JPH10255035A - Image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform a prescribed image conversion about acquired image information and to provide an image that is accurate and also has much communication information quantity. SOLUTION: A density conversion processing part 4 which inputs image data from an original image supplying part 1 and performs density conversion processing so that a density characteristic which is held by the image data and an image conversion processing characteristic which is held by a neural network image conversion processing part 2 may be adapted to each other is provided between the part 1 which outputs image data that is an object to be processed and the part 2 which inputs the image data from the part 1 and performs image conversion processing. Thereby, a prescribed image conversion is performed of the acquired image information, and an image that is accurate and has also much transmission information quantity is provided.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for irradiating or penetrating an object with an information carrier such as an electromagnetic wave or an ultrasonic wave including light or X-rays, and measuring a change in the information carrier to obtain image information. The present invention relates to an image processing apparatus such as a medical image diagnostic apparatus or a broadcast video processing apparatus or an industrial image processing apparatus or a consumer image capturing and displaying apparatus. The present invention relates to an image processing apparatus capable of providing a large amount of images.

[0002]

2. Description of the Related Art As shown in Japanese Patent Application Laid-Open No. 8-153194, a conventional image processing apparatus of this type includes an original image supply unit for outputting image data to be processed, an artificial neural element and an input layer. An image conversion processing unit consisting of a neural network configured to form a network by combining the intermediate layer and the output layer so as to form a layer structure and to perform image conversion processing by inputting image data from the original image supply unit; And a display unit for inputting and displaying the image information output after the image conversion processing by the processing unit. Then, image data is output from the original image supply unit, and the image data is directly input to an image conversion processing unit composed of a neural network. In the image conversion processing unit, the image data is converted into image conversion processing characteristics of the neural network. The image is converted based on the image data, sent to a display unit, and displayed.

[0003]

However, in such a conventional image processing apparatus, the image data output from the original image supply section is directly input to the image conversion processing section comprising a neural network. In some cases, the image conversion processing unit does not perform image conversion processing that sufficiently matches the density characteristics of the input image data. That is, the density characteristic of the image data from the original image supply unit varies depending on the X-ray irradiation conditions when the image data is data obtained from, for example, an X-ray imaging apparatus, and is obtained from a consumer image capturing and displaying apparatus. In the case of such data, it changes depending on conditions such as the aperture of the camera and weather. As described above, the density characteristic of the image data from the original image supply unit changes according to the conditions at the time of collecting the image data. And
When such image data whose density characteristics change is directly input to an image conversion processing unit composed of a neural network having a constant image conversion processing characteristic, the density characteristics of the image data and the image conversion processing unit have The image conversion characteristics do not match, and the image data is output without sufficient image conversion, and is displayed on the display unit. The image data output and displayed in this manner is inaccurate and inconsistent, for example, causing a problem that the contrast of a region of interest to be viewed is reduced, some information is lost, or noise is conspicuous. An image with a small amount of information was sometimes obtained.

Accordingly, the present invention addresses such a problem and provides an image processing apparatus which can perform a predetermined image conversion on the obtained image information and provide an image which is accurate and has a large amount of transmitted information. The purpose is to:

[0005]

In order to achieve the above object, an image processing apparatus according to a first aspect of the present invention comprises an original image supply section for outputting image data to be processed, an artificial neural element comprising an input layer and an intermediate layer. A neural network image conversion processing unit for forming a network by combining layers and an output layer so as to form a layer structure, inputting image data from the original image supply unit and performing image conversion processing, and the neural network image conversion processing A display unit for inputting and displaying the image information output by the image conversion processing in the unit, wherein the original image supply unit and the neural network image conversion processing unit Input image data from the image processing unit so that the density characteristics of the image data match the image conversion processing characteristics of the neural network image conversion processing unit It is provided with a density conversion processing section for performing density conversion processing.

The density conversion processing unit is applied to a combination of a multiplier and an adder for calculating image data input from the original image supply unit and a predetermined value, or to image data input from the original image supply unit. It consists of a table memory that stores density conversion functions.

Further, the density conversion processing section calculates a histogram by inputting image data from the original image supply section and calculates a histogram, and calculates a density conversion function using the histogram data from the histogram calculation section. A density conversion function calculation unit; and a density conversion unit that receives the image data from the original image supply unit and the density conversion function from the density conversion function calculation unit and performs density conversion on the image data. It may be.

Further, the original image supply unit outputs image data to be processed in a time series, and the neural network image conversion processing unit outputs a predetermined area including at least a pixel of interest from the original image supply unit. May be input to perform image conversion processing and output data corresponding to the target pixel.

An image processing apparatus according to a second aspect of the present invention comprises: an original image supply unit for outputting image data to be processed;
Neural network image conversion processing in which artificial neural elements are connected to form an input layer, an intermediate layer, and an output layer so as to form a layer structure, form a network, and input image data from the original image supply unit to perform image conversion processing. And a display unit for inputting and displaying the image information output and converted by the neural network image conversion processing unit, the original image supply unit and the neural network image conversion processing Image data from the original image supply unit, and performs density conversion processing such that the density characteristics of the image data match the image conversion processing characteristics of the neural network image conversion processing unit. A conversion processing unit is provided, and a neural network image conversion is provided between the neural network image conversion processing unit and the display unit. The image information output to image conversion processing management unit is provided with a density inversion processing unit for performing inverse transform processing of the transform processing in the density conversion processing section.

The density conversion processing unit is applied to a combination of a multiplier and an adder for calculating image data input from the original image supply unit and a predetermined value, or to image data input from the original image supply unit. A density conversion function storing table memory, wherein the density inverse conversion processing unit is a combination of a divider and a subtracter for calculating image information input from the neural network image conversion processing unit and a predetermined value, or a neural network image It consists of a table memory storing a density inverse conversion function to be applied to the image information input from the conversion processing unit.

Further, the density conversion processing section receives the image data from the original image supply section to calculate a histogram, and calculates a density conversion function using the histogram data from the histogram calculation section. A density conversion function calculation unit; and a density conversion unit that receives the image data from the original image supply unit and the density conversion function from the density conversion function calculation unit and performs density conversion on the image data. It may be.

Further, the original image supply unit outputs image data to be processed in a time series, and the neural network image conversion processing unit outputs a predetermined area including at least a pixel of interest from the original image supply unit. May be input to perform image conversion processing and output data corresponding to the target pixel.

[0013]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. FIG. 1 is a block diagram showing an embodiment of the image processing apparatus according to the first invention.
The image processing apparatus irradiates or penetrates an object with an information carrier such as an electromagnetic wave or an ultrasonic wave including light or X-rays, and obtains image information by measuring a change in the information carrier. It is a medical image diagnostic apparatus, a broadcast video processing apparatus, an industrial image processing apparatus, a consumer image capturing and displaying apparatus, or the like. As shown in FIG.
, A neural network image conversion processing unit 2 and a display unit 3, and further includes a density conversion processing unit 4.

The original image supply unit 1 outputs image data to be processed. For example, the original image supply unit 1 outputs medical data such as an X-ray photographing apparatus, an ultrasonic tomography apparatus, an X-ray CT apparatus, a nuclear medicine imaging apparatus, or a magnetic resonance imaging apparatus. A measuring unit that measures or captures an object in an image diagnostic device, a broadcast video processing device, an industrial image processing device, a consumer image capturing and displaying device, or the like, or stores image data measured for the object and reads out the data as necessary. It is a storage unit or the like. Also,
The neural network image conversion processing unit 2 performs a required image conversion process by inputting the image data output from the original image supply unit 1, and converts the artificial neural elements into the input layer, the intermediate layer, and the output layer. A network is formed by connecting them so as to form a structure. Further, the display unit 3
Is a device for inputting and displaying the image information output by the image conversion processing by the neural network image conversion processing unit 2, and comprises, for example, a television monitor or a printer.

The neural network image conversion processing unit 2 comprises a network 5 as shown in FIG. The network 5 receives time-series data of a predetermined area including at least a pixel of interest from the original image supply unit 1 to perform image conversion processing and output data corresponding to the pixel of interest. 6 are connected so as to form a layer structure of an input layer, an intermediate layer, and an output layer, and are configured using a linear function as an input / output function of the final output layer. FIG. 3 is an explanatory view showing the artificial neural element 6 constituting the network 5. As shown in the figure, an artificial neural element (hereinafter, referred to as a “neuron model”) 6 is a multi-input, one-output element that simulates the function of a biological neural element (neuron).
(I ₁ -In) and the output Oj by the product sum of the coupling coefficient Wji (Wj ₁ ~Wjn) is determined.

FIG. 2 is an explanatory diagram showing the hierarchical structure of the neural network image conversion processing unit 2 which is connected to form a layer structure and is formed into a network. As shown in the figure, the network 5 uses a large number of the neuron models 6 described above, and connects the input layer, the intermediate layer, and the output layer so as to form a layer structure, thereby realizing functions of signal processing and information processing. Have been. In FIG. 2, reference numeral 7 denotes a branch connecting the input layer and the intermediate layer, and each neuron model 6 in the input layer is connected to all the neuron models 6 in the intermediate layer. Reference numeral 8 denotes a branch connecting the intermediate layer and the output layer. Each neuron model 6 in the intermediate layer is connected to all the neuron models 6 in the output layer. And this network 5
The input information 9 supplied to the input layer is converted and output as output information 10 from the output layer. An identity function is used as an input / output function of the input layer of the network 5, and a sigmoid function is used as an input / output function of the hidden layer.

With such a configuration, the network 5 is an information processing mechanism having a learning function and a self-organizing function. And this network 5
By giving a teacher signal 11 and performing learning, self-organization is performed so that desired output information 10 is output from the output layer with respect to the input information 9 supplied to the input layer.

Here, in the present invention, as shown in FIG. 1, a density conversion processing section 4 is provided between the original image supply section 1 and the neural network image conversion processing section 2. The density conversion processing unit 4 receives the image data from the original image supply unit 1 so that the density characteristics of the image data match the image conversion processing characteristics of the neural network image conversion processing unit 2. The density conversion processing is performed, and the internal configuration is composed of a combination of a multiplier and an adder for calculating image data input from the original image supply unit 1 and a predetermined value, or input from the original image supply unit 1. It consists of a table memory storing density conversion functions applied to image data. As an example of a specific configuration, as shown in FIG. 4, a multiplier 13 that multiplies image data input from the original image supply unit 1 by a predetermined value set by the first input unit 12, And an adder 15 for adding a predetermined value set by the second input means 14 to the image data output from the multiplier 13. The first and second input means 12 and 14 include, for example, a volume, a keyboard or a memory.

Next, the operation of the image processing apparatus thus configured will be described. First, in FIG. 1, an original image supply unit 1 irradiates or penetrates an object with an information carrier, and outputs image data obtained by measuring a change in the information carrier as a processing target. This may be data actually measured by various medical image diagnostic apparatuses, broadcast video processing apparatuses, industrial image processing apparatuses, or consumer image capturing and displaying apparatuses, or may be measured in advance for an object and written to the storage unit. The read data may be read.

Next, the image data output from the original image supply unit 1 is input to a density conversion processing unit 4. The density conversion processing section 4 performs density conversion processing of the image data from the original image supply section 1. This density conversion process is performed by, for example, manually inputting the data to the multiplier 13 or the adder 15 by the first and second input means 12 and 14 shown in FIG. The image data from 1 is multiplied by a multiplier 13 and
The addition is performed by the adder 15 to the multiplication result. Alternatively, although not shown, the density conversion is performed by applying a density conversion function preset and stored in a table memory constituting the density conversion processing unit 4 to the image data from the original image supply unit 1. By performing the density conversion process in this manner, the image data from the original image supply unit 1 is determined to have a density characteristic that matches the image conversion processing characteristic of the neural network image conversion processing unit 2 shown in FIG. Data can be supplied to the neural network image conversion processing unit 2.

Next, the image data output from the density conversion processing section 4 is input to the neural network image conversion processing section 2. The neural network image conversion processing unit 2 performs an image conversion process on the input image data in a self-organized state learned in advance by the network 5 shown in FIG. Then, the image information output by the image conversion processing by the neural network image conversion processing unit 2 is input to the display unit 3. The display unit 3 inputs the image information output by the image conversion processing by the neural network image conversion processing unit 2 and displays it as an image.

Next, the learning operation of the network 5 in the neural network image conversion processing unit 2 will be described. FIG. 5 is a block diagram showing the overall configuration of the network 5 at the time of learning. During this learning,
Normally, the density conversion processing section 4 does not perform the density conversion processing,
The image data from the original image supply unit 1 is directly supplied to the neural network image conversion processing unit 2. At this time, in the network 5 of the neural network image conversion processing unit 2, learning is performed so that desired output information 10 is output with respect to the input information 9 shown in FIG. That is, the image data supplied from the original image supply unit 1 shown in FIG. 5 (see reference numeral 9 in FIG. 2) is compared with the teacher image supplied from the teacher image supply unit 16 (see reference numeral 11 in FIG. 2). Branches 7 and 8 connecting each layer of the layer structure of the network 5 shown in FIG. 2 so that an error from output image data (see reference numeral 10 in FIG. 2) output from the network image conversion processing unit 2 is reduced. Is changed, and self-organization is performed so that a desired output is obtained after learning. After the learning, a network 5 for realizing a desired image conversion process is constructed.

FIG. 6 shows a learning algorithm of the analog prediction network 5 using the input / output function of the output layer shown in FIG. 2 as a linear function. The meanings of the symbols in FIG. 6 are as follows. I; input information Ok; output of the k-th layer neuron model Wk; coupling coefficient θk between the k-th layer and the (k + 1) -th layer; offset T of the k-th layer neuron model T; teacher signal ηw; learning constant for the coupling coefficient W .eta..theta .; learning constant for offset .theta.k; error amount fi for correcting the (k-1) th layer fi; identity function fs; sigmoid function fl; linear function Note that input information I and output of the neuron model of the kth layer O
k, the offset θk of the neuron model of the k-th layer, the teacher signal T, and the error amount δk for correcting the (k−1) -th layer are represented by a one-dimensional vector, and the k-th layer and the (k + 1) -th layer Is represented by a two-dimensional vector. Also,
Since the network 5 shown in FIG. 2 has three layers, the first layer is an input layer, the second layer is an intermediate layer, and the third layer is an intermediate layer.
The layer is the output layer.

The learning algorithm will be described below. Mean square error E between the teacher signal T and the output O _{3 of the} output layer
Is It is. In order to reduce this, the coupling coefficient Wk between the k-th layer and the (k + 1) -th layer is changed by the following correction amount. Modify by here, In other words, δ ₃ that determines the amount of correction between the hidden layer and the output layer
Is It is. Therefore, the amount of correction is Becomes Thereby, the output from the output layer becomes an analog value as described above. Δ ₂ that determines the amount of correction between the input layer and the output layer is And the correction amount is Becomes The algorithm shown in FIG. 6 is obtained by similarly calculating the offset.

Then, based on the teacher signal 11 (see FIG. 2) supplied by the teacher image supply unit 16 shown in FIG. 5, learning of the network 5 shown in FIG. A neural network that realizes the conversion process is automatically constructed. Although an example of a three-layer neural network has been described here, a neural network having three or more layers can be configured by increasing the number of intermediate layers. In this case, the learning algorithm performs back propagation of the error amount δ. The correction amount may be obtained in a recursive manner, and it is possible to deal with image information including more complicated image conversion processing. Then, the image information converted by the neural network image conversion processing unit 2 is
Although not shown, it is converted into an appropriate density scale, input to the display unit 3 and displayed.

FIG. 7 is a block diagram showing another embodiment of the internal configuration of the density conversion processing section 4 shown in FIG. The density conversion processing unit 4 according to the present embodiment, as shown in FIG.
It comprises a histogram calculator 17, a density conversion function calculator 18 and a density converter 19. The histogram calculation unit 17 receives image data from the original image supply unit 1 shown in FIG. 1 and calculates a histogram of the image data, and includes, for example, an arithmetic unit. The density conversion function calculator 18 receives the histogram data from the histogram calculator 17 and calculates a density conversion function using the histogram data. The density conversion function calculator 18 includes, for example, an arithmetic unit. Further, the density conversion unit 19 is provided with the original image supply unit 1.
, And the density conversion function from the density conversion function calculation unit 18 is input to perform the density conversion of the image data, and includes, for example, an arithmetic unit or a table memory.

Next, the density conversion processing unit 4 having such a configuration is described.
Will be described. First, the image data from the original image supply unit 1 shown in FIG. Then, the histogram calculator 17 calculates a histogram of the effective area of the input image data. Next, the histogram data output from the histogram calculator 17 is input to the density conversion function calculator 18. Then, the density conversion function calculator 18 calculates a density conversion function using the input histogram data.

At this time, data values Gmin and Gmax of a predetermined percentage above and below the area of the histogram from the histogram calculating unit 17 are calculated as the effective density range. Then, a density conversion function that matches the preset values G ₀ min and G ₀ max is calculated. For example, when the density conversion function is a linear function such as f (x) = gain · x + offset, the above values G ₀ min and G ₀ max are replaced by G ₀ min = gain · Gmin + offset G ₀ max = gain · Gmax + offset By doing so, the parameters of the density conversion function can be calculated as in the following equation. gain = (G ₀ min−G ₀ max) / (G min−G max) offset = G ₀ min − ｛(G ₀ min−G ₀ max) / (G min−G m
ax)｝ Gmin Such a calculation is performed by the density conversion function calculation unit 18.

Next, the density conversion function obtained by the density conversion function calculation section 18 and the image data from the original image supply section 1 are input to a density conversion section 19. Then, the density conversion unit 19 performs density conversion on the image data as in the following equation. f (x) = gain · x + offset where x is the image data from the original image supply unit 1 and f (x) is the output image data after the conversion processing. Note that any function can be used as the density conversion function, and is determined based on a preset value.

FIG. 8 is a block diagram for explaining another example of the calculation of the density conversion function in the density conversion processing section 4 having the above configuration. In this example, a characteristic calculation unit 20 is connected to the neural network image conversion processing unit 2, and the characteristic calculation unit 20 controls the density conversion processing unit 4 according to the image conversion processing characteristics of the neural network image conversion processing unit 2. The density conversion function calculator 18 calculates the density conversion function. In the characteristic calculation unit 20, the same value is input to all inputs of the neural network image conversion processing unit 2 and the value is changed, and the output value of the neural network image conversion processing unit 2
The input / output characteristics of the neural network image conversion processing unit 2 are obtained. Then, the minimum value and the maximum value of the non-saturation region of the input-output characteristics obtained above, for example, the aforementioned value G ₀ min, G
By setting it to ₀ max, the density conversion function of the density conversion function calculation unit 18 can be calculated according to the image conversion processing characteristics of the neural network image conversion processing unit 2.

FIG. 9 is a block diagram for explaining another example of the calculation of the density conversion function in the density conversion processing section 4 having the configuration shown in FIG. In this example, another histogram calculation unit 21 is connected to the teacher image supply unit 16 shown in FIG. 5, and the histogram calculation unit 21 calculates a histogram of the teacher image of the teacher image supply unit 16. Then, the calculated histogram is sent to the density conversion function calculation section 18 in the density conversion processing section 4. Thus, the density conversion function calculation unit 18 calculates the values G ₀ min and G ₀ max using the histogram of the teacher image. For example, a density conversion function can be calculated by setting G ₀ min and G ₀ max to data of a predetermined percentage above and below the area of the obtained histogram.

FIG. 10 is an explanatory view showing another embodiment of the first invention. In this embodiment, it is assumed that the original image supply unit 1 outputs image data to be processed in a time series, and the neural network image conversion processing unit 2 outputs at least a pixel of interest from the original image supply unit 1. The image conversion processing is performed by inputting the time-series data of a predetermined area including the data, and the data corresponding to the pixel of interest is output. As shown in the figure, input information to the neural network image conversion processing unit 2 includes a target pixel 23 on the original image 22 and a plurality of pixels 24 near the target pixel 23.
24 are time-series data 25 obtained by extracting data in a predetermined area including a predetermined time period in the direction of the time axis t. In the example of FIG. 10, the time-series data 25 is a single original image 22.
Since there are five pixels of data in the upper predetermined area and five original images 22 have been extracted within a certain period of time, the data consists of data of 5 × 5 = 25 pixels. The value of each pixel is normalized (not shown) and input to the neural network image conversion processing unit 2. By using such input information, image conversion including spatio-temporal conversion can be performed. Then, in the neural network image conversion processing unit 2, the input information is converted, and
3 is output as the pixel data 26 corresponding to. By moving the pixel of interest 23 on the original image 22 in, for example, a raster scan, image conversion processing of the entire image is performed. Thereby, the converted image 27 is created and output and displayed on the display unit 3.

FIG. 11 is a block diagram showing an embodiment of the image processing apparatus according to the second invention. This image processing apparatus inputs image data from the original image supply unit 1 between an original image supply unit 1 and a neural network image conversion processing unit 2 and calculates density characteristics of the image data and neural network image conversion. A density conversion processing unit 4 for performing density conversion processing so that the image conversion processing characteristics of the processing unit 2 is compatible is provided, and the neural network image conversion processing unit 2 and the display unit 3 include a neural network image. A density reverse conversion processing unit 28 is provided which performs an inverse conversion process of the conversion process in the density conversion processing unit 4 on the image information output after the image conversion processing by the conversion processing unit 2.

The density conversion processing section 4 is a combination of a multiplier 13 and an adder 15 for calculating image data input from the original image supply section 1 and a predetermined value, or input from the original image supply section 1. It comprises a table memory (not shown) storing a density conversion function to be applied to the image data. The density reverse conversion processing section 28 calculates the image information input from the neural network image conversion processing section 2 and a predetermined value. It comprises a combination of the divider 29 and the subtractor 30, or a table memory (not shown) storing an inverse density conversion function applied to the image information input from the neural network image conversion processing unit 2.

According to this embodiment, the image data subjected to the density conversion processing by the density conversion processing section 4 is subjected to the density inverse conversion processing by the density inverse conversion processing section 28, and the original image data of the original image supply section 1 is processed. , And the image can be displayed on the display unit 3 in the state of the original image.

[0036]

The present invention has been configured as described above.
According to the first invention, the image data from the original image supply unit is input between the original image supply unit and the neural network image conversion processing unit, and the density characteristics of the image data and the neural network image conversion processing unit Is provided with a density conversion processing unit that performs a density conversion process so that the image conversion processing characteristics of the neural network image conversion processing unit are compatible with the neural network image conversion processing unit. The image data is subjected to sufficient image conversion, output, and displayed on the display unit.
Therefore, it is possible to provide an accurate image with a large amount of transmitted information by preventing a decrease in contrast of a region of interest to be viewed, preventing loss of information, or reducing noise.

According to the second aspect of the present invention, the image data from the original image supply unit is input between the original image supply unit and the neural network image conversion processing unit, and the density characteristics of the image data and the neural network A density conversion processing unit for performing density conversion processing so that the image conversion processing characteristics of the network image conversion processing unit is provided is provided. The neural network image conversion processing is provided between the neural network image conversion processing unit and the display unit. The density conversion processing unit performs an inverse conversion process of the conversion process in the density conversion processing unit on the image information output by the image conversion process in the unit. Can be returned to the original density characteristic of the image data of the original image supply unit by performing the density inverse conversion processing by the density inverse conversion processing unit. Therefore, the original image can be accurately displayed on the display unit as an image having a large amount of transmitted information.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of an image processing apparatus according to the first invention.

FIG. 2 is an explanatory diagram showing a hierarchical structure of a neural network image conversion processing unit which is combined into a layer structure to form a network.

FIG. 3 is an explanatory diagram showing an artificial neural element constituting a neural network.

FIG. 4 is a block diagram illustrating an example of an internal configuration of an image conversion processing unit.

FIG. 5 is a block diagram showing the overall configuration of the neural network image conversion processing unit when learning a network.

FIG. 6 is an explanatory diagram showing a learning algorithm of a neural network for analog output.

FIG. 7 is a block diagram showing another embodiment of the internal configuration of the density conversion processing unit shown in FIG. 1;

FIG. 8 is a block diagram for explaining another example of calculating a density conversion function in the density conversion processing unit having the above configuration.

9 is a block diagram for explaining still another example of the calculation of the density conversion function in the density conversion processing unit having the configuration shown in FIG. 7;

FIG. 10 is an explanatory view showing another embodiment of the first invention.

FIG. 11 is a block diagram showing an embodiment of an image processing apparatus according to the second invention.

FIG. 12 is a block diagram illustrating an example of an internal configuration of a density conversion processing unit and a density inverse conversion processing unit in the image processing apparatus.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Original image supply part 2 ... Neural network image conversion processing part 3 ... Display part 4 ... Density conversion processing part 5 ... Network 6 ... Artificial nerve element 7, 8 ... Branch 9 ... Input information 10 ... Output information 11 ... Teacher signal 12 First input means 13 Multiplier 14 Second input means 15 Adder 16 Teacher image supply unit 17 Histogram calculation unit 18 Density conversion function calculation unit 19 Density conversion unit 28 Density inverse conversion Processing unit 29: Divider 30 ... Subtractor

Claims (8)

[Claims] 1. An original image supply unit for outputting image data to be processed, and an artificial neural element connected to form an input layer, an intermediate layer, and an output layer to form a layer structure to form a network, and A neural network image conversion processing unit for inputting image data from the unit and performing image conversion processing; and a display unit for inputting and displaying image information output after image conversion processing by the neural network image conversion processing unit. In the image processing apparatus, image data from the original image supply unit is input between the original image supply unit and the neural network image conversion processing unit, and the density characteristics of the image data and the neural network image conversion processing An image processing apparatus, comprising: a density conversion processing unit that performs a density conversion process so that image conversion processing characteristics of the unit match. 2. The density conversion processing unit is applied to a combination of a multiplier and an adder for calculating image data input from an original image supply unit and a predetermined value, or to image data input from an original image supply unit. 2. The image processing apparatus according to claim 1, further comprising a table memory storing a density conversion function. 3. The density conversion processing section receives the image data from the original image supply section and calculates a histogram, and calculates a density conversion function using the histogram data from the histogram calculation section. A density conversion function calculation unit; and a density conversion unit that receives the image data from the original image supply unit and the density conversion function from the density conversion function calculation unit and performs density conversion on the image data. The image processing apparatus according to claim 1, wherein: 4. The original image supply unit outputs image data to be processed in chronological order, and the neural network image conversion processing unit outputs a predetermined area including at least a pixel of interest from the original image supply unit. 4. The image processing apparatus according to claim 1, wherein the apparatus performs image conversion processing by inputting time-series data and outputs data corresponding to the pixel of interest. 5. An original image supplying unit for outputting image data to be processed and an artificial neural element connected to form an input layer, an intermediate layer and an output layer so as to form a layer structure to form a network. A neural network image conversion processing unit for inputting image data from the unit and performing image conversion processing; and a display unit for inputting and displaying image information output after image conversion processing by the neural network image conversion processing unit. In the image processing apparatus, image data from the original image supply unit is input between the original image supply unit and the neural network image conversion processing unit, and the density characteristics of the image data and the neural network image conversion processing A density conversion processing section for performing density conversion processing so that the image conversion processing characteristics of the Between the conversion processing unit and the display unit, a density inverse conversion processing unit that performs an inverse conversion process of the conversion process in the density conversion processing unit on the image information output by the image conversion processing by the neural network image conversion processing unit. An image processing apparatus, comprising: 6. The density conversion processing unit is applied to a combination of a multiplier and an adder for calculating image data input from the original image supply unit and a predetermined value, or to image data input from the original image supply unit. A density conversion function storing table memory, wherein the density inverse conversion processing unit is a combination of a divider and a subtracter for calculating image information input from the neural network image conversion processing unit and a predetermined value, or a neural network image 6. The image processing apparatus according to claim 5, further comprising a table memory storing a density inverse conversion function applied to image information input from the conversion processing unit. 7. A density conversion processing unit which receives image data from an original image supply unit and calculates a histogram, and calculates a density conversion function using the histogram data from the histogram calculation unit. A density conversion function calculation unit; and a density conversion unit that receives the image data from the original image supply unit and the density conversion function from the density conversion function calculation unit and performs density conversion on the image data. The image processing apparatus according to claim 5, wherein: 8. The original image supply unit outputs image data to be processed in a time series, and the neural network image conversion processing unit outputs a predetermined area including at least a pixel of interest from the original image supply unit. 8. The image processing apparatus according to claim 5, wherein time series data is input, image conversion processing is performed, and data corresponding to the pixel of interest is output.