may connect to the storage device to receive data. For example, when the USB storage device is connected to the input unit 250, an image stored in the USB may be copied to the memory 260 and stored or displayed on the display 230.

According to one embodiment of the present disclosure, the input unit 250 can receive an external operation. For example, when the jog included in the input unit 250 is manipulated for the control of the camera unit 210, the camera unit 210 can change the position to be photographed or change the focus. The input unit 250 may include a switch for changing the status determination photograph. The input unit 250 may receive various user inputs.

According to one embodiment of the present disclosure, the input unit 250 may include or connect to one or more external input devices. The user can perform firmware upgrade, content storage, and the like of the device 200 by connecting an external storage medium to the input unit. In addition, a user may connect a keyboard, a mouse, and the like to the input unit 250. In addition, the USB port of the input unit 250 may support USB 2.0, and USB 3.0. In addition, the input unit 250 may include an interface such as a thumb bolt as well as a USB port. Also, the input unit 250 may receive external input by combining various input devices such as a touch screen, a button, and the like, and the control unit 270 may control the device 200 through the input.

According to one embodiment of the present disclosure, the memory 260 may store a program for operation of the controller 270, and may store input / output data (e.g., status determination information, messages, still images, May be temporarily or permanently stored. The memory 260 may store data relating to display and sound. The memory 260 may be a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory), a RAM (Random Access Memory), SRAM (Static Random Access Memory), ROM (Read Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM A disk, and / or an optical disk.

Memory 260 may store various network functions and algorithms. The memory 260 may store programs, software, instructions, codes, and the like that cause the device 200 to run. The memory 260 may temporarily store the image data including the cache, and may temporarily store programs, commands, and the like necessary for operating the apparatus 200. The memory 260 may store at least a portion of images and images captured by the camera unit 210. In addition, the memory 260 may delete, from the stored photographing data, a photographed image not yet loaded by another component such as the controller 270 or another device after a predetermined time. The memory 260 may store the photographed image divided into image patches of various sizes. The configuration of the memory 260 is merely an example, and the present disclosure is not limited thereto.

FIG. 2B is a block diagram illustrating a configuration of a control unit according to an embodiment of the present invention in more detail.

The controller 270 may include a data acquisition module 271, a computing module 272, a status determination module 273, a node and a weight determination module 274, according to one embodiment of the present disclosure. The control unit 270 may further include other modules or may be configured as a part of the modules depending on the photographed objects to be photographed, algorithms and the like. The control unit typically controls the overall operation of the state determination apparatus 200 of image data. For example, it is possible to divide an image photographed through the camera section 210 into image patches of a predetermined size, store the photographed image in the memory 260, and judge whether the photographed image is a state judgeable image .

According to one embodiment of the present disclosure, the data acquisition module 271 may obtain image data that can be determined from the image photographed by the camera section 210. The data acquisition module 271 may obtain image data that can be determined from the image input through the input unit 250. The data acquisition module 271 may divide the image captured by the camera unit 210 into image patches of a predetermined size to acquire image data. The image input to the data acquisition module 271 can be input without limitation to the formats such as jpg, png, tif, psd, and Ai, and data can be input without limitation to moving images, image files, and the like.

According to one embodiment of the present disclosure, the computing module 272 may obtain first output data by a network function based on the image data. The computation module 272 may obtain second output data by an algorithm having an effect different from the network function, based on the image data. The output result output from the operation module 272 may be a numerical value, image data, or various types of data.

According to another embodiment of the present disclosure, the computing module 272 may obtain third output data by the first network function based on the image data. The computation module 272 may obtain fourth output data by a second network function different from the first network function, based on the third output data. The calculation module 272 can calculate the output data based on the data input through the network function. The calculation module 272 can calculate the input data based on the network function and output the output data. The network function does not operate on its own, and the operation module 272 can execute the operation through the network function.

According to one embodiment of the present disclosure, the status determination module 273 can determine the status information of the image data based on the data obtained in the operation module 272. [ The status determination module 273 can determine the status information of the image data based on the first output data and the second output data. The status determination module may determine status information of the image based on the identity of the first output data and the second output data. The status information of the image data may be information classified into a plurality of classes or categories. In addition, the status information of the image data may be normal or bad information.

According to another embodiment of the present disclosure, the status determination module can determine the status information of the image data based on the image data and the fourth output data. The status determination module 273 may determine the status information of the image based on the fourth output data and the identity of the image data. The status determination module can compare the image data based on the identity and determine the status information without being limited to the data type.

In accordance with one embodiment of the present disclosure, the node and weight determination module 274 may determine at least one of a node, a weight, and a connection state of a network function. The node and weight determination module 274 may determine at least one of a node, a weight, and a connection state so that the most accurate image state determination result may be obtained based on preset image data. The node and weight determination module 274 may be used only in the pre-learning phase and may not be used in the image determination step. In the pre-learning stage, the node and weight determination module 274 may repeatedly add, delete, or change nodes, weights, connection states, etc. to supplement the network function so that the most accurate image state determination result is obtained.

As shown in FIGS. 2A and 2B, the controller 270 can communicate with all of the other components described above, so that they can organically control their operations.

The various embodiments described herein may be embodied in a recording medium readable by a computer or similar device using, for example, software, hardware, or a combination thereof. According to a hardware implementation, the embodiments described herein may be implemented as application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays May be implemented using at least one of a processor, controllers, micro-controllers, microprocessors, and other electronic units for performing other functions. In some cases, For example, according to one embodiment of the present disclosure, if the pre-learning step is not included, the node and weight determination module 274 may be omitted, According to a software implementation, embodiments such as the procedures and functions described herein may be implemented in separate software modules Each of the software modules may perform one or more of the functions and operations described herein. Software code may be implemented in a software application written in a suitable programming language. And may be executed by the control unit 270. [

3 is a flowchart illustrating a method of determining a state of image data using a network function learned for at least one pattern according to an exemplary embodiment of the present invention.

The method of determining image data according to an embodiment of the present invention can be used, for example, to determine the state of an irregular pattern. The non-standardized pattern may refer to, for example, a pattern associated with image data of a blood vessel structure of a human body, a pattern of leather, an appearance of a specific animal (for example, a fish), human handwriting, .

The method of determining image data according to an exemplary embodiment of the present invention can be used, for example, to classify the state of image data to be determined. The state of the image data to be judged may be two or more. For example, the damage or defective portion of the pattern of the leather, the fish species, the letters intended by the human handwriting, etc. may be the state of the image data.

3, a method of determining image data according to an exemplary embodiment of the present invention includes determining (S310) at least one of a node, a weight, and a connection state of a network function based on preset image data (S310) (S330) of acquiring second output data by an algorithm having an effect different from that of the network function (S330), and acquiring first output data of the image data based on the first output data and the second output data And determining status information (S340).

Here, the network function according to an embodiment of the present invention may include the above-described neural network having output data and input data. In addition, the network function may include a plurality of links having a plurality of nodes and weights, and may include at least one input node and one output node. In addition, the network function may include any function used in machine learning or pattern recognition. A network function according to an embodiment of the present invention may refer to a network function that meets the conditions appropriate to accomplish the object of the present invention. For example, a network function may include an autoencoder, an autoassociator, or a Diabolo network.

The auto encoder 400 may include one input layer 410, an output layer 450, and at least one hidden layer. The number of nodes constituting the input layer of the auto encoder 400 and the number of nodes constituting the output layer may be the same. The auto-encoder can be a network function usable in an unsupervised learning model, consistent with the definition of the neural network described above. In other words, an auto-encoder can be designed to reconstruct the input given to a given input layer. In one embodiment, the number of nodes constituting the hidden layers constituting the auto encoder 400 may be smaller than the number of nodes constituting the input layer or the output layer. That is, the network function constituting the auto encoder 400 may include a set of nodes composed of a number of nodes equal to or less than the number of nodes of the set of input nodes.

4 is a diagram schematically illustrating the structure of an auto-encoder among network functions according to an embodiment of the present invention. As shown in FIG. 3, the auto encoder 400 may include at least one hidden layer while satisfying the definition of the neural network described above. A subset of the layers that make up the auto-encoder 400 itself can satisfy the definition of a neural network. For example, a set of input layers 410 and hidden layers 420 and 430 of the auto encoder 400 may be recognized as one neural network. In this case, the portion can be referred to as an encoder 400A. Similarly, the hidden layers 430 and 440 of the auto encoder 400 and the set of output layers 450 can be recognized as one neural network. In this case, the portion can be referred to as a decoder 400B. As shown in FIG. 4, the output layer of the encoder 400A may be the input layer of the decoder 400B

The auto encoder 400 encodes data inputted to the input layer 410 through the hidden layer 420 of the encoder 400A and outputs the result to the hidden layer 440 of the decoder 400B And perform a decoding function. Encoder 400A and decoder 400B may each include one or more hidden layers.

The auto encoder 400 may have a predetermined neural network structure. The neural network structure can be defined as the number of nodes constituting the neural network, the connection relation of the links connecting the nodes, and the weight assigned to each link.

Referring back to FIG. 3, the image data determination method according to an exemplary embodiment of the present invention may determine the structure of a network function network based on preset image data (S310). Here, the network function may include an auto encoder satisfying the definition of the neural network. The structure of the neural network of the network function may refer to the nodes of the network function, the link state of the links connecting the nodes, and the weights given thereto, as described above. In step S310, determining the structure of the neural network may refer to an operation of determining the structure of the neural network with a structure suitable for the image determination method according to an embodiment of the present invention. The operation can be performed through unsupervised training. The meaning of the operation of the network function and the determination of the structure of the network function may mean that the module of the control unit or the control unit operates according to at least one of the predetermined network functions and algorithms. For example, the node and weight determination module 274 may operate along at least one of the predetermined network functions and algorithms to determine the node and the weight.

For example, in step S310, input data obtained based on image data may be input to an input layer of a given network function initially. In step S310, the image data may be divided into one or more image patches having a size suitable for input to the network function. Here, the image data may be an image captured by a photographing device such as a camera, an infrared ray photographing, an X-ray, a camcorder, or the like, but is not limited thereto.

As described above, the input data may be information (for example, R, G, B values of each pixel constituting the image data) containing information of the image data without loss. Or the input data may be data configured based on the feature value of the image data. The input data may be converted to output data via a network function.

As described above, the network function may be the auto-encoder of FIG. The image patch input to the network function can be converted to output data based on the structure of the network function. When the network function has the structure of the auto encoder, the number of nodes of the input layer and the number of nodes of the output layer may be the same. Thus, the image data, which is input data, and the output data may include mutually comparable information. For example, based on the output data, an output image patch having the same or similar size as the image patch that is input data can be generated.

In step S310, in order to generate a network function suitable for the method of determining image data according to an embodiment of the present invention, the apparatus 200 for determining the state of image data may change the structure of the network function. The operation of changing the structure of a network function may mean changing at least one of the number of nodes of the network function, the connection relationship of the links and the weight of the links.

For example, for determining image data according to an embodiment of the present invention, the network function used in step S310 may be designed to preserve the features shared by two or more image data of a similar type. For example, if image data is generated based on data having a pattern of fabric, leather, etc., the network function may be designed such that the output data retains the pattern that appears in the fabric or leather. Alternatively, when the image data is generated based on an image (for example, an image of CT, MRI, X-ray, or the like) containing medical information of the human body, (For example, a pattern in a steady state such as a blood vessel, a normal pattern such as a long-term inner wall without a tumor or abnormal tissue, etc.). Or when the image data is generated based on the data of the appearance of the fish, the network function determines whether the output data is characteristic of the fish of the same species (for example, the shape of fish scales, the arrangement and proportion of the fins, etc.) . ≪ / RTI >

The network function suitable for the method of determining image data according to an embodiment of the present invention can satisfy the definition of the auto encoder. In this case, depending on the characteristics of the auto-encoder, the network function may include one or more hidden layers having fewer nodes than the number of nodes included in the input and output layers. Due to this characteristic, some data may be lost in the network function (auto encoder) of the input data. In this process, a network function designed to preserve one characteristic shared by two or more image data may lose data that does not belong to the characteristic. In this case, the output data in which the specific data is lost may have a relatively low similarity with the input data as compared to the data with no or little data loss.

For example, data sharing certain attributes may be defined as normal data, and data not shared may be defined as abnormal data. For example, when a pattern of leather or the like follows a normal normal leather pattern, data based on the image of the leather can be defined as normal data. Likewise, data based on an image of a pattern of a defective leather such as scratch or dirt can be defined as abnormal data. In this case, the network function receiving the normal data can output the output data highly similar to the normal data. On the other hand, the network function receiving the abnormal data can output the output data having relatively low similarity with the abnormal data. However, the state of the data may be defined as three or more. For example, depending on the design, the state of the leather may be determined as three states: normal, abnormal due to contamination, and abnormal due to scratches.

In step S310, if the initial structure of the predetermined network function does not meet the above-described design conditions, the state determination apparatus 200 for image data according to an embodiment of the present invention can change the structure of the network function have. Changing the structure of a network function may generally mean adjusting the weight assigned to the link that constitutes the network function. By adjusting the weights given to the link, the network function can output output data with little information loss to the input data having specific attributes.

In step S310, the controller 270 of the apparatus 200 for determining the state of image data according to an exemplary embodiment of the present invention may perform an operation of adjusting a network function. The control unit 270 may determine the structure (at least one of the node, the weight, and the connection state) of the network function to be used in the subsequent steps S320, S330, and S340 by adjusting the network function (S310). The control unit 270 may adjust the network function based on a predetermined network function adjustment algorithm or a signal input through the input unit 250. [ In step S310, the control unit 270 may compare the input and the output of the network function to determine the similarity. If the input data is normal data, the controller 270 may adjust its structure to enhance the similarity of the input and the output. The more input data is input to the network function, the more coordination steps are performed, and thus the network function that can be determined by the node and weight determination module 274 is suitable for the image data determination method according to an embodiment of the present invention. .

As described above, the image data may be divided into a plurality of input patches, and thus, a plurality of input patches may be obtained from the image data for one object. Suitability of the network function can be increased as more data to be input into the network function operated by the operation module 272 is. Therefore, a large number of input data suitable for comparison yarn learning can be obtained with a small number of image data.

The network function having the structure determined through the node and weight determination module 274 may be referred to as a learned network.

If the network function is determined in step S310, the process may proceed to step S320 to determine the status of the image data.

In step S320, the image data may be input to the network function (determined in S310). The image data may be generated based on image data photographed directly by the camera unit 210 of the apparatus 200 for determining the state of image data according to an embodiment of the present invention or received from the outside via the network connection unit 240 And the input unit 250, as shown in FIG. As described above, the image data can be divided into image patches of a size that can be input to the network function. Or the image data itself can be preset to a size that can be input to the network function. The computation module 272 may obtain the first output data from the output of the network function.

According to one embodiment of the present disclosure, the first output data obtained by the network function may be equal to the size of the extracted image patch. As described above, according to one embodiment of the present disclosure, the network function learned for at least one pattern can output first output data having consistent characteristics when normal data is input. If the learned network function for one pattern receives defective data, the defective data may be less similar to the output data obtained through the network function than for the normal data. That is, when the abnormal data having a defect is output by the operation module 272, the similarity between the input data and the output data may be low. Alternatively, when the abnormal data passes through the network function, the output data may have a loss rate of the information that is relatively larger than the output data of the normal data.

FIG. 5A is a graphical illustration of exemplary input data of a method for determining the state of image data according to an embodiment of the present disclosure. As shown in Fig. 5A, the input data may be image data obtained by photographing a photograph such as leather. Such a photograph may include an image pattern having a consistent feature. In Figure 5A, a consistent feature may refer to a pattern of leather. As shown in FIG. 5A, the pattern of leather can be applied to the entire image data while maintaining a consistent feature. In other words, the image shown in Fig. 5A keeps the normal pattern of the leather throughout the image. This image can be defined as being " normal data ". As described above, when the normal function is used as the input data, the network function can output the output data with a relatively small data loss as compared with the normal data as the input data. As described above, the network function can be trained (S310) to output the output data with relatively little data loss with respect to the normal data.

FIG. 5B is a diagram showing, in a method for determining the state of image data according to an embodiment of the present disclosure, a visual representation of exemplary data output by a network function based on the input data shown in FIG. 5A. FIG. Based on the normal image shown in FIG. 5A, a trained (e.g., comparative learned) network function may generate output data. As described above, when generating the output data based on the normal data (the image data in FIG. 5A), the network function can output data in which the loss amount of the information included in the normal data is relatively small. For example, if the network function is an auto-encoder 400, information loss may occur due to the presence of at least one hidden layer 430 having fewer nodes than the input layer 410. However, in the case of the auto-encoder 400 according to an embodiment of the present invention, the structure of the network function is optimized so that it is suitable to preserve the coherent characteristics contained in the normal data (the data that keeps the coherent characteristic throughout the data) (S310). Accordingly, the auto-encoder 400 can output the output data (see FIG. 5B) that maintains a consistent characteristic contained in the normal data (the image data in FIG. 5A). In the case of Fig. 5B, on the basis of Fig. 5A, consistent features (patterns defined according to the bending of the leather) existing in the pattern of the leather in the output data are maintained. However, depending on the characteristics of the auto-encoder 400 or any other network in which there is a loss of data, the image loss may also be present in the output data based on the normal data. For example, as shown in FIG. 5B, a decrease in resolution and a loss of contrast due to data loss in a high frequency region may occur. However, the manner of data loss is not limited thereto.

6A is a diagram illustrating another exemplary input data of a method for determining the state of image data according to an embodiment of the present disclosure. As shown in Fig. 6A, similar to Fig. 5A, the input data may be image data obtained by photographing a photograph such as leather. However, unlike FIG. 5A, the image shown in FIG. 6A may include portions of the image with a " different feature " having attributes that are not included in the coherent feature present in the pattern of the leather. For example, as shown in FIG. 6A, the image 600A may include an image portion 610 that is displayed in a straight line on the skin. In general, the pattern of the leather is defined according to the bending of the leather, as shown in Fig. 5A. For example, the bending of the leather may have a similar width to the regions produced by the bending of each leather, The curvature of the curve formed by the curvature is generated within a certain range, and so on. However, the pattern of the leather can be defined as having any characteristic other than the above-mentioned characteristic, and is not limited by the above-mentioned examples. However, the image portion 610 of FIG. 6A, which segments the region in a straight line, can be regarded as outside the scope of this " consistent feature ".

FIG. 6B is a diagram showing a visual representation of exemplary data output by the network function based on the input data shown in FIG. 6A, in the method of determining the state of image data according to an embodiment of the present disclosure. As described above, the network function (or the auto encoder 400) has a relatively low information loss rate for the " coherent feature " existing in the input data, and for the data having the " (S310). ≪ / RTI > According to one embodiment of the present invention, the network function 400 can output the output data (the image data shown in Fig. 6B) based on the image data based on the image shown in Fig. 6A. As shown in FIG. 6B, the network function according to an exemplary embodiment of the present invention can output output data having a relatively low information loss with respect to a consistent feature of the leather pattern across input data. However, in the case of the image portion 610 of the input data 600A having a different pattern, the information loss rate on the output data may be relatively large. 6B, the image portion 600B of the output data corresponding to the image portion 610 having a different pattern may include information (e.g., a portion 620 indicated by a straight line) according to a different pattern, Can be lost.

7A is a diagram illustrating another exemplary input data of a method for determining the state of image data according to an embodiment of the present disclosure. As shown in Fig. 7A, similar to Fig. 5A, the input data may be image data obtained by photographing a photograph such as leather. However, unlike FIG. 5A, the image shown in FIG. 7A may include portions of the image having " different features " with attributes not included in the coherent features present in the pattern of the leather. For example, as shown in FIG. 7A, the image 700A may include an image portion 710 that is displayed as a circular black dot on the skin. In general, the pattern of the leather is defined according to the bending of the leather, as shown in Fig. 5A. For example, the bending of the leather may have a similar width to the regions produced by the bending of each leather, The curvature of the curve formed by the curvature is generated within a certain range, and so on. However, the pattern of the leather can be defined as having any characteristic other than the above-mentioned characteristic, and is not limited by the above-mentioned examples. However, the image portion 710 of FIG. 7A, which segments the region in a straight line, can be regarded as outside the scope of such a " consistent feature ".

FIG. 7B is a diagram showing, in a method of determining the state of image data according to an embodiment of the present disclosure, a graphical representation of exemplary data output by a network function based on the input data shown in FIG. 7A. FIG. As described above, the network function (or the auto encoder 400) has a relatively low information loss rate for the " coherent feature " existing in the input data, and for the data having the " (S310). ≪ / RTI > According to one embodiment of the present invention, the network function 400 can output the output data (the image data shown in Fig. 7B) based on the image data based on the image shown in Fig. 7A. As shown in FIG. 7B, the network function according to an exemplary embodiment of the present invention can output output data with a relatively small information loss with respect to a consistent feature of the leather pattern over the input data. However, in the case of the image portion 710 of the input data 700A having a different pattern, the information loss rate on the output data may be relatively large. 7B, the image portion 700B of the output data corresponding to the image portion 710 having a different pattern may include information according to a different pattern (for example, the portion 720 indicated by the circular dot) ) Can be lost.

Returning to Fig. 3, in step S330, the same image data as input to the network function in step S320 may be modulated by an algorithm having a different effect from the network function. An algorithm having a different effect from the network function according to an embodiment of the present disclosure may be any algorithm that loses or modifies at least a portion of the input image data. Thus, the second output data may be an image with lower resolution than the input image data or an image including a crushing effect. The second output data can be quantified in various ways. For example, the number of closed curves of a specific color of the second output data can be analyzed and quantified. Further, the brightness value can be measured based on the brightness of the whole color of the second output data. For example, an algorithm with different effects may be an image reduction and enlargement algorithm that loses information of the input image data. However, an algorithm having a different effect is not limited by the above-described example, and any algorithm that loses information or lowers resolution of input image data may be used.

For example, according to an embodiment of the present disclosure, an algorithm having a different effect from a network function divides image data into images of a predetermined size, and then, based on the average data value of the divided images, Lt; / RTI > The second output data may be data that can be compared with the first output data through loss or modification of the data based on the input image data. For example, an image patch of 256 × 256 size may be divided into 4096 image patches of 4 × 4, and an average of colors of 4 × 4 image patches classified may be expressed to make the entire image mosaic. As another example, the algorithm may be an algorithm that reduces input image data to 1 / n magnification and again magnifies n times to obtain second output data whose data has been modified through size variation. The image data inputted through the algorithm may be deformed in size to include a crushing effect of the second output data. Here, the value of the magnification n to be reduced or enlarged may be determined based on the size of the image patch extracted from the network function. As another example, the algorithm may be an algorithm that transforms the color mode to obtain a transformation of the input image data. If the input image data is the RGB color mode, the second output data may be obtained by changing to the CMYK color mode and then changing to the RAB color mode. The present invention is not limited to the above-described algorithm for losing or modifying the input image data only by way of example.

When step S330 is performed, the operation module 272 can obtain the second output data.

In step S340, the control unit 270 can determine the status information of the image data based on the first data and the second data. In step S310, for example, when the network function is an auto-encoder, due to the characteristics of the auto-encoder having a hidden layer having fewer nodes than the number of nodes of the input layer, Experience. As described above, depending on the state of the input image data, the degree of information loss may be different. However, in general, the input image data may experience degradation in resolution due to loss of information.

As described above, when the characteristics of the input image data match the characteristics of the network function that is learned by the comparison in step S310, information loss and distortion may be relatively small. That is, when the input image data is, for example, normal data, the output data passed through the network function may have a high similarity with the normal data. However, even in this case, there may be a difference between the input data and the output data due to information loss and distortion. Therefore, in step S320, the state determination module 273 directly comparing the input data with the output data may degrade the accuracy of the state determination of the image data.

Thus, the computing module 272 can obtain second output data with loss of information, which is suitable for comparison with the first output data. The different algorithm may be an algorithm having a data loss rate similar to the general data loss rate of the network function of step S320. For example, the different algorithm of step S330 may be an algorithm having a resolution lowering rate similar to the resolution lowering rate due to the network function of step S320. For example, if the different algorithm of step S330 is a reduction and enlargement algorithm of image data, the magnification of reduction and enlargement can be determined such that the resolution lowering rate is similar to the resolution lowering rate of the network function of step S320.

Since the second output data acquired by the calculation module 272 has a data loss rate similar to that of the first output data, direct comparison can be facilitated. Therefore, the state determination module 273 can determine the state of the image data based on the similarity of the first output data and the second output data.

In step S340, the state determination module 273 may use any algorithm to determine the similarity of the two data. For example, the status determination module 273 can determine the similarity of data based on the difference image between the first output data and the second output data. In another embodiment, the status determination module 273 may calculate a Normalized Cross Correlation (NCC) value in various kernels based on the first output data and the second output data. The status determination module 273 may determine that the NCC value is a defect when the NCC value is smaller than a predetermined reference value, and may determine that the NCC value is normal when the NCC value is greater than the reference value. The kernel may have various sizes such as 16x16, 24x24, and 32x32. However, this disclosure is not limited by the size of the kernel.

According to one embodiment of the present disclosure, the status determination module 273 can determine the status information of the image data based on the difference image of the first output data and the second output data. Based on the first-order image data, which is the difference image between the first output data and the input image data, and the second-order image data that is the difference image between the second output data and the input image data, the status determination module 273 determines, Can be determined. The state determination module 273 may calculate the difference image value between the first and second image data and the second image data, and may calculate the NCC value for the kernels of various sizes. The state determination module 273 may calculate the difference image as black if the compared pixels are the same and include other colors, and if the ratio of black of the difference image is not less than the reference value (for example, 95%) The status information of the image data can be determined from the normal image data. The method of comparing the first output data with the second output data includes a method of calculating a difference by calculating a binarization value for a specific point, a method of comparing the number of closed curves by searching for a contour, And a method of calculating the distance can be used. The present disclosure is not limited to the above methods of comparing the first output data and the second output data and the ratio value.

FIG. 8A is a diagram illustrating exemplary data obtained by obtaining a difference image based on the images of FIGS. 5A and 5B in the method of determining the state of image data according to an embodiment of the present disclosure. The difference image of FIG. 8A may be the result of obtaining a difference image of the image of FIG. 5B, which is exemplary data of the output data having characteristics consistent with the image of FIG. 5A, which is input data. Since Fig. 5A is normal input data without a defect part, output data which is an image of Fig. 5B may have a low data loss. As a result, the difference between the input data and the output data is small, so that the ratio of the black portion of the difference image may be larger than that obtained from the abnormal image. The smaller the difference, the more black areas can appear. Therefore, if the data loss of the output data obtained through the network function is small, the proportion of the black portion may be relatively large. The image data may be determined based on the difference image in FIG. 8A, and the image data may be determined based on the calculation of the NCC value or the like. The method of comparing the image data is merely an example, and the present disclosure is not limited thereto.

FIG. 8B is a diagram showing another exemplary data obtained by obtaining a difference image based on the images of FIGS. 6A and 6B in the method of determining the state of image data according to an embodiment of the present disclosure. The difference image of FIG. 8B may be the result of obtaining the difference image of the image of FIG. 6B, which is exemplary data among the output data having characteristics different from the image of FIG. 6A, which is input data. 6A is data having a defective portion, the output data, which is an image of FIG. 6B, may have a relatively large data loss rate. As a result, the difference between the input data and the output data may be large, and the ratio of the black portion of the difference image may be smaller than the ratio obtained from the normal image. The smaller the difference, the more black areas can appear. Specifically, the portion corresponding to the image portion 610 indicated by the straight line in FIG. 6A may not be displayed in black in FIG. 8B. Therefore, if the data loss of the output data obtained through the network function is large, the proportion of the black portion may be relatively small. The state of the image data may be determined based on the difference image of Fig. 8B, and the state of the image data may be determined based on the calculation of the NCC value or the like. The method of comparing the image data is merely an example, and the present disclosure is not limited thereto.

FIG. 8C is a diagram that visually shows other exemplary data obtained from a difference image based on the images of FIGS. 7A and 7B in the method of determining the state of image data according to an embodiment of the present disclosure. The difference image of FIG. 8C may be the result of obtaining the difference image of the image of FIG. 7B, which is exemplary data among the output data having the characteristics different from the image of FIG. 7A which is the input data. 7A is data having a defective portion, the output data, which is the image of FIG. 7B, may have a relatively large data loss rate. As a result, the difference between the input data and the output data may be large, and the ratio of the black portion of the difference image may be smaller than the ratio obtained from the normal image. The smaller the difference, the more black areas can appear. Specifically, a portion corresponding to the image portion 710 indicated by the circular black point in FIG. 7A may not be displayed in black in FIG. 8C. Therefore, if the data loss of the output data obtained through the network function is large, the proportion of the black portion may be relatively small. The image data may be determined based on the difference image in FIG. 8C, and the image data may be determined based on the calculation of the NCC value or the like. The method of comparing the image data is merely an example, and the present disclosure is not limited thereto.

According to one embodiment of the present disclosure, a method for determining the state of image data further includes determining at least one of a node, a weight, and a connection state of a network function based on preset image data prior to obtaining the first output data You may. The node and weight determination module 274 may modify at least one of the node, the weight, and the connection state of the network function to learn the network function on the preset image data. The node and weight determination module 274 may change at least one of the node, weight, and connection state of the network function based on the same or similar portion of the feature portion of the preset image data.

According to an embodiment of the present disclosure, a method for determining the state of image data may further include repeating the step of determining at least one of a node, a weight, and a connection state of a network function, May be one or more normal image data. The step of determining at least one of the node, the weight, and the connection state of the network function may be repeated until there is no change in the network function. The node and weight determination module 274 may repeat the operation of determining nodes, weights, and connection states for each of the various image patch sizes extracted by the network function. The node and weight determination module 274 determines the network function as a learned network function for at least one pattern if there is no change in the network function, inputs the image data to the learned network function, Can be obtained. The preset image data may be normal image data without defects and may not be the same data. The calculation module 272 can output the first output data which is relatively different from the image data inputted through the learned network function using the normal image data. The node and weight determination module 274 determines at least one of the nodes of the network function, the weighted connection state, so that the difference between the input image data and the first output data is reduced, and sends a decision to reduce the difference to the other image data Can be repeated.

FIG. 9 is a conceptual diagram for explaining a state determination structure of image data according to an embodiment of the present disclosure; FIG.

The same image data may be included in the network function and the network function and the different algorithm. The first output data includes an output value by a network function, and the second output data includes an output value by an algorithm different from the network function. The first output data and the second output data may have at least one of image resolution, size, color mode, and aspect ratio equal or comparable. The status information of the image data may be status information determined based on the first output data and the second output data. The above example of the same data is merely an example, and the present disclosure is not limited thereto.

The status determination module may generate status information of the image data based on the first output data and the second output data. The status information of the image data may be normal, defective, or may include one or more classes of information. The status information of the image data may also include specific bad data or comments on the status information.

10 is a flowchart of a method for determining the state of image data according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, a method for determining the state of image data can be configured without including a pre-learning step. The method for determining the state of image data includes the steps of acquiring first output data by a network function (S620), acquiring second output data by an algorithm having an effect different from that of the network function (S630) And determining status information of the image data based on the second output data (S640). Each step can be similarly described in correspondence with the steps of FIG. 3 described above.

11 is a flowchart illustrating a method of determining the state of image data using two or more network functions according to another embodiment of the present disclosure.

According to another embodiment of the present disclosure, a method for determining the state of image data may include obtaining (S710) the third output data by the first network function. The network function learned for one pattern can output consistent third similar output data when normal data is input. If the learned network function for one pattern receives defective data, the defective data may have different compression results than the normal data. Also, when the learned network function receives the defective data, the third output data obtained by restoring the normal image data may be different from the third output data obtained by restoring the defective image data.

According to another embodiment of the present disclosure, a method for determining the state of image data may include obtaining (S720) the fourth output data by the second network function different from the first network function. And the step of acquiring the fourth output data may be performed based on the third output data. The obtaining of the fourth output data may be by a second network function to recover at least one lost portion of the lost third output data. For example, the second network function that is the restoring network function may be a network function that restores the resolution reduced by the first network function to the same resolution as the input image data. The second network function may be a combination of a plurality of network functions, and a network function for compressing and / or restoring another network function may be connected after the second network function.

According to another embodiment of the present disclosure, the second network function may be a restoration network function that restores at least a portion of the image data lost by the first network function. The fourth output data obtained through the restoration network function may have the same image resolution, size, and the like as the image data. The fourth output data, and the image data may include configurations having the same or similar values so that they can be compared. The restoration network function may modify at least a portion of the third output data to restore at least a portion of the image data.

According to another embodiment of the present disclosure, a method for determining the state of image data may include determining (S730) the state information of the image data based on the image data and the fourth output data. The image data input to the first network function and the fourth output data may be data in which at least one of image resolution, size, color mode, and aspect ratio is the same. The image data may be compared with the fourth output data to detect an actual defect, and the status information of the image data may be determined based on the same portion. Specifically, the method of determining the state information of the image data may be a method of determining the NCC value, the difference image, the binarization, the closed curve of the contour, the relative distance between the reference lines, etc., and determining the difference based on the reference value. The example of the same data and the method of determining the state information of the image data are merely examples, and the present disclosure is not limited thereto.

12 is a conceptual diagram for explaining a status determination structure of image data according to another embodiment of the present disclosure;

The second network function 830 may be a restoration network function that restores at least a portion of the image data lost by the first network function 810 in accordance with another embodiment of the present disclosure. The fourth output data 840 obtained through the restoration network function may have the same image resolution, size, and the like as the image data. The fourth output data 840 and the image data may include constructions having the same numerical value so that they are comparable. The restoration network function may modify at least a portion of the third output data 820 to restore at least a portion of the image data.

The fourth output data 840 may be data obtained by restoring data of at least a part of the lost third output data 820. [ The fourth output data 840 may include portions similar to image data. For example, the fourth output data has the same resolution as the image data, and the status determination module can generate the status information of the image data based on the same resolution. The second network function 830 may be a network function that causes the image data and at least some configuration values to be configured the same or similar.

13 is a conceptual diagram illustrating a network function to which two or more network functions are connected according to another embodiment of the present disclosure.

The first network function 910 and the second network function 920 may be connected to form a network function according to an embodiment of the present disclosure. The first network function 910 may comprise third output data that is a collection of output nodes. The second network function 920 may include fourth output data that is a collection of other output nodes. The first network function may analyze input image data and assign it to a set of input nodes. The value input to the node can be calculated to the next node by a link or a weight. The number of nodes of the network function, the connection status, and the layer structure can be arranged in a configurable combination of combinations. The network functions are not arranged in a layer form but can be configured according to the arrangement of arbitrary nodes. The network function may include a network function of the extended algorithm rather than a network function of the algorithm being compressed. Not all nodes of the network function may be connected to the link, and there may be nodes that are not connected.

A computer program stored in a computer-readable storage medium executed by one or more processors of the apparatus for determining the state of image data in accordance with one embodiment of the present disclosure may comprise a plurality of instructions. The computer program comprising instructions for causing a computation module to obtain first output data by a network function based on the image data, instructions for causing the computation module to perform an algorithm based on the image data, Instructions for obtaining second output data, and instructions for the status determination module to determine status information of the image data based on the first output data and the second output data.

According to another embodiment of the present disclosure, a computer program stored in a computer-readable storage medium executed by one or more processors of the apparatus for determining the state of image data may comprise a plurality of instructions. The computer program comprising instructions for obtaining third output data by a first network function based on the image data, instructions for obtaining a third output data by a second network function different from the first network function based on the third output data, 4 output data, and instructions for determining the status information of the image data based on the image data and the fourth output data.

Those of ordinary skill in the art will understand that information and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that may be referenced in the above description may include voltages, currents, electromagnetic waves, magnetic fields or particles, Particles or particles, or any combination thereof.

Those skilled in the art will appreciate that the various illustrative logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented or performed with a specific purpose, (Which may be referred to herein as "software") or a combination of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the design constraints imposed on the particular application and the overall system. Those skilled in the art may implement the described functions in various ways for each particular function, but such implementation decisions should not be interpreted as being outside the scope of the present disclosure.

The various embodiments presented herein may be implemented as a method, apparatus, or article of manufacture using standard programming and / or engineering techniques. The term "article of manufacture" includes a computer program, carrier, or media accessible from any computer-readable device. The medium may comprise a storage medium or a communication medium. For example, the computer-readable storage medium can be a magnetic storage device (e.g., a hard disk, a floppy disk, a magnetic strip, etc.), an optical disk (e.g., CD, DVD, But are not limited to, memory devices (e. G., EEPROM, card, stick, key drive, etc.).

Communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal, such as a carrier wave or other transport mechanism, Media. The term modulated data signal refers to a signal that has one or more of its characteristics set or changed to encode information in the signal. By way of example, and not limitation, communication media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared, or other wireless media.

It will be appreciated that the particular order or hierarchy of steps in the presented processes is an example of exemplary approaches. It will be appreciated that, based on design priorities, a particular order or hierarchy of steps in the processes may be rearranged within the scope of this disclosure. The appended method claims provide elements of the various steps in a sample order, but are not meant to be limited to the specific order or hierarchy presented.

The description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present disclosure should not be construed as limited to the embodiments set forth herein, but is to be accorded the widest scope consistent with the principles and novel features presented herein.

Claims (1)

A method for determining the state of image data.

Images