WO2022120899A1 - 图像重建方法和装置、电子设备以及机器可读存储介质 - Google Patents

图像重建方法和装置、电子设备以及机器可读存储介质 Download PDF

Info

Publication number
WO2022120899A1
WO2022120899A1 PCT/CN2020/136550 CN2020136550W WO2022120899A1 WO 2022120899 A1 WO2022120899 A1 WO 2022120899A1 CN 2020136550 W CN2020136550 W CN 2020136550W WO 2022120899 A1 WO2022120899 A1 WO 2022120899A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
image
image data
residual
target image
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.)
Ceased
Application number
PCT/CN2020/136550
Other languages
English (en)
French (fr)
Inventor
梁栋
朱庆永
崔卓须
柯子文
丘志浪
刘元元
刘新
郑海荣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Institute of Advanced Technology of CAS
Original Assignee
Shenzhen Institute of Advanced Technology of CAS
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shenzhen Institute of Advanced Technology of CAS filed Critical Shenzhen Institute of Advanced Technology of CAS
Publication of WO2022120899A1 publication Critical patent/WO2022120899A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T12/00Tomographic reconstruction from projections

Definitions

  • the present invention belongs to the technical field of image processing, and in particular, relates to an image reconstruction method, an image reconstruction apparatus, an electronic device and a machine-readable storage medium.
  • Magnetic resonance imaging as an important imaging diagnostic technique, has the main drawback of slow imaging speed, which limits its further development in advanced clinical applications such as multi-contrast imaging and dynamic cardiac cine imaging. Therefore, it has always been a difficult problem in the field of magnetic resonance to study how to shorten the imaging time while maintaining the high resolution of the image.
  • compressed sensing uses a much lower sampling amount than Nyquist under the condition that the signal is sparse and the sampling matrix is irrelevant to the sparse transform basis. data, a good restoration of the original image is obtained by a nonlinear reconstruction algorithm.
  • the standard compressed sensing imaging method is to solve an L1 norm regularization method, also known as lasso (LASSO) regression.
  • L1 norm regularization method also known as lasso (LASSO) regression.
  • This kind of algorithm has a good application for weak convexity and even some non-convexity problems.
  • an iterative thresholding algorithm based on Plug-and-Play Prior replaces the original threshold shrinkage with an image denoising filter (the filter prior is the Plug-and-Play prior), which is coupled Image denoising to a forward model-based image restoration framework.
  • the iterative thresholding algorithm based on plug-and-play priors shows superior imaging quality than traditional optimization algorithms.
  • the traditional iterative threshold algorithm based on plug-and-play prior is only an iterative threshold algorithm based on a single type of plug-and-play prior, and does not consider more types of plug-and-play prior, so it is impossible to obtain the
  • the reconstructed image is well-balanced in terms of artifact suppression and structure preservation, so that the image reconstruction performance cannot be improved.
  • the present invention provides an image reconstruction method and an image reconstruction device that combine more types of plug-and-play priors.
  • An image reconstruction method provided according to an aspect of an embodiment of the present invention includes: performing the following loop process until a loop end condition is satisfied:
  • the image data and data residuals of the target image obtained in the previous cycle are filtered to obtain the first image data in this cycle;
  • the obtained target image data and data residual are filtered to obtain the second image data in the current cycle;
  • the target in the current cycle is obtained according to the first image data and the second image data in the current cycle.
  • the image data of the image; the data residual in this cycle is obtained according to the under-sampling data of the K-space, the image data of the target image in the current cycle, and the target image data and data residual obtained in the previous cycle;
  • the image data and data residual of the target image obtained in this loop process serve as the image data and data residual of the target image in the next loop process.
  • the method is obtained according to the under-sampling data in K-space, the image data of the target image in the current cycle, and the target image data and data residual obtained in the previous cycle.
  • the data residual in this cycle includes: obtaining the first residual correction in this cycle according to the image data and data residual of the target image obtained in the previous cycle and through the divergence estimation of the target image structure filtering factor; according to the image data and data residual of the target image obtained in the previous cycle and through the divergence estimation of guided image structure filtering, the second residual correction factor in this cycle is obtained; according to the described in this cycle
  • the first residual error correction factor and the second residual error correction factor obtain the residual error correction factor in the current cycle; according to the undersampling data of the K-space, the image data of the target image in the current cycle process, and the residual error correction factor to obtain Data residuals during this cycle.
  • the image data and data residual of the target image obtained in the previous loop process are filtered based on the target image structure filter and according to the following formula 1 to obtain the present image the first image data during the loop,
  • the target image data obtained in the last cycle process and the data residual are filtered to obtain the second image data in the cycle process;
  • the following formula 3 is used to obtain the target image in the current cycle image data
  • represents the weight parameter
  • the following formula 4 is used to obtain the first residual error correction in the current cycle process factor
  • A represents the undersampling Fourier transform matrix
  • represents the metric constant of the degree of problem underdetermination
  • x t-1 represents the image data of the target image obtained in the previous cycle
  • z t-1 The data residual obtained in the previous cycle
  • A represents the undersampled Fourier transform matrix.
  • o t represents the residual correction factor in this cycle
  • represents the weight parameter
  • the current cycle is calculated by using the following formula 7 according to the under-sampling data of the K-space, the image data of the target image in the current cycle, and the residual correction factor data residuals in the process,
  • z t is the data residual in this cycle
  • b represents the under-sampled data in K space.
  • An image reconstruction apparatus provided according to another aspect of the embodiments of the present invention includes: a target image structure filter, a guide image structure filter, a target image data acquisition module, and a data residual acquisition module that operate in a loop until the loop end condition is satisfied;
  • the target image structure filter is used to filter the image data and data residuals of the target image obtained in the previous cycle to obtain the first image data in the current cycle;
  • the guide image structure filter is used to The target image data and data residual obtained in one cycle process are filtered to obtain the second image data in this cycle process;
  • the target image data acquisition module is used to obtain the target image data according to the first image data and the current cycle process.
  • the second image data obtains the image data of the target image in the current cycle;
  • the data residual acquisition module is used to obtain the image data of the target image according to the K-space undersampling data, the image data of the target image in the current cycle and the previous cycle process.
  • the target image data and data residual obtained in the process get the data residual in this cycle;
  • the image data and data residual of the target image obtained in this loop process serve as the image data and data residual of the target image in the next loop process.
  • an electronic device comprising: at least one processor, and a memory coupled to the at least one processor, the memory storing instructions, when the instructions are executed by the When executed by the at least one processor, the at least one processor is caused to execute the image reconstruction method as described above.
  • Still another aspect of an embodiment according to the present invention provides a machine-readable storage medium storing executable instructions that, when executed, cause the machine to perform the image reconstruction method as described above.
  • the present invention adopts a composite plug-and-play prior, which can help to obtain an image that is well balanced in terms of artifact suppression and structure protection, and can also improve image reconstruction performance.
  • FIG. 1 is a flowchart illustrating an image reconstruction method according to an embodiment of the present invention
  • FIG. 2 is a flowchart illustrating an exemplary method for obtaining data residuals in this cycle in the image reconstruction method according to an embodiment of the present invention
  • FIG. 3 is a block diagram illustrating an image reconstruction apparatus according to an embodiment of the present invention.
  • FIG. 4 is a block diagram illustrating an electronic device implementing an image reconstruction method according to an embodiment of the present invention.
  • the term "including” and variations thereof represent open-ended terms meaning “including but not limited to”.
  • the terms “based on”, “depending on” and the like mean “based at least in part on”, “based at least in part on”.
  • the terms “one embodiment” and “an embodiment” mean “at least one embodiment.”
  • the term “another embodiment” means “at least one other embodiment.”
  • the terms “first”, “second”, etc. may refer to different or the same objects. Other definitions, whether explicit or implicit, may be included below. The definition of a term is consistent throughout the specification unless the context clearly dictates otherwise.
  • the iterative threshold algorithm based on plug-and-play prior is only an iterative threshold algorithm based on a single type of plug-and-play prior, and does not consider more types of plug-and-play prior. In this case, it is impossible to obtain a reconstructed image that is well balanced in terms of artifact suppression and structure protection, so that the image reconstruction performance cannot be improved.
  • embodiments of the present invention provide a method for image reconstruction that combines more types of plug-and-play priors for image reconstruction.
  • Image reconstruction method and image reconstruction device are described in detail below.
  • the image reconstruction method can be performed by an electronic device, and the electronic device performs the following loop process until the loop end condition is satisfied: filtering the image data and data residual of the target image obtained in the previous loop process based on the target image structure filter In order to obtain the first image data in the current cycle process; the target image data and data residuals obtained in the previous cycle process are filtered based on the guided image structure filter to obtain the second image data in the current cycle process; The first image data and the second image data in the cycle process obtain the image data of the target image in the cycle process; according to the undersampling data of the K space, the image data of the target image in the cycle process and the previous cycle The target image data and data residual obtained in the loop process obtain the data residual in this loop process; wherein, when the loop end condition is not met, the image data and data residual of the target image obtained in this loop process The difference serves as the image data and data residual for the target image in the next cycle.
  • the composite plug-and-play prior of the target image structure filter and the guided image structure filter is used to filter the image, which can help to obtain good results in artifact suppression and structure protection. Balanced images, and can also improve image reconstruction performance.
  • magnetic resonance imaging is used as an example for description.
  • some basic concepts and process derivations applied to the image reconstruction method according to the embodiment of the present invention are described by taking magnetic resonance imaging as an example.
  • the image reconstruction task of accelerated magnetic resonance imaging can be considered as solving a LASSO regression problem with regular L 1 norm.
  • the cost function for solving the LASSO regression problem includes the data fitting term based on the assumption of noise Gaussian distribution and the coefficients of the image in the transform domain. L1 norm constraint.
  • PPP Plug and Play Prior
  • AMP Approximate Message Passing
  • Equation 1 the magnetic resonance imaging data acquisition process based on the K-space undersampling mechanism can be discretely expressed as Equation 1 below.
  • is a positive constant that balances data fitting and sparse regularity.
  • R( ) represents a sparse variation domain.
  • Equation 2 A series of iterative algorithms with lower computational cost are proposed to solve the convex optimization problem of Equation 2, among which the representative algorithm is the Iterative Soft-thresholding Algorithm (ISTA).
  • the specific algorithm can be expressed as the following formula Sub 3 and Equation 4.
  • the iterative soft thresholding algorithm S ⁇ (y) (
  • x t represents the t-th x estimation (that is, the image data of the target image reconstructed after the t-th iteration)
  • represents the threshold parameter.
  • the approximate message passing algorithm is based on the introduction of the Onsager correction term to Gaussianize the residual z t , thereby further improving the performance of the algorithm. Therefore, the approximate message passing algorithm can be expressed as Equation 5, Equation 6, and Equation 7 below.
  • S ⁇ ′ is the derivative of S ⁇
  • is the metric constant of the degree of underdetermination of the problem
  • ⁇ > represents the vector mean operation.
  • Equation 8 Equation 9
  • two different types of denoising priors will be coupled to construct a composite plug-and-play prior based approximate message passing algorithm with improved performance for image reconstruction.
  • the first type of plug-and-play prior selects an algorithm based on Block Matching and 3D Filtering (BM3D) as the self-structure filtering prior from the target image x, where the BM3D algorithm aims to The metric matching of the Euclidean distance is performed on the adjacent image blocks, and a three-dimensional matrix is constructed for all similar image blocks.
  • BM3D Block Matching and 3D Filtering
  • the second class of plug-and-play priors selects the Mutually Guided Image Filtering (muGIF) algorithm as the structural filtering prior for guided images x r from other modalities or parameters
  • muGIF aims to
  • the anatomical structure information shared between the target image and the guide image is obtained through the interactive measurement of the structure information of the guide image similar to the target image, which can effectively suppress the image artifacts and strengthen the main structural features of the target image.
  • the interactive guided filtering can fully introduce the structural information of the guided image, and on the other hand, it can avoid the detail filtering deviation caused by the different image content.
  • the image reconstruction method according to the embodiment of the present invention may be performed by an electronic device, and the electronic device may include a smart phone, a tablet computer, a personal computer, a cloud server, a server, and the like.
  • FIG. 1 is a flowchart illustrating an image reconstruction method according to an embodiment of the present invention.
  • step S110 the image data and data residual of the target image obtained in the previous cycle process are filtered based on the target image structure filter to obtain the first image data in the current cycle process.
  • the target image structure filter has the aforementioned block matching based 3D filtering algorithm, which can implement filtering a priori on the self-structure from the target image.
  • the image data and data residual of the target image obtained in the previous cycle process are filtered to obtain the first image data in the current cycle process.
  • step S120 filter processing is performed on the target image data and the data residual obtained in the previous cycle process based on the guided image structure filter to obtain the second image data in the current cycle process.
  • the guided image structure filter has the above-described interactive guided image filtering algorithm, which can implement structure filtering priors on guided images from other modalities or parameters.
  • the guide image is similar to the target image.
  • the target image data and the data residual obtained in the previous loop process are filtered based on the guided image structure filter and according to the following formula 12 to obtain the second image data in the current loop process.
  • step S130 image data of the target image in the current cycle is obtained according to the first image data and the second image data in the current cycle.
  • the following formula 13 is used to obtain the image data of the target image in the current cycle.
  • represents the weight parameter
  • step S140 the data residual in the current loop is obtained according to the under-sampling data in the K-space, the image data of the target image in the current loop, and the target image data and data residual obtained in the previous loop.
  • the under-sampled data in k-space may be, for example, data acquired by under-sampling by a magnetic resonance imaging device.
  • FIG. 2 is a flow chart illustrating an exemplary method for obtaining data residuals in the current cycle in the image reconstruction method according to an embodiment of the present invention.
  • step S141 the first residual correction factor in the current loop is obtained according to the image data and data residual of the target image obtained in the previous loop and through the divergence estimation of the structure filtering of the target image.
  • the first residual correction factor in the current loop process is obtained according to the image data of the target image and the data residual obtained in the previous loop process, and the following formula 14 is used.
  • represents the metric constant of the degree of problem underdetermination
  • step S142 the second residual correction factor in the current loop is obtained according to the image data and data residual of the target image obtained in the previous loop and through the divergence estimation of the guided image structure filtering.
  • the second residual correction factor in the current loop process is obtained according to the image data of the target image and the data residual obtained in the previous loop process, and the following formula 15 is used.
  • step S143 the residual correction factor in the current loop process is obtained according to the first residual error correction factor and the second residual error correction factor in the current loop process.
  • the following formula 16 is used to obtain the residual correction factor in the current cycle.
  • o t represents the residual correction factor in this cycle.
  • step S144 the data residual in the current cycle is obtained according to the under-sampling data in the K-space, the image data of the target image in the current cycle, and the residual correction factor.
  • the data residual in this cycle is obtained by calculating the following formula 17 according to the under-sampling data in K-space, the image data of the target image in the current cycle, and the residual correction factor.
  • z t represents the data residual in this cycle
  • b represents the undersampled data in K space.
  • step S150 it is determined whether the loop end condition is satisfied. If yes, end the reconstruction; if no, use the image data and data residuals of the target image obtained in this loop as the image data and data residuals of the target image in the next loop, and go to step S110.
  • the loop end condition can be specified.
  • the loop end condition may include reaching a predetermined number of loops (or iterations).
  • FIG. 3 is a block diagram illustrating an image reconstruction apparatus according to an embodiment of the present invention.
  • the image reconstruction apparatus 300 is applied in an electronic device to be executed by the electronic device.
  • the image reconstruction apparatus 300 includes: a target image structure filter 310 , a guide image structure filter 320 , a target image data acquisition module 330 and a data residual acquisition module 340 .
  • the target image structure filter 310 , the guide image structure filter 320 , the target image data acquisition module 330 and the data residual acquisition module 340 operate in a loop until the loop end condition is satisfied.
  • the loop end condition can be specified.
  • the loop end condition may include reaching a predetermined number of loops (or iterations).
  • the target image structure filter 310 is configured to perform filtering processing on the image data and data residuals of the target image obtained in the previous cycle to obtain the first image data in the current cycle.
  • the target image structure filter 310 may be configured to perform filtering processing on the target image data and data residual obtained in the previous loop process according to the above formula 11 to obtain the first image data in the current loop process .
  • the guiding image structure filter 320 is configured to perform filtering processing on the target image data and data residual obtained in the previous cycle to obtain the second image data in the current cycle.
  • the guiding image structure filter 320 may be configured to perform filtering processing on the target image data and data residual obtained in the previous loop process according to the above formula 12 to obtain the second image data in the current loop process .
  • the target image data acquisition module 330 is configured to obtain image data of the target image in the current cycle according to the first image data and the second image data in the current cycle.
  • the target image data acquisition module 330 may be configured to obtain the target in the current cycle by calculating the above formula 13 according to the first image data and the second image data in the current cycle Image data for the image.
  • the data residual acquisition module 340 is configured to obtain the data in the current loop process according to the under-sampled data in the K-space, the image data of the target image in the current loop process, and the target image data and data residual obtained in the previous loop process. residual.
  • the data residual acquisition module 340 may be configured to obtain the data residuals according to the under-sampled data of K-space, the image data of the target image in the current loop, and the target image data and data residuals obtained in the previous loop. Using the above formula 15, formula 16 and formula 17 to calculate the data residual in this cycle process.
  • the image reconstruction apparatus may be implemented by hardware, or may be implemented by software or a combination of hardware and software. Taking software implementation as an example, a device in a logical sense is formed by reading the corresponding computer program instructions in the memory into the memory for operation by the processor of the device where it is located. In the embodiment of the present invention, the apparatus for image reconstruction using, for example, an electronic device may be used to implement.
  • FIG. 4 is a block diagram illustrating an electronic device implementing an image reconstruction method according to an embodiment of the present invention.
  • an electronic device 400 may include at least one processor 410 , memory (eg, non-volatile memory) 420 , memory 430 , and communication interface 440 , and at least one processor 410 , memory 420 , memory 430 , and communication interface 440 Connected together via bus 450 .
  • At least one processor 410 executes at least one computer-readable instruction stored or encoded in memory (ie, the above-described elements implemented in software).
  • computer-executable instructions are stored in memory that, when executed, cause at least one processor 410 to perform the following loop process until a loop end condition is met: filter based on the target image structure filter on the target obtained in the previous loop process The image data and data residual of the image are filtered to obtain the first image data in the current cycle; the target image data and data residual obtained in the previous cycle are filtered based on the guided image structure filter to obtain the current image.
  • the second image data in the loop process obtain the image data of the target image in the loop process according to the first image data and the second image data in the loop process;
  • the image data of the target image in the process and the target image data and data residual obtained in the previous cycle process obtain the data residual in the current cycle process; wherein, when the cycle end condition is not met, the current cycle process
  • the image data and data residuals of the target image obtained in serve as the image data and data residuals of the target image in the next cycle.
  • a program product eg, a machine-readable medium
  • the machine-readable medium may have instructions (ie, the above-described elements implemented in software) that, when executed by a machine, cause the machine to perform the various operations described in connection with FIGS. 1-3 above in various embodiments of the present invention and function.
  • a system or an apparatus equipped with a readable storage medium may be provided, on which software program codes for realizing the functions of any of the above-described embodiments are stored, and a computer or a computer of the system or apparatus may be provided.
  • the processor reads and executes the instructions stored in the readable storage medium.
  • the program code itself read from the readable medium can implement the functions of any one of the above-described embodiments, and thus the machine-readable code and the readable storage medium storing the machine-readable code constitute the present invention part of the example.
  • Examples of readable storage media include floppy disks, hard disks, magneto-optical disks, optical disks (eg, CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD-RW), magnetic tape, non- Volatile memory cards and ROMs.
  • the program code may be downloaded from a server computer or the cloud over a communications network.
  • the device structure described in the above embodiments may be a physical structure or a logical structure, that is, some units may be implemented by the same physical entity, or some units may be implemented by multiple physical entities, or may be implemented by multiple physical entities. Some components in separate devices are implemented together.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)

Abstract

一种图像重建方法,其包括:执行以下循环过程,直至满足循环结束条件:基于目标图像结构滤波器对上一循环过程中的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据(S110);基于引导图像结构滤波器对上一循环过程中的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据(S120);根据第一图像数据和第二图像数据得到本循环过程中的目标图像的图像数据(S130);根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差得到本次循环过程中的数据残差(S140);在未满足循环结束条件时,本次循环过程中的目标图像的图像数据和数据残差充当下一循环过程中的图像数据和数据残差。

Description

图像重建方法和装置、电子设备以及机器可读存储介质 技术领域
本发明属于图像处理技术领域,具体地讲,涉及一种图像重建方法、图像重建装置、电子设备以及机器可读存储介质。
背景技术
磁共振成像作为一种重要的影像诊断技术,其成像速度缓慢的主要缺陷限制了包括在多对比度成像、动态心脏电影成像等高级临床应用领域的进一步发展。因此,研究如何在保持图像高分辨的情况下缩短成像时间,一直是磁共振领域的重难点问题。
压缩感知作为一种经典的基于信号欠采样机制的加速磁共振成像方法,在满足信号稀疏以及采样矩阵与稀疏变换基不相关的条件下,使用远低于奈奎斯特(Nyquist)采样量的数据,通过非线性重建算法得到原始图像的良好恢复。标准的压缩感知成像方法是求解一个L 1范数正则化的方法,也被称为套索(LASSO)回归。而对于LASSO类问题的优化算法研究主要集中在具有较低计算代价以及较快收敛特性的一阶优化算法。例如,在引入Nesterov加速策略后拥有o(1/k 2)收敛速率的快速迭代软阈值算法;结合变量可分离求解大规模问题的交替方向乘子法。这类算法对于弱凸性甚至包括一些非凸性问题均有良好的应用。
近来,基于即插即用先验(Plug-and-Play Prior,PPP)的迭代阈值算法以一个图像去噪滤波(滤波先验即为即插即用先验)替代原始的阈值收缩,即耦合图像去噪到基于前向模型的图像恢复框架。基于即插即用先验的迭代阈值算法显示出比传统的优化算法更为优越的成像质量。
然而,传统的基于即插即用先验的迭代阈值算法,仅是基于单一类型的即插即用先验的迭代阈值算法,没有考虑更多类型的即插即用先验,如此无法获得在伪影抑制与结构保护方面得到良好平衡的重建图像,从而无法提升图像重 建性能。
发明内容
为了解决上述现有技术存在的技术问题,本发明提供了一种结合了更多类型的即插即用先验的图像重建方法和图像重建装置。
根据本发明的实施例的一方面提供的图像重建方法包括:执行以下循环过程,直至满足循环结束条件:
基于目标图像结构滤波器对上一循环过程中得到的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据;基于引导图像结构滤波器对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据;根据本循环过程中的所述第一图像数据和所述第二图像数据得到本循环过程中的目标图像的图像数据;根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差得到本次循环过程中的数据残差;
其中,在未满足所述循环结束条件时,本次循环过程中得到的目标图像的图像数据和数据残差充当下一循环过程中的目标图像的图像数据和数据残差。
在上述一方面提供的图像重建方法的一个示例中,所述根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差得到本次循环过程中的数据残差,包括:根据上一循环过程中得到的目标图像的图像数据和数据残差并通过目标图像结构滤波的散度估计得到本循环过程中的第一残差修正因子;根据上一循环过程中得到的目标图像的图像数据和数据残差并通过引导图像结构滤波的散度估计得到本循环过程中的第二残差修正因子;根据本循环过程中的所述第一残差修正因子和所述第二残差修正因子得到本循环过程中的残差修正因子;根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及残差修正因子得到本次循环过程中的数据残差。
在上述一方面提供的图像重建方法的一个示例中,基于目标图像结构滤波器并根据下面的式子1对上一循环过程中得到的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据,
Figure PCTCN2020136550-appb-000001
其中,
Figure PCTCN2020136550-appb-000002
表示本循环过程中的所述第一图像数据,
Figure PCTCN2020136550-appb-000003
表示基于目标图像结构滤波器的滤波算子,x t-1表示上一循环过程中得到的目标图像的图像数据,z t-1上一循环过程中得到的数据残差,A表示欠采样傅里叶变换矩阵;
和/或,基于引导图像结构滤波器并根据下面的式子2对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据;
Figure PCTCN2020136550-appb-000004
其中,
Figure PCTCN2020136550-appb-000005
表示本循环过程中的所述第二图像数据,
Figure PCTCN2020136550-appb-000006
表示基于引导图像结构滤波器的滤波算子,x t-1表示上一循环过程中得到的目标图像的图像数据,z t-1上一循环过程中得到的数据残差,A表示欠采样傅里叶变换矩阵。
在上述一方面提供的图像重建方法的一个示例中,根据本循环过程中的所述第一图像数据和所述第二图像数据,并利用下面的式子3计算得到本循环过程中的目标图像的图像数据,
Figure PCTCN2020136550-appb-000007
其中,α表示权重参数,且0<α<1,x t表示本循环过程中的目标图像的图像数据。
在上述一方面提供的图像重建方法的一个示例中,根据上一循环过程中得到的目标图像的图像数据和数据残差,并利用下面的式子4得到本循环过程中的第一残差修正因子,
Figure PCTCN2020136550-appb-000008
其中,
Figure PCTCN2020136550-appb-000009
表示本循环过程中的基于目标图像结构滤波器滤波先验的第一残差修正因子,δ表示问题欠定程度的度量常数,
Figure PCTCN2020136550-appb-000010
表示基于目标图像结构滤波器的滤波算子对应的散度算子,x t-1表示上一循环过程中得到的目标图像的图像数据,z t-1上一循环过程中得到的数据残差,A表示欠采样傅里叶变换矩阵;
和/或,根据上一循环过程中得到的目标图像的图像数据和数据残差,并利用下面的式子5得到本循环过程中的基于引导图像结构滤波器滤波先验的第二残差修正因子,
Figure PCTCN2020136550-appb-000011
其中,
Figure PCTCN2020136550-appb-000012
表示本循环过程中的基于引导图像结构滤波器滤波先验的第二残差修正因子,δ表示问题欠定程度的度量常数,
Figure PCTCN2020136550-appb-000013
表示基于引导图像结构滤波器的滤波算子对应的散度算子,x t-1表示上一循环过程中得到的目标图像的图像数据,z t-1上一循环过程中得到的数据残差,A表示欠采样傅里叶变换矩阵。
在上述一方面提供的图像重建方法的一个示例中,根据本循环过程中的所述第一残差修正因子和所述第二残差修正因子,并利用下面的式子6计算得到本循环过程中的残差修正因子,
Figure PCTCN2020136550-appb-000014
其中,o t表示本循环过程中的残差修正因子,α表示权重参数,且0<α<1。
在上述一方面提供的图像重建方法的一个示例中,根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及残差修正因子,并利用下面的式子7计算得到本次循环过程中的数据残差,
[7]  z t=b-Ax t+o t
其中,z t本次循环过程中的数据残差,b表示K空间的欠采样数据。
根据本发明的实施例的另一方面提供的图像重建装置包括:循环操作直至满足循环结束条件的目标图像结构滤波器、引导图像结构滤波器、目标图像数据获取模块、数据残差获取模块;
所述目标图像结构滤波器用于对上一循环过程中得到的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据;所述引导图像结构滤波器用于对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据;所述目标图像数据获取模块用于根据本循环过程中的所述第一图像数据和所述第二图像数据得到本循环过程中的目标图像的图像数据;所述数据残差获取模块用于根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差得到本次循环过程中的数据残差;
其中,在未满足所述循环结束条件时,本次循环过程中得到的目标图像的图像数据和数据残差充当下一循环过程中的目标图像的图像数据和数据残差。
根据本发明的实施例的又一方面提供了一种电子设备,其包括:至少一个处理器,以及与所述至少一个处理器耦合的存储器,所述存储器存储指令,当所述指令被所述至少一个处理器执行时,使得所述至少一个处理器执行如上所述的图像重建方法。
根据本发明的实施例的再一方面提供了一种机器可读存储介质,其存储有可执行指令,所述指令当被执行时使得所述机器执行如上所述的图像重建方法。
有益效果:本发明采用复合型即插即用先验,能够帮助获得在伪影抑制与结构保护方面得到良好平衡的图像,并且还能够提升图像重建性能。
附图说明
通过结合附图进行的以下描述,本发明的实施例的上述和其它方面、特点和优点将变得更加清楚,附图中:
图1是示出了根据本发明的实施例的图像重建方法的流程图;
图2是示出了根据本发明的实施例的图像重建方法中获取本次循环过程中的数据残差的一个示例性方法的流程图;
图3是示出了根据本发明的实施例的图像重建装置的方框图;
图4是示出了根据本发明的实施例的实现图像重建方法的电子设备的方框图。
具体实施方式
以下,将参照附图来详细描述本发明的具体实施例。然而,可以以许多不同的形式来实施本发明,并且本发明不应该被解释为限制于这里阐述的具体实施例。相反,提供这些实施例是为了解释本发明的原理及其实际应用,从而使本领域的其他技术人员能够理解本发明的各种实施例和适合于特定预期应用的各种修改。
如本文中使用的,术语“包括”及其变型表示开放的术语,含义是“包括但不限于”。术语“基于”、“根据”等表示“至少部分地基于”、“至少部分地根据”。术语“一个实施例”和“一实施例”表示“至少一个实施例”。术语“另一个实施例”表示“至少一个其他实施例”。术语“第一”、“第二”等可以指代不同的或相同的对象。下面可以包括其他的定义,无论是明确的还是隐含的。除非上下文中明确地指明,否则一个术语的定义在整个说明书中是一致的。
如上所述,基于即插即用先验的迭代阈值算法,仅是基于单一类型的即插即用先验的迭代阈值算法,没有考虑更多类型的即插即用先验。在这种情况下,无法获得在伪影抑制与结构保护方面得到良好平衡的重建图像,从而也就无法提升图像重建性能。
为了获得在伪影抑制与结构保护方面得到良好平衡的重建图像,从而提升图像重建性能,根据本发明的实施例提供了一种结合了更多类型的即插即用先验来进行图像重建的图像重建方法和图像重建装置。该图像重建方法可以由电子设备执行,该电子设备执行以下循环过程,直至满足循环结束条件:基于目标图像结构滤波器对上一循环过程中得到的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据;基于引导图 像结构滤波器对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据;根据本循环过程中的所述第一图像数据和所述第二图像数据得到本循环过程中的目标图像的图像数据;根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差得到本次循环过程中的数据残差;其中,在未满足所述循环结束条件时,本次循环过程中得到的目标图像的图像数据和数据残差充当下一循环过程中的目标图像的图像数据和数据残差。
因此,在该图像重建方法中,采用目标图像结构滤波器和引导图像结构滤波器的复合型即插即用先验来对图像进行过滤处理,能够帮助获得在伪影抑制与结构保护方面得到良好平衡的图像,并且还能够提升图像重建性能。
在对根据本发明的实施例的图像重建方法的描述中,以磁共振成像为例进行说明。接下来,以磁共振成像为例对根据本发明的实施例的图像重建方法中应用到的一些基本概念及过程推导进行说明。
磁共振加速成像的图像重建任务可认为是求解一个L 1范数正则的LASSO回归问题,该LASSO回归问题的求解代价函数包括基于噪声高斯分布假设的数据拟合项以及关于图像在变换域系数的L 1范数约束。
首先构造求解上述优化问题的PPP(即插即用先验)-AMP(近似消息传递,Approximate Message Passing)算法框架,重点将设计以及耦合两类即插即用先验:一类为来自目标图像自身的结构滤波先验,另一类为来其他模态或参数的引导图像结构滤波先验。加权两类即插即用先验作为复合即插即用先验,并耦合到AMP算法的前向模型框架。
具体地,基于K空间欠采样机制的磁共振成像数据采集过程可以离散表示为下面的式子1。
[式子1]  b=Ax+ξ
这里,
Figure PCTCN2020136550-appb-000015
为待重建的目标图像的图像数据,
Figure PCTCN2020136550-appb-000016
为K空间欠采样数据,
Figure PCTCN2020136550-appb-000017
为欠采样傅里叶变换矩阵,
Figure PCTCN2020136550-appb-000018
为假设服从高斯分布的噪声。
对于上述欠定问题的求解往往是困难的,而在压缩感知理论框架下,基于 目标图像在某个稀疏变换域系数的L 1范数正则化,可以考虑从K空间欠采样数据b近乎完美地恢复(或称重建)原始目标图像x。针对求解而言,构造一个无约束的优化模型,其被表示为下面的式子2。
Figure PCTCN2020136550-appb-000019
这里,λ为平衡数据拟合以及稀疏正则的正数常量。R(·)表示某个稀疏变化域。
一系列拥有较低计算代价的迭代算法被提出用于求解式子2的凸优化问题,其中代表性算法为迭代软阈值算法(Iterative Soft-thresholding Algorithm,ISTA)具体算法可以被表示为下面的式子3和式子4。
[式子3]  x t=S τ(A Hz t-1+x t-1)
[式子4]  z t=b-Ax t
这里,迭代软阈值算法S τ(y)=(|y|-τ) +sign(y),x t表示第t次的x估计(即第t次迭代后重建的目标图像的图像数据),τ表示阈值参数。近似消息传递算法作为ISTA的变种,其核心在于引入了Onsager修正项以便对残差z t进行高斯化,从而进一步提高算法性能。因此,近似消息传递算法可以被表示为下面的式子5、式子6和式子7。
[式子5]  x t=S τ(A Hz t-1+x t-1)
Figure PCTCN2020136550-appb-000020
[式子7]  z t=b-Ax t+o t
这里,
Figure PCTCN2020136550-appb-000021
为Onsager修正项,S τ′为S τ的导数,δ为问题欠定程度的度量常数,<·>表示向量均值运算。
接着,引入广义去噪算子(即滤波算子)构造PPP-AMP算法框架,则构造的PPP-AMP算法可以被表示为下面的式子8、式子9和式子10。
Figure PCTCN2020136550-appb-000022
Figure PCTCN2020136550-appb-000023
[式子10]  z t=b-Ax t+o t
这里,
Figure PCTCN2020136550-appb-000024
为去噪算子,
Figure PCTCN2020136550-appb-000025
为去噪算子对应的散度算子。
在根据本发明的实施例中,将耦合两种不同类型的去噪先验,以构造性能提升的基于复合型即插即用先验的近似消息传递算法来进行图像的重建。
在一个示例中,第一类即插即用先验选用基于块匹配3D滤波(Block Matching and 3D Filtering,BM3D)算法作为来自目标图像x的自身结构滤波先验,其中,BM3D算法旨在通过与相邻图像块进行欧氏距离的度量匹配,所有相似图像块构造三维矩阵,在三维空间整体滤波,再将滤波结果反变换到二维图像。
在一个示例中,第二类即插即用先验选用交互性引导图像滤波(Mutually Guided Image Filtering,muGIF)算法作为来自其他模态或参数的引导图像x r的结构滤波先验,muGIF旨在通过与目标图像相似的引导图像结构信息的交互性度量,获得目标图像与引导图像间共享的解剖结构信息,在有效抑制图像伪影的同时强化目标图像的主要结构特征。交互性引导滤波一方面实现充分引入引导图像结构信息,另一方面避免因存在差异的图像内容带来的细节滤波偏差。
以上是以磁共振成像为例对应用到根据本发明的实施例的图像重建方法中的一些基本概念及过程推导进行的详细描述。
接下来,将结合附图来详细描述根据本发明的实施例的结合了复合型即插即用先验来进行图像重建的图像重建方法和图像重建装置。
根据本发明的实施例的图像重建方法可以由电子设备执行,该电子设备可以包括智能手机、平板电脑、个人计算机、云服务器、服务器等。
图1是示出了根据本发明的实施例的图像重建方法的流程图。
参照图1,在步骤S110中,基于目标图像结构滤波器对上一循环过程中得到的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据。
在一个示例中,目标图像结构滤波器具有上述的基于块匹配3D滤波算法,其可以实现对来自目标图像的自身结构滤波先验。
在一个示例中,基于目标图像结构滤波器并根据下面的式子11对上一循环过程中得到的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据。
Figure PCTCN2020136550-appb-000026
其中,
Figure PCTCN2020136550-appb-000027
表示本循环过程(第t次迭代)中的所述第一图像数据,
Figure PCTCN2020136550-appb-000028
表示基于目标图像结构滤波器的滤波算子,x t-1表示上一循环过程(第t-1次迭代)中得到的目标图像的图像数据,z t-1上一循环过程中得到的数据残差,A表示欠采样傅里叶变换矩阵。
在步骤S120中,基于引导图像结构滤波器对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据。
在一个示例中,引导图像结构滤波器具有上述的交互性引导图像滤波算法,其可以实现对来自其他模态或参数的引导图像的结构滤波先验。在一个示例中,引导图像与目标图像相似。
在一个示例中,基于引导图像结构滤波器并根据下面的式子12对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据。
Figure PCTCN2020136550-appb-000029
其中,
Figure PCTCN2020136550-appb-000030
表示本循环过程中的所述第二图像数据,
Figure PCTCN2020136550-appb-000031
表示基于引导图像结构滤波器的滤波算子。
在步骤S130中,根据本循环过程中的所述第一图像数据和所述第二图像数据得到本循环过程中的目标图像的图像数据。
在一个示例中,根据本循环过程中的所述第一图像数据和所述第二图像数据,并利用下面的式子13计算得到本循环过程中的目标图像的图像数据。
Figure PCTCN2020136550-appb-000032
其中,α表示权重参数,且0<α<1,x t表示本循环过程中的目标图像的图像数据。
在步骤S140中,根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差得到本次循环过程中的数据残差。
在一个示例中,K空间的欠采样数据可以例如是由磁共振成像装置欠采样采集的数据。
图2是示出了根据本发明的实施例的图像重建方法中获取本次循环过程中的数据残差的一个示例性方法的流程图。
参照图2,在步骤S141中,根据上一循环过程中得到的目标图像的图像数据和数据残差并通过目标图像结构滤波的散度估计得到本循环过程中的第一残差修正因子。
在一个示例中,根据上一循环过程中得到的目标图像的图像数据和数据残差,并利用下面的式子14得到本循环过程中的第一残差修正因子。
Figure PCTCN2020136550-appb-000033
其中,
Figure PCTCN2020136550-appb-000034
表示本循环过程中的基于目标图像结构滤波器滤波先验的第一残差修正因子,δ表示问题欠定程度的度量常数,
Figure PCTCN2020136550-appb-000035
表示基于目标图像结构滤波器的滤波算子对应的散度算子。
在步骤S142中,根据上一循环过程中得到的目标图像的图像数据和数据残差并通过引导图像结构滤波的散度估计得到本循环过程中的第二残差修正因子。
在一个示例中,根据上一循环过程中得到的目标图像的图像数据和数据残差,并利用下面的式子15得到本循环过程中的第二残差修正因子。
Figure PCTCN2020136550-appb-000036
其中,
Figure PCTCN2020136550-appb-000037
表示本循环过程中的基于引导图像结构滤波器滤波先验的第二残差修正因子,
Figure PCTCN2020136550-appb-000038
表示基于引导图像结构滤波器的滤波算子对应的散度算子。
在步骤S143中,根据本循环过程中的所述第一残差修正因子和所述第二残差修正因子得到本循环过程中的残差修正因子。
在一个示例中,根据本循环过程中的所述第一残差修正因子和所述第二残差修正因子,并利用下面的式子16计算得到本循环过程中的残差修正因子。
Figure PCTCN2020136550-appb-000039
其中,o t表示本循环过程中的残差修正因子。
在步骤S144中,根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及残差修正因子得到本次循环过程中的数据残差。
在一个示例中,根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及残差修正因子,并利用下面的式子17计算得到本次循环过程中的数据残差。
[式子17]  z t=b-Ax t+o t
其中,z t表示本次循环过程中的数据残差,b表示K空间的欠采样数据。
继续参照图1,在步骤S150中,判断是否满足循环结束条件。如果是,结束重建;如果否,将本次循环过程中得到的目标图像的图像数据和数据残差充当下一循环过程中的目标图像的图像数据和数据残差,并进入步骤S110。
这里,循环结束条件可以是指定的。在一个示例中,循环结束条件可以包括达到预定循环次数(或称迭代次数)。
图3是示出了根据本发明的实施例的图像重建装置的方框图。
图像重建装置300应用于电子设备中,以由电子设备来执行。参照图3,图像重建装置300包括:目标图像结构滤波器310、引导图像结构滤波器320、目标图像数据获取模块330以及数据残差获取模块340。目标图像结构滤波器310、引导图像结构滤波器320、目标图像数据获取模块330以及数据残差获取模块340循环操作,直至满足循环结束条件。其中,循环结束条件可以是指定的。在一个示例中,循环结束条件可以包括达到预定循环次数(或称迭代次数)。
目标图像结构滤波器310被配置为对上一循环过程中得到的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据。在一个示例中,目标图像结构滤波器310可以被配置为根据上面的式子11对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据。
引导图像结构滤波器320被配置为对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据。在一个示例中,引导图像结构滤波器320可被配置为根据上面的式子12对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据。
目标图像数据获取模块330被配置为用于根据本循环过程中的所述第一图像数据和所述第二图像数据得到本循环过程中的目标图像的图像数据。在一个示例中,目标图像数据获取模块330可以被配置为根据本循环过程中的所述第一图像数据和所述第二图像数据,并利用上面的式子13计算得到本循环过程中的目标图像的图像数据。
数据残差获取模块340被配置为根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差得到本次循环过程中的数据残差。在一个示例中,数据残差获取模块340可以被配置为根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差,并利用上面的式子15、式子16和式子17计算得到本次循环过程中的数据残差。
以上参照图1到图3,对根据本发明的实施例的图像重建方法和图像重建装置进行了描述。
根据本发明的实施例的图像重建装置可以采用硬件实现,也可以采用软件或者硬件和软件的组合来实现。以软件实现为例,作为一个逻辑意义上的装置,是通过其所在设备的处理器将存储器中对应的计算机程序指令读取到内存中运行形成的。在本发明的实施例中,使用进行图像重建的装置例如可以利用电子设备来实现。
图4是示出了根据本发明的实施例的实现图像重建方法的电子设备的方框图。
参照图4,电子设备400可以包括至少一个处理器410、存储器(例如,非易失性存储器)420、内存430和通信接口440,并且至少一个处理器410、存储器420、内存430和通信接口440经由总线450连接在一起。至少一个处理器410执行在存储器中存储或编码的至少一个计算机可读指令(即,上述以软件形式实现的元素)。
在一个示例中,在存储器中存储计算机可执行指令,其当执行时使得至少一个处理器410执行以下循环过程,直至满足循环结束条件:基于目标图像结构滤波器对上一循环过程中得到的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据;基于引导图像结构滤波器对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据;根据本循环过程中的所述第一图像数据和所述第二图像数据得到本循环过程中的目标图像的图像数据;根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差得到本次循环过程中的数据残差;其中,在未满足所述循 环结束条件时,本次循环过程中得到的目标图像的图像数据和数据残差充当下一循环过程中的目标图像的图像数据和数据残差。
应该理解,在存储器中存储的计算机可执行指令当执行时使得至少一个处理器410在进行根据本发明的各个实施例中结合以上图1至图3描述的各种操作和功能。
根据一个实施例,提供了一种例如机器可读介质的程序产品。机器可读介质可以具有指令(即,上述以软件形式实现的元素),该指令当被机器执行时,使得机器执行本发明的各个实施例中的结合以上图1至图3描述的各种操作和功能。
具体地,可以提供配有可读存储介质的系统或者装置,在该可读存储介质上存储着实现上述实施例中任一实施例的功能的软件程序代码,且使该系统或者装置的计算机或处理器读出并执行存储在该可读存储介质中的指令。
在这种情况下,从可读介质读取的程序代码本身可实现上述实施例中任何一项实施例的功能,因此机器可读代码和存储机器可读代码的可读存储介质构成了本发明的实施例的一部分。
可读存储介质的实施例包括软盘、硬盘、磁光盘、光盘(如CD-ROM、CD-R、CD-RW、DVD-ROM、DVD-RAM、DVD-RW、DVD-RW)、磁带、非易失性存储卡和ROM。可选择地,可以由通信网络从服务器计算机上或云上下载程序代码。
上述对本发明的特定实施例进行了描述。其它实施例在所附权利要求书的范围内。在一些情况下,在权利要求书中记载的动作或步骤可以按照不同于实施例中的顺序来执行并且仍然可以实现期望的结果。另外,在附图中描绘的过程不一定要求示出的特定顺序或者连续顺序才能实现期望的结果。在某些实施方式中,多任务处理和并行处理也是可以的或者可能是有利的。
上述各流程和各系统结构图中不是所有的步骤和单元都是必须的,可以根据实际的需要忽略某些步骤或单元。各步骤的执行顺序不是固定的,可以根据需要进行确定。上述各实施例中描述的装置结构可以是物理结构,也可以是逻辑结构,即,有些单元可能由同一物理实体实现,或者,有些单元可 能分由多个物理实体实现,或者,可以由多个独立设备中的某些部件共同实现。
在整个本说明书中使用的术语“示例性”、“示例”等意味着“用作示例、实例或例示”,并不意味着比其它实施例“优选”或“具有优势”。出于提供对所描述技术的理解的目的,具体实施方式包括具体细节。然而,可以在没有这些具体细节的情况下实施这些技术。在一些实例中,为了避免对所描述的实施例的概念造成难以理解,公知的结构和装置以框图形式示出。
以上结合附图详细描述了本发明的实施例的可选实施方式,但是,本发明的实施例并不限于上述实施方式中的具体细节,在本发明的实施例的技术构思范围内,可以对本发明的实施例的技术方案进行多种简单变型,这些简单变型均属于本发明的实施例的保护范围。
本说明书内容的上述描述被提供来使得本领域任何普通技术人员能够实现或者使用本说明书内容。对于本领域普通技术人员来说,对本说明书内容进行的各种修改是显而易见的,并且,也可以在不脱离本说明书内容的保护范围的情况下,将本文所定义的一般性原理应用于其它变型。因此,本说明书内容并不限于本文所描述的示例和设计,而是与符合本文公开的原理和新颖性特征的最广范围相一致。

Claims (10)

  1. 一种图像重建方法,其特征在于,所述图像重建方法包括:
    执行以下循环过程,直至满足循环结束条件:
    基于目标图像结构滤波器对上一循环过程中得到的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据;
    基于引导图像结构滤波器对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据;
    根据本循环过程中的所述第一图像数据和所述第二图像数据得到本循环过程中的目标图像的图像数据;
    根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差得到本次循环过程中的数据残差;
    其中,在未满足所述循环结束条件时,本次循环过程中得到的目标图像的图像数据和数据残差充当下一循环过程中的目标图像的图像数据和数据残差。
  2. 根据权利要求1所述的图像重建方法,其特征在于,所述根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差得到本次循环过程中的数据残差,包括:
    根据上一循环过程中得到的目标图像的图像数据和数据残差并通过目标图像结构滤波的散度估计得到本循环过程中的第一残差修正因子;
    根据上一循环过程中得到的目标图像的图像数据和数据残差并通过引导图像结构滤波的散度估计得到本循环过程中的第二残差修正因子;
    根据本循环过程中的所述第一残差修正因子和所述第二残差修正因子得到本循环过程中的残差修正因子;
    根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及残差修正因子得到本次循环过程中的数据残差。
  3. 根据权利要求1或2所述的图像重建方法,其特征在于,基于目标图像结构滤波器并根据下面的式子1对上一循环过程中得到的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据,
    Figure PCTCN2020136550-appb-100001
    其中,
    Figure PCTCN2020136550-appb-100002
    表示本循环过程中的所述第一图像数据,
    Figure PCTCN2020136550-appb-100003
    表示基于目标图像结构滤波器的滤波算子,x t-1表示上一循环过程中得到的目标图像的图像数据,z t-1上一循环过程中得到的数据残差,A表示欠采样傅里叶变换矩阵;
    和/或,基于引导图像结构滤波器并根据下面的式子2对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据;
    Figure PCTCN2020136550-appb-100004
    其中,
    Figure PCTCN2020136550-appb-100005
    表示本循环过程中的所述第二图像数据,
    Figure PCTCN2020136550-appb-100006
    表示基于引导图像结构滤波器的滤波算子,x t-1表示上一循环过程中得到的目标图像的图像数据,z t-1上一循环过程中得到的数据残差,A表示欠采样傅里叶变换矩阵。
  4. 根据权利要求3所述的图像重建方法,其特征在于,根据本循环过程中的所述第一图像数据和所述第二图像数据,并利用下面的式子3计算得到本循环过程中的目标图像的图像数据,
    Figure PCTCN2020136550-appb-100007
    其中,α表示权重参数,且0<α<1,x t表示本循环过程中的目标图像的图像数据。
  5. 根据权利要求2所述的图像重建方法,其特征在于,根据上一循环过程中得到的目标图像的图像数据和数据残差,并利用下面的式子4得到本循环过程中的第一残差修正因子,
    Figure PCTCN2020136550-appb-100008
    其中,
    Figure PCTCN2020136550-appb-100009
    表示本循环过程中的基于目标图像结构滤波器滤波先验的第一残差修正因子,δ表示问题欠定程度的度量常数,
    Figure PCTCN2020136550-appb-100010
    表示基于目标图像结构滤波器的滤波算子对应的散度算子,x t-1表示上一循环过程中得到的目标图像的图像数据,z t-1上一循环过程中得到的数据残差,A表示欠采样傅里叶变换矩阵;
    和/或,根据上一循环过程中得到的目标图像的图像数据和数据残差,并利用下面的式子5得到本循环过程中的第二残差修正因子,
    Figure PCTCN2020136550-appb-100011
    其中,
    Figure PCTCN2020136550-appb-100012
    表示本循环过程中的基于引导图像结构滤波器滤波先验的第二残差修正因子,δ表示问题欠定程度的度量常数,
    Figure PCTCN2020136550-appb-100013
    表示基于引导图像结构滤波器的滤波算子对应的散度算子,x t-1表示上一循环过程中得到的目标图像的图像数据,z t-1上一循环过程中得到的数据残差,A表示欠采样傅里叶变换矩阵。
  6. 根据权利要求5所述的图像重建方法,其特征在于,根据本循环过程中的所述第一残差修正因子和所述第二残差修正因子,并利用下面的式子6计算得到本循环过程中的残差修正因子,
    Figure PCTCN2020136550-appb-100014
    其中,o t表示本循环过程中的残差修正因子,α表示权重参数,且0<α<1。
  7. 根据权利要求6所述的图像重建方法,其特征在于,根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及残差修正因子,并利用下面的式子7计算得到本次循环过程中的数据残差,
    [7]  z t=b-Ax t+o t
    其中,z t本次循环过程中的数据残差,b表示K空间的欠采样数据。
  8. 一种图像重建装置,其特征在于,所述图像重建装置包括循环操作直至满足循环结束条件的目标图像结构滤波器、引导图像结构滤波器、目标图像数据获取模块、数据残差获取模块;
    所述目标图像结构滤波器用于对上一循环过程中得到的目标图像的图像数据和数据残差进行滤波处理以得到本循环过程中的第一图像数据;
    所述引导图像结构滤波器用于对上一循环过程中得到的目标图像数据和数据残差进行滤波处理以得到本循环过程中的第二图像数据;
    所述目标图像数据获取模块用于根据本循环过程中的所述第一图像数据和所述第二图像数据得到本循环过程中的目标图像的图像数据;
    所述数据残差获取模块用于根据K空间的欠采样数据、本循环过程中的目标图像的图像数据以及上一循环过程中得到的目标图像数据和数据残差得到本次循环过程中的数据残差;
    其中,在未满足所述循环结束条件时,本次循环过程中得到的目标图像的图像数据和数据残差充当下一循环过程中的目标图像的图像数据和数据残差。
  9. 一种电子设备,其特征在于,包括:
    至少一个处理器,以及
    与所述至少一个处理器耦合的存储器,所述存储器存储指令,当所述指令被所述至少一个处理器执行时,使得所述至少一个处理器执行如权利要求1至7中任一所述的图像重建方法。
  10. 一种机器可读存储介质,其存储有可执行指令,其特征在于,所述指令当被执行时使得所述机器执行如权利要求1到7中任一所述的图像重建方法。
PCT/CN2020/136550 2020-12-07 2020-12-15 图像重建方法和装置、电子设备以及机器可读存储介质 Ceased WO2022120899A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011440541.8 2020-12-07
CN202011440541.8A CN112489155B (zh) 2020-12-07 2020-12-07 图像重建方法和装置、电子设备以及机器可读存储介质

Publications (1)

Publication Number Publication Date
WO2022120899A1 true WO2022120899A1 (zh) 2022-06-16

Family

ID=74941504

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/136550 Ceased WO2022120899A1 (zh) 2020-12-07 2020-12-15 图像重建方法和装置、电子设备以及机器可读存储介质

Country Status (2)

Country Link
CN (1) CN112489155B (zh)
WO (1) WO2022120899A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113192151B (zh) * 2021-04-08 2022-12-27 广东工业大学 一种基于结构相似性的mri图像重建方法
CN115754857B (zh) * 2021-09-03 2026-02-13 中国科学院深圳先进技术研究院 一种低场磁共振快速成像方法、终端设备及计算机存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150287223A1 (en) * 2014-04-04 2015-10-08 The Board Of Trustees Of The University Of Illinois Highly accelerated imaging and image reconstruction using adaptive sparsifying transforms
CN109671129A (zh) * 2018-12-14 2019-04-23 深圳先进技术研究院 一种自适应参数学习的动态磁共振图像重建方法和装置
CN110148215A (zh) * 2019-05-22 2019-08-20 哈尔滨工业大学 一种基于平滑约束和局部低秩约束模型的四维磁共振图像重建方法
CN111192214A (zh) * 2019-12-27 2020-05-22 上海商汤智能科技有限公司 图像处理的方法、装置、电子设备及存储介质
CN111798391A (zh) * 2020-06-29 2020-10-20 东软医疗系统股份有限公司 图像处理方法及装置、医学影像设备及系统

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106485764B (zh) * 2016-11-02 2019-11-22 中国科学技术大学 Mri图像的快速精确重建方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150287223A1 (en) * 2014-04-04 2015-10-08 The Board Of Trustees Of The University Of Illinois Highly accelerated imaging and image reconstruction using adaptive sparsifying transforms
CN109671129A (zh) * 2018-12-14 2019-04-23 深圳先进技术研究院 一种自适应参数学习的动态磁共振图像重建方法和装置
CN110148215A (zh) * 2019-05-22 2019-08-20 哈尔滨工业大学 一种基于平滑约束和局部低秩约束模型的四维磁共振图像重建方法
CN111192214A (zh) * 2019-12-27 2020-05-22 上海商汤智能科技有限公司 图像处理的方法、装置、电子设备及存储介质
CN111798391A (zh) * 2020-06-29 2020-10-20 东软医疗系统股份有限公司 图像处理方法及装置、医学影像设备及系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZONG, CHUNMEI ET AL.: "Compressed Sensing Magnetic Resonance Imaging Based on Deep Priors and Non-Local Similarity", JOURNAL OF COMPUTER APPLICATIONS, vol. 40, no. 10, 10 October 2020 (2020-10-10), pages 3054 - 3059, XP055940730 *

Also Published As

Publication number Publication date
CN112489155A (zh) 2021-03-12
CN112489155B (zh) 2024-01-26

Similar Documents

Publication Publication Date Title
Elad et al. Image denoising: The deep learning revolution and beyond—a survey paper
Ongie et al. Deep learning techniques for inverse problems in imaging
Ongie et al. Off-the-grid recovery of piecewise constant images from few Fourier samples
Huang et al. Iterative weighted sparse representation for X‐ray cardiovascular angiogram image denoising over learned dictionary
Schlemper et al. A deep cascade of convolutional neural networks for dynamic MR image reconstruction
US10671939B2 (en) System, method and computer-accessible medium for learning an optimized variational network for medical image reconstruction
KR102673155B1 (ko) 스코어 기반의 확산 모델을 이용한 자기공명영상 복원 방법 및 그 장치
CN105761216B (zh) 一种图像去噪处理方法及装置
Zosso et al. A primal-dual projected gradient algorithm for efficient Beltrami regularization
CN109375125B (zh) 一种修正正则化参数的压缩感知磁共振成像重建方法
CN104574456B (zh) 一种基于图正则化稀疏编码的磁共振超欠采样k数据成像方法
CN107945129B (zh) 一种mri图像重构方法
CN107845065A (zh) 超分辨率图像重建方法和装置
CN112489155B (zh) 图像重建方法和装置、电子设备以及机器可读存储介质
CN106056554B (zh) 一种梯度域卷积稀疏编码的磁共振快速成像方法
CN115270892B (zh) 一种信号重建方法、系统、装置及存储介质
CN109064402A (zh) 基于增强非局部总变分模型先验的单幅图像超分辨率重建方法
CN116502441A (zh) 一种基于低秩约束的gap快速反演方法、存储介质及设备
US8582909B2 (en) Adaptive multi-grid contrast optical flow
Ma et al. Adaptive image denoising method based on diffusion equation and deep learning
Zou et al. Video matting via sparse and low-rank representation
Ye et al. FAST TOTAL VARIATION WAVELET INPAINTING VIA APPROXIMATED PRIMAL-DUAL HYBRID GRADIENT ALGORITHM.
Karaoglu et al. A plug-and-play deep denoiser prior model for accelerated MRI reconstruction
CN116128995B (zh) 一种融合信号模型与深度学习的磁共振重建方法及装置
CN115754857B (zh) 一种低场磁共振快速成像方法、终端设备及计算机存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20964832

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20964832

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 27.05.2024)

122 Ep: pct application non-entry in european phase

Ref document number: 20964832

Country of ref document: EP

Kind code of ref document: A1