CN111260562A - Digital holographic image reconstruction method based on deep learning - Google Patents
Digital holographic image reconstruction method based on deep learning Download PDFInfo
- Publication number
- CN111260562A CN111260562A CN202010147099.3A CN202010147099A CN111260562A CN 111260562 A CN111260562 A CN 111260562A CN 202010147099 A CN202010147099 A CN 202010147099A CN 111260562 A CN111260562 A CN 111260562A
- Authority
- CN
- China
- Prior art keywords
- digital holographic
- holographic image
- low
- neural network
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4046—Scaling the whole image or part thereof using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4053—Super resolution, i.e. output image resolution higher than sensor resolution
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Holo Graphy (AREA)
Abstract
The invention discloses a digital holographic image reconstruction method based on deep learning, which comprises the following steps: A. down-sampling the reference digital holographic image to obtain a low-resolution sample set; B. establishing a neural network model; C. grouping reference digital holographic images into a linear correlation group and a nonlinear correlation group with a low-resolution sample set, and scrambling two groups of data through the low-resolution sample set to obtain two groups of training data sets; D. inputting two groups of training data sets from two input layers respectively, and training the neural network model until the output layer reaches the preset precision; E. and inputting the digital holographic image to be processed and the low-resolution sample set thereof into a neural network model, and processing the digital holographic image to be processed by the neural network model to obtain a reconstructed digital holographic image. The invention can improve the defects of the prior art and improve the robustness of the reconstruction method.
Description
Technical Field
The invention relates to the technical field of digital holographic images, in particular to a digital holographic image reconstruction method based on deep learning.
Background
The digital holographic image is a holographic image recorded by using an optical sensor and generated in a computer, is convenient to store, is suitable for various post-treatments according to use requirements, and is widely applied to multiple fields of biomedicine, scientific research experiments and the like. In order to improve the definition of the digital holographic image, the digital holographic image is generally required to be reconstructed, and the existing reconstruction method is not high in robustness, has more requirements on the input image and limits the application of the digital holographic image.
Disclosure of Invention
The invention aims to provide a digital holographic image reconstruction method based on deep learning, which can overcome the defects of the prior art and improve the robustness of the reconstruction method.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows.
A digital holographic image reconstruction method based on deep learning comprises the following steps:
A. down-sampling the reference digital holographic image to obtain a low-resolution sample set;
B. establishing a neural network model, wherein the neural network model consists of two input layers, a convolutional layer, a feedback layer, an activation layer, a pooling layer and an output layer;
C. grouping reference digital holographic images into a linear correlation group and a nonlinear correlation group with a low-resolution sample set, and scrambling two groups of data through the low-resolution sample set to obtain two groups of training data sets;
D. inputting two groups of training data sets from two input layers respectively, and training the neural network model until the output layer reaches the preset precision;
E. and inputting the digital holographic image to be processed and the low-resolution sample set thereof into a neural network model, and processing the digital holographic image to be processed by the neural network model to obtain a reconstructed digital holographic image.
Preferably, in step B, the loss function L of the neural network model is,
where n is the number of training passes, F is the mapping of the low resolution image to the high resolution image, xiScrambling the processed data for sets linearly related to the low resolution sample set, yiScrambling the processed data by a group of nonlinear correlations with the low-resolution sample set, wherein theta is a weight;
the activation function H of the neural network model is,
wherein k is1And k2Is a constant, Y is a saturation value,
is xiIs determined by the average value of (a) of (b),
is yiS is the system input.
Preferably, the convolution kernel of the convolutional layer has a size of m × n, m: the value of n is equal to the resolution of the low resolution samples.
Preferably, in step C, the digital hologram image is subjected to low-pass filtering processing before being grouped.
Preferably, in step D, after each training, the weight θ is adjusted, the single adjustment amount of the weight θ is not greater than a first set threshold, the total adjustment amount of the weight θ is not greater than a second set threshold, and the second set threshold is smaller than the first set threshold.
Adopt the beneficial effect that above-mentioned technical scheme brought to lie in: the invention adopts a deep learning mode to reconstruct the digital holographic image. In order to improve the robustness of the reconstruction process, the invention adopts a double-input neural network model, and effectively avoids the problem that the traditional neural network model depends too much on training convergence by optimizing the structure and parameters of the neural network model, thereby improving the robustness of the input image.
Drawings
FIG. 1 is a schematic diagram of one embodiment of the present invention.
Detailed Description
Referring to fig. 1, one embodiment of the present invention includes the steps of:
A. down-sampling the reference digital holographic image to obtain a low-resolution sample set;
B. establishing a neural network model, wherein the neural network model consists of two input layers, a convolutional layer, a feedback layer, an activation layer, a pooling layer and an output layer;
C. grouping reference digital holographic images into a linear correlation group and a nonlinear correlation group with a low-resolution sample set, and scrambling two groups of data through the low-resolution sample set to obtain two groups of training data sets;
D. inputting two groups of training data sets from two input layers respectively, and training the neural network model until the output layer reaches the preset precision;
E. and inputting the digital holographic image to be processed and the low-resolution sample set thereof into a neural network model, and processing the digital holographic image to be processed by the neural network model to obtain a reconstructed digital holographic image.
In step B, the loss function L of the neural network model is,
where n is the number of training passes, F is the mapping of the low resolution image to the high resolution image, xiScrambling the processed data for sets linearly related to the low resolution sample set, yiScrambling the processed data by a group of nonlinear correlations with the low-resolution sample set, wherein theta is a weight;
the activation function H of the neural network model is,
wherein k is1And k2Is a constant, Y is a saturation value,
is xiIs determined by the average value of (a) of (b),
is yiS is the system input.
The convolution kernel of the convolution layer has a size of m × n, m: the value of n is equal to the resolution of the low resolution samples.
And C, performing low-pass filtering processing on the digital holographic images before grouping the digital holographic images.
In step D, after each training, the weight θ is adjusted, the single adjustment amount of the weight θ is not greater than the first set threshold, the total adjustment amount of the weight θ is not greater than the second set threshold, and the second set threshold is smaller than the first set threshold.
The foregoing shows and describes the general principles and broad features of the present invention and advantages thereof. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (5)
1. A digital holographic image reconstruction method based on deep learning is characterized by comprising the following steps:
A. down-sampling the reference digital holographic image to obtain a low-resolution sample set;
B. establishing a neural network model, wherein the neural network model consists of two input layers, a convolutional layer, a feedback layer, an activation layer, a pooling layer and an output layer;
C. grouping reference digital holographic images into a linear correlation group and a nonlinear correlation group with a low-resolution sample set, and scrambling two groups of data through the low-resolution sample set to obtain two groups of training data sets;
D. inputting two groups of training data sets from two input layers respectively, and training the neural network model until the output layer reaches the preset precision;
E. and inputting the digital holographic image to be processed and the low-resolution sample set thereof into a neural network model, and processing the digital holographic image to be processed by the neural network model to obtain a reconstructed digital holographic image.
2. The deep learning-based digital holographic image reconstruction method according to claim 1, wherein: in step B, the loss function L of the neural network model is,
where n is the number of training passes, F is the mapping of the low resolution image to the high resolution image, xiScrambling the processed data for sets linearly related to the low resolution sample set, yiScrambling the processed data by a group of nonlinear correlations with the low-resolution sample set, wherein theta is a weight;
the activation function H of the neural network model is,
wherein k is1And k2Is a constant, Y is a saturation value,
is xiIs determined by the average value of (a) of (b),
is yiS is the system input.
3. The deep learning-based digital holographic image reconstruction method according to claim 2, wherein: the convolution kernel of the convolution layer has a size of m × n, m: the value of n is equal to the resolution of the low resolution samples.
4. The deep learning-based digital holographic image reconstruction method according to claim 3, wherein: and C, performing low-pass filtering processing on the digital holographic images before grouping the digital holographic images.
5. The deep learning-based digital holographic image reconstruction method according to claim 4, wherein: in step D, after each training, the weight θ is adjusted, the single adjustment amount of the weight θ is not greater than the first set threshold, the total adjustment amount of the weight θ is not greater than the second set threshold, and the second set threshold is smaller than the first set threshold.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010147099.3A CN111260562A (en) | 2020-03-05 | 2020-03-05 | Digital holographic image reconstruction method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010147099.3A CN111260562A (en) | 2020-03-05 | 2020-03-05 | Digital holographic image reconstruction method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111260562A true CN111260562A (en) | 2020-06-09 |
Family
ID=70952940
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010147099.3A Withdrawn CN111260562A (en) | 2020-03-05 | 2020-03-05 | Digital holographic image reconstruction method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111260562A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114002931A (en) * | 2021-10-08 | 2022-02-01 | 清华大学深圳国际研究生院 | Large-view-field holographic projection method and system based on deep learning accelerated calculation |
| CN116432244A (en) * | 2023-06-15 | 2023-07-14 | 杭州海康威视数字技术股份有限公司 | Image processing method, device, equipment and system |
-
2020
- 2020-03-05 CN CN202010147099.3A patent/CN111260562A/en not_active Withdrawn
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114002931A (en) * | 2021-10-08 | 2022-02-01 | 清华大学深圳国际研究生院 | Large-view-field holographic projection method and system based on deep learning accelerated calculation |
| CN116432244A (en) * | 2023-06-15 | 2023-07-14 | 杭州海康威视数字技术股份有限公司 | Image processing method, device, equipment and system |
| CN116432244B (en) * | 2023-06-15 | 2023-09-19 | 杭州海康威视数字技术股份有限公司 | Image processing method, device, equipment and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108765296B (en) | Image super-resolution reconstruction method based on recursive residual attention network | |
| CN108596841B (en) | Method for realizing image super-resolution and deblurring in parallel | |
| CN109118432B (en) | Image super-resolution reconstruction method based on rapid cyclic convolution network | |
| CN110223304B (en) | Image segmentation method and device based on multipath aggregation and computer-readable storage medium | |
| CN114140353A (en) | Swin-Transformer image denoising method and system based on channel attention | |
| CN109389556A (en) | The multiple dimensioned empty convolutional neural networks ultra-resolution ratio reconstructing method of one kind and device | |
| Luo et al. | Lattice network for lightweight image restoration | |
| CN111260562A (en) | Digital holographic image reconstruction method based on deep learning | |
| CN111667444B (en) | Image compressed sensing reconstruction method based on multi-channel residual error network | |
| CN110706154A (en) | Image super-resolution method based on hierarchical residual error neural network | |
| CN111932461A (en) | Convolutional neural network-based self-learning image super-resolution reconstruction method and system | |
| CN113902658B (en) | RGB image-to-hyperspectral image reconstruction method based on dense multiscale network | |
| CN110796251A (en) | Image compression optimization method based on convolutional neural network | |
| CN111402138A (en) | Image super-resolution reconstruction method of supervised convolutional neural network based on multi-scale feature extraction fusion | |
| CN112215755A (en) | Image super-resolution reconstruction method based on back projection attention network | |
| CN110533591B (en) | Super-resolution image reconstruction method based on codec structure | |
| CN113837959B (en) | Image denoising model training method, image denoising method and system | |
| CN110738660A (en) | Spine CT image segmentation method and device based on improved U-net | |
| CN111986092B (en) | Dual-network-based image super-resolution reconstruction method and system | |
| CN114821100A (en) | Image compressed sensing reconstruction method based on structural group sparse network | |
| CN111563843B (en) | Image super-resolution reconstruction method, system and related device | |
| CN116797541A (en) | Transformer-based lung CT image super-resolution reconstruction method | |
| CN114529482B (en) | Image compressed sensing reconstruction method based on wavelet multichannel depth network | |
| CN113793267B (en) | Self-supervision single remote sensing image super-resolution method based on cross-dimension attention mechanism | |
| CN111915492B (en) | Multi-branch video super-resolution method and system based on dynamic reconstruction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200609 |
|
| WW01 | Invention patent application withdrawn after publication |