CN110008585B - Scattered wavefront light field estimation acceleration method based on momentum gradient - Google Patents
Scattered wavefront light field estimation acceleration method based on momentum gradient Download PDFInfo
- Publication number
- CN110008585B CN110008585B CN201910263248.XA CN201910263248A CN110008585B CN 110008585 B CN110008585 B CN 110008585B CN 201910263248 A CN201910263248 A CN 201910263248A CN 110008585 B CN110008585 B CN 110008585B
- Authority
- CN
- China
- Prior art keywords
- depth
- light field
- scattered
- gradient
- estimation
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a scattering wavefront light field estimation acceleration method based on momentum gradient, which comprises the following steps: at depth z in a scattered light field imaging system 1 、z 2 The lower collected light intensity is Y 1 | 2 、|Y 2 | 2 Two light intensity images of (a); the arithmetic square root of the light intensity is taken as the depth z 1 、z 2 The initial value of the scattered wave front light field is estimated, and the corresponding initial value of the momentum gradient is set as a zero matrix with the same dimension as the two light intensities; estimating the scattered wave front light field of two depths by using a scattered wave front light field estimation iterative algorithm based on light field amplitude constraint; and updating momentum gradient by using the initial gradient obtained by the propagation and amplitude replacement of the light field in the iterative process, and updating the scattered wavefront light field estimated value by using the momentum gradient until iteration convergence to obtain two-depth scattered wavefront light field estimated value Y' 1 And Y' 2 (ii) a Y 'using a quadratic propagation algorithm of the light field' 1 And Y' 2 And transmitting to the target depth, and averaging the obtained two transmitted scattered wave front light fields in a complex field to obtain a scattered wave front light field estimated value of the target depth.
Description
Technical Field
The invention relates to a scattered wavefront light field estimation acceleration method based on a momentum gradient, and belongs to the field of computational photography.
Background
Scattering media widely present in nature are capable of changing the direction of propagation of incident light, whereas human and camera vision systems rely on the straight-line propagation of light for imaging. Therefore, imaging objects shielded by scattering media is a promising technique. However, this task is still faced with a number of problems to be solved when the scattering medium has a scattering strength and a thickness such that it scatters substantially all the light passing through it.
Current scatter imaging schemes through strongly scattering layers mainly include the following four: the method comprises a wave front modulation method, a phase recovery method based on speckle scanning, a phase recovery method based on single acquisition and a system point spread function deconvolution method. The wavefront modulation method uses a spatial light modulator to compensate phase disturbance caused by a scattering layer to emergent light, so that one point of an object plane in the system corresponds to a single point of a phase plane, and direct imaging can be realized like a traditional optical imaging system. The phase recovery method based on speckle scanning and the phase recovery method based on single acquisition can recover clear images of simple samples in a non-invasive manner. The three methods described above can only reconstruct the imaging target at a single depth in one imaging process, the imaging depth of field is limited to a small range around the acquisition depth, and the range of the imaging depth of field is proportional to the square of the object distance, so the range of the depth of field in the microscopic imaging system with a small object distance is further reduced. The system point spread function deconvolution method can realize multi-depth reconstruction by acquiring multi-depth system point spread functions, a speckle estimation algorithm based on a training sample can also estimate the depth of the training sample in a non-invasive manner to obtain the speckle distribution of the depth of the training sample for imaging target reconstruction, but the multi-depth acquisition cannot be realized under many conditions, and the calculated amount of the multi-depth speckle estimation and the requirement on the training sample distribution are obviously increased along with the increase of the depth number. The depth of field expanding method based on speckle scaling can expand the depth of field of a scattering perspective imaging system under the condition of collecting or estimating a point spread function of a single-depth system. The light field estimation method based on the double-depth light field amplitude constraint and the light field propagation can improve the effect of field depth expansion under the condition of collecting or estimating the point spread function of the double-depth system, but the time consumption of the light field propagation iteration process of the method is long, so that the application prospect of the method is limited.
The above background disclosure is only for the purpose of assisting understanding of the inventive concept and technical solutions of the present invention, and does not necessarily belong to the prior art of the present patent application, and should not be used for evaluating the novelty and inventive step of the present application in the case that there is no clear evidence that the above content is disclosed before the filing date of the present patent application.
Disclosure of Invention
The invention mainly aims to overcome the defects of the prior art and provides a scattering wavefront light field estimation acceleration method based on momentum gradient, which is used for improving the iterative optimization speed of the light field estimation method based on double-depth light field amplitude constraint and light field propagation.
The invention provides the following technical scheme for achieving the purpose:
a scattering wavefront light field estimation acceleration method based on momentum gradient comprises the following steps:
a1, in a scattered light field imaging system, at a depth z 1 The lower collected light intensity is Y 1 | 2 At a depth z 2 Lower collection light intensity | Y 2 | 2 Another light intensity image of (a);
a2, with light intensity | Y 1 | 2 、|Y 2 | 2 Arithmetic square root of (1) | Y 1 I and Y 2 L as depth z respectively 1 、z 2 And setting the depth z 1 、z 2 The initial values of the corresponding momentum gradients are respectively equal to two light intensities Y 1 | 2 、|Y 2 | 2 Zero value matrices having the same dimensions;
a3, estimating the scattered wave front light field with two depths by using a scattered wave front light field estimation iterative algorithm based on light field amplitude constraint; in the iteration process, updating momentum gradient by using the initial gradient obtained by the propagation and amplitude replacement of the light field, and updating the estimated value of the scattered wavefront light field by using the momentum gradient until iteration convergence; when iterative convergence is carried out, scattering wavefront light field estimated values of two depths are obtained and are respectively recorded as Y 1 ' and Y 2 '；
A4, using a secondary propagation algorithm of the light field to enable the scattered wave front light field to be at two depths z 1 、z 2 Estimate of (d) Y 1 ' and Y 2 The method comprises the steps of' transmitting to a target depth to correspondingly obtain two transmitted and scattered wavefront light fields;
and A5, averaging the two transmitted scattered wavefront light fields in a complex field to obtain a scattered wavefront light field estimation value of the target depth.
According to the scattering wavefront light field estimation acceleration method based on the momentum gradient, provided by the technical scheme of the invention, the momentum gradient is introduced in the iteration process, and the value of the momentum gradient is updated by the initial gradient generated in each iteration process, so that the scattering wavefront light field estimation is continuously optimized by using the momentum gradient as a new iteration optimization direction, the convergence speed of the algorithm is finally improved, and the time consumption of the algorithm is reduced.
Drawings
FIG. 1 is a flow chart of a method for estimating acceleration of a scattered wave front light field based on momentum gradient provided by the present invention;
FIG. 2 is a schematic diagram of an iterative process of an iterative algorithm for scattered wave front light field estimation based on light field amplitude constraints.
Detailed Description
The invention is further described with reference to the following figures and detailed description of embodiments.
The specific embodiment of the invention provides a scattering wavefront light field estimation acceleration method based on momentum gradient, which can greatly improve the iteration speed of a scattering wavefront light field estimation algorithm, and referring to fig. 1, the method comprises the following steps of A1-A5:
step A1, in the scattered light field imaging system, at the depth z 1 The lower collection light intensity is Y 1 | 2 At a depth z 2 Lower collection light intensity | Y 2 | 2 Another intensity image of (a).
Step A2, using light intensity | Y 1 | 2 、|Y 2 | 2 Arithmetic square root of (1) | Y 1 I and Y 2 L as depth z respectively 1 、z 2 And setting the depth z 1 、z 2 The initial values of the corresponding momentum gradients are respectively equal to two light intensities Y 1 | 2 、|Y 2 | 2 Zero value matrices having the same dimensions;
a3, estimating the scattered wave front light field with two depths by using a scattered wave front light field estimation iterative algorithm based on light field amplitude constraint; in the iteration process, the momentum gradient is updated by using the initial gradient obtained by the propagation and amplitude replacement of the light field, and the estimated value of the scattered wave front light field is updated by using the momentum gradient until the iterationConverging; when iterative convergence is carried out, scattering wavefront light field estimated values of two depths are obtained and are respectively marked as Y 1 ' and Y 2 '；
Step A4, using a secondary propagation algorithm of the light field to enable the scattered wave front light field to be at two depths z 1 、z 2 Estimated value Y of 1 ' and Y 2 The method comprises the steps of' transmitting to a target depth to correspondingly obtain two transmitted and scattered wavefront light fields;
and A5, averaging the two transmitted scattered wavefront light fields in a complex field to obtain a scattered wavefront light field estimation value of the target depth.
In particular, a light intensity sensor is adopted to collect a distance scattering layer z in a scattered light field imaging system 1 、z 2 Two intensity images of light intensity Y 1 | 2 And | Y 2 | 2 (ii) a Wherein, z is set 2 ＞z 1 . The two acquired intensity images have the size (alpha, beta) and each pixel has the size (delta l) x ,δl y ). By imaging system aperture size and z 2 Nyquist Interval (δ l) of computing System n ,δl n ). If δ l x ＞δl n Or δ l y ＞δl n Then, two light intensity images are resampled to (alpha delta l) by using a bilinear interpolation method x /δl n ,βδl y /δl n ) The image size of (2). Set depth z 1 And z 2 The corresponding scattered wave front light field estimation initial value Y 1 0 、
Is the arithmetic square root of the intensity after resampling, i.e. Y 1 0 ＝|Y 1 |，
Setting the corresponding initial value of momentum gradient
Is equal to two light intensities | Y 1 | 2 、|Y 2 | 2 Zeros of the same dimensionA matrix of values; set light intensity | Y 1 | 2 、|Y 2 | 2 Is the arithmetic square root of (c) | Y 1 I and Y 2 And | serving as the amplitude constraint of the scattered wave front light field estimation iterative algorithm based on the light field amplitude constraint. Thus, the initialization of the scattered wavefront light field estimation iterative algorithm is completed.
And then, starting iteration of the algorithm, updating the momentum gradient by using the initial gradient obtained by the propagation and amplitude replacement of the light field, and updating the estimated value of the scattered wavefront light field by using the momentum gradient until iteration is converged. When the iteration converges, the depth z is obtained 1 、z 2 The corresponding scattered wavefront light field estimate. Referring to fig. 2, a specific iterative process includes:
first, the depth z is measured 1 Scattered wavefront light field estimation Y obtained from the (k-1) th iteration 1 k-1 Forward propagation to depth z 2 Then obtain
To pair
Is carried out with | Y 2 I is amplitude replacement of amplitude constraint to obtain amplitude replacement result
Wherein, the arg () is a function for acquiring radian included by the input signal in a complex field and a real axis; j denotes an imaginary unit.
Then, the above-obtained
Calculating the depth z 2 Initial gradient of kth iteration
Reuse of the above-obtained
Updating the depth z 2 Momentum gradient of k-th iteration
Wherein gamma is the attenuation coefficient of the momentum gradient, the smaller gamma is, the faster gamma is, and the value range is 0.2-0.4; eta is the update coefficient of the current initial gradient to the momentum gradient, and the value range is between 1-gamma and 1.
Then using the updated momentum gradient
Updating the depth z 2 To obtain the depth z by scattered wavefront light field estimation 2 Scattered wavefront light field estimation for the kth iteration
Then, the depth z is measured 2 Scattered wavefront light field estimation from the kth iteration
Back propagation to depth z 1 Then obtain
To pair
Is carried out with | Y 1 I is amplitude replacement of amplitude constraint to obtain amplitude replacement result
Based on the foregoing
To obtain finally
The depth z is obtained by the following equations (6) to (8) 1 Scattered wavefront light field estimate Y for the kth iteration 1 k ：
With continued reference to FIG. 2, from Y 1 k-1 The forward propagation starts and Y is finally obtained 1 k Y obtained for one iteration 1 k And performing forward propagation as an initial value of the next iteration. Repeating the iteration by analogy, and if convergence is caused during the Nth iteration, obtaining the scattering wavefront light field estimated value Y during the Nth iteration 1 N And
i.e. depth z 1 、z 2 The corresponding scattered wave front light field estimated values are respectively marked as Y 1 ' and Y 2 '。
To obtain an estimate of the scattered wavefront light field at any other depth, two acquisition depths z 1 、z 2 The estimation results of the scattered wave front light field are respectively propagated to any target depth for the second time, and then the result is averaged in a complex field to be used as the estimation result of the scattered wave front light field of the target depth. In particular, a secondary propagation algorithm using a light field scatters the wavefront light field at two depths z 1 、z 2 Estimated value Y of 1 ' and Y 2 ' propagation to a target depth z 3 Correspondingly obtaining two propagation scattered wavefront light fields Y 1-3 、Y 2-3 Then to Y 1-3 、Y 2-3 Averaging in a complex field to obtain the target depth z 3 Scattered wave front light field estimation value Y 3 ＝(Y 1-3 +Y 2-3 )/2。
The scattered wave front light field estimation result can be applied to a deconvolution scattering perspective imaging process to improve the imaging depth of field of the system. The scattering wavefront light field estimation acceleration method based on the momentum gradient can greatly improve the iteration speed of a scattering wavefront light field estimation algorithm.
The foregoing is a further detailed description of the invention in connection with specific preferred embodiments and it is not intended to limit the invention to the specific embodiments described. For those skilled in the art to which the invention pertains, several equivalent substitutions or obvious modifications can be made without departing from the spirit of the invention, and all the properties or uses are considered to be within the scope of the invention.
Claims (6)
1. A scattering wavefront light field estimation acceleration method based on momentum gradient is characterized by comprising the following steps:
a1, in a scattered light field imaging system, at a depth z 1 The lower collected light intensity is Y 1 | 2 Is measured by the light intensity image of (a),at a depth z 2 Lower collection light intensity | Y 2 | 2 Another light intensity image of (a);
a2, with light intensity | Y 1 | 2 、|Y 2 | 2 Is the arithmetic square root of (c) | Y 1 I and Y 2 L as depth z respectively 1 、z 2 And setting the depth z 1 、z 2 The initial values of the corresponding momentum gradients are respectively equal to two light intensities Y 1 | 2 、|Y 2 | 2 Zero value matrices having the same dimensions;
a3, estimating the scattered wave front light field with two depths by using a scattered wave front light field estimation iterative algorithm based on light field amplitude constraint; in the iteration process, updating momentum gradient by using the initial gradient obtained by the propagation and amplitude replacement of the light field, and updating the estimated value of the scattered wavefront light field by using the momentum gradient until iteration convergence; when iterative convergence is carried out, scattering wavefront light field estimated values of two depths are obtained and are respectively recorded as Y 1 'and Y' 2 ；
Setting z 2 ＞z 1 Then, the kth iteration in step A3 includes:
a31, depth z 1 Scattered wavefront light field estimation Y obtained by the (k-1) th iteration 1 k-1 Forward propagation to depth z 2 Then proceed with | Y 2 I is amplitude replacement of amplitude constraint to obtain amplitude replacement result
A32, utilization
Calculating the depth z 2 Initial gradient of kth iteration
A33, utilization
Updating the depth z 2 Momentum gradient of k-th iteration
A34, utilization
Updating the depth z 2 To obtain the depth z by scattered wavefront light field estimation 2 Scattered wavefront light field estimation for the kth iteration
A35, depth z 2 Scattered wavefront light field estimation obtained from the kth iteration
Counter-propagating to depth z 1 Then proceed with | Y 1 I is amplitude replacement of amplitude constraint to obtain amplitude replacement result
A36, utilization
Calculating the depth z 1 Initial gradient of kth iteration
A37, use of
Updating the depth z 1 Momentum gradient of k-th iteration
A38, utilization
Updating the depth z 1 To obtain the depth z by scattered wavefront light field estimation 1 Scattered wavefront light field estimate Y for the kth iteration 1 k Completing the kth iteration;
a4, scattering wave front light field at two depths z by using secondary propagation of the light field 1 、z 2 Estimated value Y of 1 'and Y' 2 Transmitting to a target depth to correspondingly obtain two transmitted and scattered wavefront light fields;
and A5, averaging the two transmitted scattered wavefront light fields in a complex field to obtain a scattered wavefront light field estimation value of the target depth.
2. The method of claim 1, wherein the depth z is a depth z for acceleration estimation of the scattered wavefront optical field based on momentum gradients 1 、z 2 Initial gradient of kth iteration
And
is defined as:
3. the method of claim 1 wherein the method of acceleration estimation is based on a scattered wavefront light field with momentum gradients
And
updating the momentum gradient of the kth iteration
And
the method comprises the following steps:
4. The method of claim 1, wherein the method further comprises,
depth z in A31 1 Scattered wavefront light field estimation Y obtained from the (k-1) th iteration 1 k-1 Forward propagation to depth z 2 Then obtain
To pair
Is carried out with | Y 2 Is obtained after amplitude substitution with the I being amplitude constraint
Depth z in A35 2 Scattered wavefront light field estimation obtained from the kth iteration
Counter-propagating to depth z 1 Then obtain
For is to
Is carried out with | Y 1 Is obtained after amplitude substitution with the I being amplitude constraint
Wherein, the arg () is a function for acquiring radian included by the input signal in a complex field and a real axis; j denotes an imaginary unit.
5. The method of claim 1, wherein the method further comprises,
utilization in A34
Updating the depth z 2 To obtain the depth z by scattered wavefront light field estimation 2 Scattered wavefront light field estimation for the kth iteration
The method comprises the following steps:
utilization in A38
Updating the depth z 1 To obtain the depth z by scattered wavefront light field estimation 1 Scattered wavefront light field estimate Y for the kth iteration 1 k The method comprises the following steps:
6. the method for estimating and accelerating scattered wavefront optical field based on momentum gradient as claimed in claim 1, wherein in step A1, the distance scattering layer z in the scattered optical field imaging system is collected by using the light intensity sensor 1 、z 2 Two intensity images.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910263248.XA CN110008585B (en) | 2019-04-02 | 2019-04-02 | Scattered wavefront light field estimation acceleration method based on momentum gradient |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910263248.XA CN110008585B (en) | 2019-04-02 | 2019-04-02 | Scattered wavefront light field estimation acceleration method based on momentum gradient |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110008585A CN110008585A (en) | 2019-07-12 |
| CN110008585B true CN110008585B (en) | 2022-12-23 |
Family
ID=67169663
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910263248.XA Active CN110008585B (en) | 2019-04-02 | 2019-04-02 | Scattered wavefront light field estimation acceleration method based on momentum gradient |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110008585B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116506714B (en) * | 2023-06-26 | 2023-09-26 | 之江实验室 | Imaging method, device, system, storage medium and electronic equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107247332A (en) * | 2017-08-04 | 2017-10-13 | 清华大学深圳研究生院 | It is a kind of to estimate that the non-intrusion type with deconvolution scatters imaging method based on speckle |
| CN107424195A (en) * | 2017-07-20 | 2017-12-01 | 清华大学深圳研究生院 | A kind of light field method for estimating distance |
| CN108364309A (en) * | 2018-02-09 | 2018-08-03 | 清华大学深圳研究生院 | A kind of spatial light field restoration methods based on hand-held light-field camera |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020160470A1 (en) * | 2000-11-13 | 2002-10-31 | Genoptix | Methods and apparatus for generating and utilizing linear moving optical gradients |
-
2019
- 2019-04-02 CN CN201910263248.XA patent/CN110008585B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107424195A (en) * | 2017-07-20 | 2017-12-01 | 清华大学深圳研究生院 | A kind of light field method for estimating distance |
| CN107247332A (en) * | 2017-08-04 | 2017-10-13 | 清华大学深圳研究生院 | It is a kind of to estimate that the non-intrusion type with deconvolution scatters imaging method based on speckle |
| CN108364309A (en) * | 2018-02-09 | 2018-08-03 | 清华大学深圳研究生院 | A kind of spatial light field restoration methods based on hand-held light-field camera |
Non-Patent Citations (2)
| Title |
|---|
| Underwater optical communications using orbital angular momentum-based spatial division multiplexing;Alan E. Willner 等;《Optics Communications》;20171231;全文 * |
| 基于点衍射干涉仪的分布式随机并行梯度下降算法;付强等;《红外与激光工程》;20111225(第12期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110008585A (en) | 2019-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| He et al. | A nonlinear least square technique for simultaneous image registration and super-resolution | |
| US7930152B2 (en) | Method for signal and image processing with lattice gas processes | |
| CN105931196A (en) | Fourier optical modeling-based coded aperture camera image restoration method | |
| CN103761712B (en) | The Image Blind convolution method of rebuilding based on ADAPTIVE OPTICS SYSTEMS point spread function | |
| CN110796616B (en) | Turbulence degradation image recovery method based on norm constraint and self-adaptive weighted gradient | |
| CN109003234B (en) | For the fuzzy core calculation method of motion blur image restoration | |
| CN114418883B (en) | Blind image deblurring method based on depth priori | |
| CN108416750A (en) | A kind of image recovery method | |
| Abubakar et al. | A multiplicative regularization approach for deblurring problems | |
| CN103310472A (en) | Limited angle photoacoustic imaging reconstruction method and device on basis of regularization iteration | |
| CN110008585B (en) | Scattered wavefront light field estimation acceleration method based on momentum gradient | |
| CN112634380B (en) | Multi-target color scattering imaging method of single-frame super-optical memory effect | |
| CN112465725B (en) | Infrared image frame rate up-conversion method based on PWC-Net | |
| CN112991504A (en) | Improved method for filling holes based on TOF camera three-dimensional reconstruction | |
| Fedorczak et al. | Tomographic reconstruction of tokamak plasma light emission from single image using wavelet-vaguelette decomposition | |
| CN100437640C (en) | Blind image restoring method based on adaptive judgment feedback balance technique | |
| CN108230251A (en) | Combined type image recovery method and device | |
| Lim et al. | Visual hull based 3D reconstruction of shocks in under-expanded supersonic bevelled jets | |
| WO2023279249A1 (en) | Non-line-of-sight imaging via neural transient field | |
| Zhang et al. | Expanding the memory effect in scattering imaging by manipulating photon distributions | |
| Ding et al. | Curvature regularization for non-line-of-sight imaging from under-sampled data | |
| CN108550120B (en) | Method for restoring underwater image under variable frame | |
| Chen et al. | Super-resolution reconstruction for underwater imaging | |
| Xie et al. | Joint reconstruction and calibration using regularization by denoising | |
| Huang et al. | Restoration of Missing Data in Limited Angle Tomography Based on Consistency Conditions |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |