KR20240105039A - Method and apparatus for converting reconstruction wavelength for hologram pattern - Google Patents

Abstract

A method and device for converting wavelengths to reproduce holographic patterns are disclosed. When the reproduction wavelength converter receives a hologram pattern of a first wavelength, it converts the hologram pattern of the first wavelength into a hologram pattern of a second wavelength through an artificial neural network learned to convert the reproduction wavelength and outputs it. This can effectively reduce individual calculations for different holographic displays.

Description

Method and apparatus for converting wavelength for hologram pattern reproduction

Embodiments of the present invention relate to a method and device for converting the reproduction wavelength of a holographic pattern, and more specifically, to a method and device for converting the reproduction wavelength of a holographic pattern using deep learning technology.

Hologram is a technology that records and reproduces object light using the interference phenomenon of light. Computer generated holograms, which are commonly used recently, are created by assuming a target image at a certain depth from a holographic display and calculating a hologram pattern using a numerical propagation model of light. At this time, there are various methods of numerical propagation model, but the most representative method is to assume each pixel of the target image as a point light source and add all the diverging diffraction patterns. Among them, the Angular spectrum method is the most representative.

When calculating a numerical propagation model, the specifications of the holographic display system, such as the wavelength of the light source and the pixel spacing of the display, must be considered. For this reason, in order to reproduce the same content on different holographic display systems, there is a disadvantage that the same calculation process must be performed individually for each system. Holographic pattern synthesis has as a bottleneck a huge computational amount proportional to the product of the number of pixels in the target image and the diffusion angle of the holographic display, so this is an unnecessary computation.

The technical problem to be achieved by the embodiment of the present invention is to replace repetitive calculations using deep learning technology and to synthesize hologram patterns that reflect other holographic display specifications by extracting and interpreting playback image information from existing hologram patterns. The object is to provide a wavelength conversion method and device for reproducing holographic patterns.

An example of a method for converting wavelengths to reproduce a holographic pattern according to an embodiment of the present invention to achieve the above technical problem includes receiving a holographic pattern of a first wavelength; Converting the hologram pattern of the first wavelength into a hologram pattern of the second wavelength through the artificial neural network learned to convert the reproduced wavelength; and outputting a hologram pattern of the second wavelength.

An example of a holographic pattern reproduction wavelength conversion device according to an embodiment of the present invention to achieve the above technical problem includes an artificial neural network learned to convert the reproduction wavelength; and a conversion unit that converts a hologram pattern of a first wavelength into a hologram pattern of a second wavelength through the artificial neural network.

According to an embodiment of the present invention, not only can individual calculations for different holographic displays be effectively reduced, but it can also be used to enhance hologram transmission speed using a deep learning algorithm.

1 is a diagram showing an example of a numerical propagation model used in the present invention;
Figure 2 is a diagram showing an example of a hologram reproduction simulation result according to the reproduction wavelength of the hologram;
3 is a diagram illustrating an example of a hologram pattern generation method according to an embodiment of the present invention;
Figure 4 is a diagram showing an example of an artificial neural network according to an embodiment of the present invention;
5 and 6 are diagrams showing an example of a method for generating training data for an artificial neural network according to an embodiment of the present invention;
Figure 7 is a diagram showing the configuration of an example of a wavelength conversion device for reproducing a holographic pattern according to an embodiment of the present invention, and
8 to 10 are diagrams showing experimental examples of a wavelength conversion method for reproducing a hologram pattern according to an embodiment of the present invention.

Hereinafter, a method and device for converting wavelengths to reproduce holographic patterns according to an embodiment of the present invention will be described in detail with reference to the attached drawings.

Figure 1 is a diagram showing an example of a numerical propagation model used in the present invention.

Referring to Figure 1, holographic display is a technology that can restore three-dimensional information by modulating the wavefront of light. A computer-generated hologram (hereinafter referred to as a hologram) is an interference pattern that is reproduced on a holographic display, allowing the wavefront of light to be modulated on a pixel-by-pixel basis. A hologram can be generated by numerically calculating the free space propagation from the target image plane to the hologram plane. Numerical propagation models that can be used at this time include Angular spectrum method, Fresnel propagation, and Fourier transform. This embodiment uses the Angular spectrum method as a hologram synthesis model, but this is only one of the hologram synthesis methods and other numerical propagation models can be used.

When generating a hologram using a numerical propagation model, the physical elements of the display system are considered. For example, the maximum diffraction angle of a holographic pattern is proportional to the ratio of the display pixel spacing and the wavelength of the light source. For this reason, in order to reproduce one piece of content on a holographic display with different specifications, only the values are different for each individual system, and the same computational process must be performed repeatedly using the holographic numerical propagation model. If the hologram created for wavelength λ ₁ is reproduced with wavelength λ ₂ without repeating the calculation, the target depth will vary and accurate focus control cannot be provided.

Figure 2 is a diagram showing an example of a hologram reproduction simulation result according to the reproduction wavelength of the hologram.

Referring to FIG. 2, simulation results are shown when the hologram is reproduced at the same wavelength as the generation wavelength and when the hologram is reproduced at a different wavelength. A hologram was generated by setting the target wavelength to 638 nm, and both (a) and (b) were reproduced at the same depth in the hologram plane. (a) is the result of reproduction at the target wavelength of 638 nm, and (b) is the result of reproduction at 520 nm. If you look at a partial enlarged image, you can see that (a) is in focus and clear, while (b) is out of focus, so the lines are not clear.

Figure 3 is a diagram illustrating an example of a method for generating a hologram pattern according to an embodiment of the present invention.

Referring to FIG. 3, the reproduction wavelength of the hologram pattern is converted using the artificial neural network 300. The artificial neural network 300 can be implemented with various conventional deep neural networks such as CNN (Convolutional Neural Network). An example of the artificial neural network 300 for this embodiment is shown in FIG. 4.

The artificial neural network 300 requires a learning process using training data. The artificial neural network 300 can be trained using a supervised learning method using training data including a dataset of the input hologram pattern of the first wavelength and the correct hologram pattern of the second wavelength. When the trained artificial neural network 300 receives the hologram pattern 310 of the first wavelength, it outputs the hologram pattern 320 of the second wavelength.

A holographic pattern can be divided into a real part and an imaginary part. Therefore, the data input to the artificial neural network 300 may be data obtained by concatenating the real part and the imaginary part of the hologram pattern in the channel direction. In the case of currently used holographic displays (i.e., spatial light modulators), complex modulation is not possible and must be encoded as a real number pattern. However, the artificial neural network 300 of this embodiment outputs a complex hologram, so encoding in the form desired by the user can be used.

Supervised learning-based deep learning technology is a form of learning inherent statistical distributions from multiple images, so it requires large datasets. In particular, since this embodiment targets a multi-depth hologram, a large amount of RGB images and corresponding depth map data, that is, 3D data, are required. The method of generating a large dataset for training the artificial neural network 300 is reviewed again in FIGS. 5 and 6.

Figure 4 is a diagram illustrating an example of an artificial neural network according to an embodiment of the present invention.

Referring to FIG. 4, the artificial neural network network 400 for converting the reproduction wavelength of the holographic pattern includes one convolutional layer, five residual blocks, and one output branch. may include. Each residual block 410 may include two convolutional layers, a batch normalization layer, and an activation function. The output branch 420 may be configured to include a convolutional layer. This embodiment is only an example to aid understanding, and the artificial neural network that converts the reproduction wavelength of the hologram pattern can be implemented in various forms and is not limited to this embodiment.

Figures 5 and 6 are diagrams showing an example of a method for generating training data for an artificial neural network according to an embodiment of the present invention.

Referring to FIGS. 5 and 6 together, a holographic pattern reproduction wavelength conversion device (hereinafter referred to as 'device') includes an intensity map 600 and a depth map (600) of randomly distributed points or shapes. Create a depth map (610) (S500). For example, the device generates an intensity-map 600 in which dots, circles, squares, diamonds, etc. are randomly placed in an image of a certain size, and a depth-map 610 corresponding to each intensity-map 600. creates . The method itself of generating the intensity-map 600 and the depth-map 610 can use various conventional methods. This embodiment is characterized by generating a three-dimensional image including an intensity-map 600 and a depth-map 610 in which various shapes such as dots and circles are randomly arranged.

The device generates training data including an input hologram pattern of the first wavelength and a correct answer hologram pattern of the second wavelength based on the intensity-map 600 and the depth-map 610 (S510). The device applies a numerical propagation method to the intensity-map 600 and depth-map 610 located on the target image plane to generate a hologram pattern (i.e., complex hologram pattern) in the spatial light modulator plane (i.e., hologram plane). can be created. At this time, the device generates hologram patterns for various wavelengths and uses them as training data. The device uses training data to train an artificial neural network using a supervised learning method (S520).

Figure 7 is a diagram showing the configuration of an example of a wavelength conversion device for reproducing a hologram pattern according to an embodiment of the present invention.

Referring to FIG. 7, the device 700 includes a training data generation unit 710, a learning unit 720, an artificial neural network 730, and a transformation unit 740. In one embodiment, the device 700 may be implemented as a computing device including a memory, a processor, and an input/output device. In this case, each configuration can be implemented as software, loaded into memory, and then performed by a processor.

The training data generator 710 generates a dataset for training the artificial neural network 730 (S710). For example, the training data generator 710 may generate training data using the method shown in FIGS. 5 and 6. In another embodiment, if training data is predefined, the training data generator 710 may be omitted.

The learning unit 720 trains the artificial neural network 730 using training data. The learning unit 720 can train the artificial neural network 730 using a supervised learning method.

The conversion unit 740 converts the hologram pattern of the first wavelength into a hologram pattern of the second wavelength using a trained artificial neural network. In other words, the conversion unit 740 inputs a hologram of the first wavelength to the artificial neural network 730, which performs reproducible wavelength conversion between the first wavelength and the second wavelength, and converts the hologram of the second wavelength from the artificial neural network 730. Obtain a pattern.

8 to 10 are diagrams showing experimental examples of a wavelength conversion method for reproducing a hologram pattern according to an embodiment of the present invention.

Referring to FIG. 8, the wavelengths of the light source used in the experiment were 638 nm, 532 nm, and 460 nm in the order of red, green, and blue, and the pixel size of the spatial light modulator (SLM) was 8 μm. The depth range distribution of the test data used in the experimental results assumed a total of five depth planes, from a plane 5 mm away from the spatial light modulator to a plane 15 mm away.

Referring to FIG. 9, the results of test inference on a test image using the artificial neural network of this embodiment are shown. Since the present invention targets not only 2D but also hologram patterns for 3D objects, test inference was conducted on 3D objects. The result (900) of reproducing a hologram pattern generated according to the conversion target wavelength of 638 nm using a conventional numerical propagation model (ASM), and a hologram with the reproduced wavelength converted to 638 nm using the artificial neural network of this embodiment. The result 910 of reproducing the pattern is shown. Enlarged pictures of the in-focus portion at each depth (902, 904, 906, 912, 914, 916) are shown. It can be seen that the result 910 of converting the reproduction wavelength of the hologram pattern using the artificial neural network of this embodiment and the result 900 of converting the reproduction wavelength using the conventional numerical propagation model are almost similar.

Referring to FIG. 10, the results of reproducing a hologram pattern generated using the artificial neural network of this embodiment through an optical experiment are shown. The result of optically reproducing a hologram pattern whose reproduction wavelength was converted through a numerical propagation model (ASM) (1000) and the result of optically reproducing a hologram pattern whose reproduction wavelength was converted through the artificial neural network of this embodiment (1010) are almost identical. You can see the similarity. Enlarged pictures of the in-focus portion at each depth (1002, 1004, 1006, 1008, 1012, 1014, 1016, 1018) are shown.

The present invention can also be implemented as computer-readable program code on a computer-readable recording medium. Computer-readable recording media include all types of recording devices that store data that can be read by a computer system. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage devices. Additionally, computer-readable recording media can be distributed across networked computer systems so that computer-readable code can be stored and executed in a distributed manner.

So far, the present invention has been examined focusing on its preferred embodiments. A person skilled in the art to which the present invention pertains will understand that the present invention may be implemented in a modified form without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments should be considered from an illustrative rather than a restrictive perspective. The scope of the present invention is indicated in the claims rather than the foregoing description, and all differences within the equivalent scope should be construed as being included in the present invention.

Claims (9)

In the reproduction wavelength conversion method performed by the reproduction wavelength conversion device of the holographic pattern,
Receiving a hologram pattern of a first wavelength;
Converting the hologram pattern of the first wavelength into a hologram pattern of the second wavelength through the artificial neural network learned to convert the reproduced wavelength; and
A method for converting wavelengths to reproduce a hologram pattern, comprising: outputting a hologram pattern of the second wavelength. The method of claim 1, wherein the converting step includes:
Generating a holographic pattern of the second wavelength by inputting data obtained by combining the real and imaginary parts of the holographic pattern of the first wavelength in the channel direction into the artificial neural network; converting the reproduction wavelength of the holographic pattern, comprising: method. According to clause 1,
Reproducing the holographic pattern further comprising training the artificial neural network using a supervised learning method using training data including a dataset of the input holographic pattern of the first wavelength and the correct holographic pattern of the second wavelength. Wavelength conversion method. According to clause 3,
Generating an input hologram pattern and a correct answer hologram pattern of the training data based on an intensity-map and a depth-map of randomly distributed points or shapes. A method for converting wavelengths of a holographic pattern to reproduce the hologram pattern, further comprising: An artificial neural network trained to convert reproduced wavelengths; and
A conversion unit that converts a holographic pattern of a first wavelength into a holographic pattern of a second wavelength through the artificial neural network. The method of claim 5, wherein the conversion unit,
A wavelength conversion device for reproducing a holographic pattern, characterized in that the holographic pattern of the second wavelength is generated by inputting data obtained by combining the real and imaginary parts of the holographic pattern of the first wavelength in the channel direction into the artificial neural network. According to clause 5,
A learning unit that trains the artificial neural network using a supervised learning method using training data including a dataset of the input hologram pattern of the first wavelength and the correct answer hologram pattern of the second wavelength; Reproduction wavelength converter. According to clause 7,
A training data generator that generates an input hologram pattern and a correct answer hologram pattern of the training data based on an intensity-map and a depth-map of randomly distributed points or shapes; a reproduction wavelength of the hologram pattern, characterized in that it further includes a. converter. A computer-readable recording medium recording a computer program for performing the method described in claim 1.

Images