WO2017067130A1 - 一种气动光学热辐射噪声校正方法与系统 - Google Patents

一种气动光学热辐射噪声校正方法与系统 Download PDF

Info

Publication number
WO2017067130A1
WO2017067130A1 PCT/CN2016/076616 CN2016076616W WO2017067130A1 WO 2017067130 A1 WO2017067130 A1 WO 2017067130A1 CN 2016076616 W CN2016076616 W CN 2016076616W WO 2017067130 A1 WO2017067130 A1 WO 2017067130A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
scale
deviation field
restored
degraded
Prior art date
Application number
PCT/CN2016/076616
Other languages
English (en)
French (fr)
Inventor
李乐仁瀚
桑农
高常鑫
颜露新
王金
张士伟
邵远杰
彭军才
Original Assignee
华中科技大学
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 华中科技大学 filed Critical 华中科技大学
Priority to US15/126,271 priority Critical patent/US9904987B2/en
Publication of WO2017067130A1 publication Critical patent/WO2017067130A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00Image enhancement or restoration
    • G06T5/70Denoising; Smoothing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0004Industrial image inspection
    • G06T7/001Industrial image inspection using an image reference approach
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00Geometric image transformations in the plane of the image
    • G06T3/40Scaling of whole images or parts thereof, e.g. expanding or contracting
    • G06T3/4007Scaling of whole images or parts thereof, e.g. expanding or contracting based on interpolation, e.g. bilinear interpolation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00Geometric image transformations in the plane of the image
    • G06T3/40Scaling of whole images or parts thereof, e.g. expanding or contracting
    • G06T3/4053Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
    • G06T3/4076Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution using the original low-resolution images to iteratively correct the high-resolution images
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10048Infrared image
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20076Probabilistic image processing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30108Industrial image inspection

Definitions

  • the invention belongs to the field of pattern recognition technology, and more particularly to a method and system for correcting aero-optical thermal radiation noise based on multi-scale deviation field estimation.
  • the existing pneumatic optical thermal radiation noise correction methods mainly include two types: (1) based on photoelectric correction, the method is based on the flight trajectory characteristics of the aircraft to establish the relationship between shock radiation and image degradation and working wavelength, and then guide Optimized analysis of the working wavelength of the head to select a reasonable working wavelength; (2) Based on the optical window design correction, the method mainly reduces the aerodynamic heating degree of the optical hood in the turbulent flow by designing an appropriate window structure and cooling method. To correct the aerodynamic thermal radiation noise.
  • the disadvantage of the above two methods is that both methods adopt the "pre-prevention" method to suppress the aerodynamic thermal radiation noise at the source, but the optical detection image that has generated the thermal radiation noise cannot be restored.
  • the present invention provides a pneumatic optical thermal radiation noise correction method and system based on multi-scale deviation field estimation, wherein the present invention directly performs aero-optical thermal radiation noise correction on the degraded image itself, especially through
  • the multi-scale bias field estimation method that is, the above-mentioned one-scale estimation result is used as the next scale estimation initial value to obtain the update solution of the optimal solution.
  • the procedure improves the convergence speed of the algorithm, and can effectively solve the problem that the correction effect is poor, the method complexity is high, and the thermal radiation noise correction cannot be performed at the image level.
  • a method for correcting a pneumatic optical thermal radiation noise comprising the following steps:
  • the multi-scale degraded image set is obtained by performing equal-ratio ⁇ downsampling on the degraded image g 0 by bilinear interpolation.
  • estimating the deviation field b j of the current scale and the restored image i j includes:
  • ⁇ and ⁇ are data constraint term coefficients (also called penalty coefficients), and the values of ⁇ and ⁇ can be adjusted according to experimental requirements;
  • the optimization solution model is decomposed into two sub-problems, namely the i-sub problem and the b-sub problem.
  • the Split Bregman iterative method is introduced when solving the i-sub problem.
  • step (3) the deviation field estimation value and the restored image estimation value are upsampled according to the downsampling ratio ⁇ .
  • a pneumatic optical thermal radiation noise correction system comprising:
  • the processing module includes a multi-scale downsampling and upsampling unit cell, wherein the downsampling unit for preprocessing degraded image g 0, g 0 performed on the degraded image by bilinear interpolation method or the like to obtain the down-sampling ratio ⁇ degradation multiscale
  • the image set ⁇ g 0 , g 1 , . . . , g s ⁇ , the upsampling unit is configured to perform the upsampling process on the current field deviation field estimation value and the restored image estimation value output by the deviation field estimation module, and feed back to the deviation field estimation.
  • the deviation field estimation module is configured to obtain the deviation field estimation value of the current scale and the restored image estimation value according to the degraded image of the minimum scale in the multi-scale degraded image set, and update the iteration until the deviation field estimation value of the original scale is obtained. And Output to the image restoration module;
  • An image restoration module for estimating a deviation field from the original scale
  • the degraded image g 0 is restored to obtain a pneumatic optical thermal radiation noise corrected image i 0 .
  • the deviation field estimation module includes an image selection unit and an iterative processing unit, and the image selection unit is configured to select the minimum-scale degraded image g s in the set and serve as an initialization input of the iterative processing unit;
  • the iterative processing unit is configured to obtain the deviation field estimation value of the current scale and the restored image estimation value by updating the iteration until the deviation field estimation value of the original scale is obtained And Output to the image restoration module;
  • the multi-scale degraded image set is obtained by the multi-scale processing module performing equal-ratio ⁇ downsampling on the degraded image g 0 by bilinear interpolation.
  • the upsampling unit upsamples the deviation field estimation value and the restored image estimation value of the current scale according to the downsampling ratio ⁇ .
  • the technical solution of the present invention can correct the aero-optical thermal radiation noise only by the degraded image obtained by the optical detection system, and can obtain a more accurate and ideal image restoration effect;
  • the present invention uses the estimation result of the previous scale as the input value of the next scale estimation by using the multi-scale deviation field estimation method, so that each estimation can obtain the optimal solution faster. Significantly improved the convergence speed of the algorithm;
  • the optimization is solved by the Split Bregman iteration, and the calculation is further improved.
  • the convergence speed of the method is more concise and clear, so the invention is particularly suitable for the field of pneumatic optical thermal radiation noise correction.
  • FIG. 1 is a general flow chart of a pneumatic optical thermal radiation noise correction method of the present invention
  • Figure 2 is a flow chart for solving the solution by iterative optimization
  • Figure 3 is a structural view of a pneumatic optical thermal radiation noise correction system of the present invention.
  • Fig. 4 is a schematic view showing the effect of correcting the aerodynamic thermal radiation noise of the present invention.
  • Deviation field is also called a non-uniform field, which is represented by the unevenness of brightness.
  • the concept of the deviation field is widely used in medical image processing.
  • Uneven brightness means that the pixels in the same area of the image exhibit a change in brightness, and the spatial variation of such image brightness is caused by inherent inhomogeneities of artifacts and tissue properties.
  • the intensity or color inhomogeneity caused by the bias field is a difficult problem for many image processing algorithms, such as image segmentation, image retrieval, target tracking and recognition, etc., all of which face this problem.
  • image segmentation usually uses the assumption that the target image is a piecewise constant. If the image is affected by the bias field, the intensity of the image will gradually change.
  • Multi-scale The multi-scale method is faster when the image is smaller in scale, and the obtained estimated value is used as the initial value of the next scale after up-sampling, so that the energy function can converge to the minimum value more quickly. Optimal solution.
  • FIG. 1 is a general flowchart of a method for correcting a pneumatic optical thermal radiation noise according to the present invention
  • the method of the present invention specifically includes the following steps:
  • an optimization solution model can be established, as shown in formula (1):
  • ⁇ and ⁇ are the penalty coefficients, and their size determines the smoothness of the deviation field and the retention of the detailed information of the restored image structure.
  • the larger the value of ⁇ , the smoother the estimated deviation field, and the value of ⁇ The larger the estimated structural details of the restored image, the more it remains.
  • the values of ⁇ and ⁇ can be adjusted according to experimental needs.
  • Deviation field estimate in equation (1) And restored image estimates It is a variable that needs to be optimized. Therefore, it is optimized to solve one problem, that is, one variable is fixed, and the other variable is optimized and solved alternately, so equation (1) is decomposed into two sub-problems, namely “i- Sub-problems and "b-sub-questions”.
  • Equation (5) can be optimized by iterative updating of variables i, d, r.
  • the specific solution process is as shown in formula (6):
  • Equation (7) can find the optimal solution directly by the method of least squares solution.
  • FIG. 3 is a structural diagram of the system corresponding to FIG. 1.
  • a pneumatic optical thermal radiation noise correction system includes:
  • Multiscale image processing module is divided into the downsampling unit and an image sampling unit, downsampling unit for preprocessing degraded image g 0, g 0 performed on the degraded image by bilinear interpolation method or the like to give a ratio of down-sampling the multi-scale ⁇
  • the degraded image set ⁇ g 0 , g 1 , . . . , g s ⁇ , the upsampling unit is configured to perform upsampling on the current scale deviation field estimation value and the restored image estimation value output by the deviation field estimation module, and feed back to the deviation field estimation.
  • the upsampling unit upsamples the current scale deviation field estimate and the restored image estimate according to the downsampling ratio ⁇ .
  • the deviation field estimation module is configured to obtain the deviation field estimation value of the current scale and the restored image estimation value according to the degraded image of the minimum scale in the multi-scale degraded image set, and update the iteration until the deviation field estimation value of the original scale is obtained. And Output to the image restoration module;
  • the deviation field estimation module specifically includes an image selection unit and an iterative processing unit, wherein the image selection unit is configured to select a minimum-scale degraded image g s in the set and serve as an initialization input of an iterative processing unit; the iterative processing unit For obtaining the deviation field estimation value of the current scale and the restored image estimation value by updating the iteration until the deviation field estimation value of the original scale is obtained And Output to the image restoration module.
  • An image restoration module is coupled to the deviation field estimation module for estimating a deviation field based on the original scale
  • the degraded image g 0 is restored to obtain a pneumatic optical thermal radiation noise corrected image i 0 .
  • FIG. 4 is a schematic diagram of the effect of the correction of the aero-optical thermal radiation noise.
  • the degraded image itself is downsampled first, and the multi-scale deviation field estimation is performed. Obtain the estimated value of the deviation field of the original scale, and finally restore the degraded image. Comparing the restored image with the original image in FIG. 4, it can be seen that the technical solution of the present invention can obtain a more accurate and ideal image restoration effect, and can significantly improve the convergence speed of the algorithm while ensuring more accurate image restoration, and thus is particularly suitable for pneumatic optical heat. Radiation noise correction field.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Processing (AREA)

Abstract

一种气动光学热辐射噪声校正方法,包括以下步骤:首先将退化图像进行预处理得到多尺度退化图像集合;然后,根据上述多尺度退化图像集合,以上一尺度估计结果作为下一尺度估计初始值来获得最优解的更新迭代过程,完成原始尺度的偏差场估计;最后,根据原始尺度的偏差场估计值对退化图像进行复原,得到气动光学热辐射噪声校正图像。还涉及一种气动光学热辐射噪声校正系统,包括多尺度处理模块、偏差场估计模块和图像复原模块。该方法和系统有效解决了现有方法中存在的校正效果差、方法复杂度高、无法在图像层面进行热辐射噪声校正的问题,适用于气动光学热辐射噪声图像复原的应用场合。

Description

一种气动光学热辐射噪声校正方法与系统 [技术领域]
本发明属于模式识别技术领域,更具体地,涉及一种基于多尺度偏差场估计的气动光学热辐射噪声校正方法与系统。
[背景技术]
当高速飞行器在超声速状态下飞行时,其光学窗口被气动加热而处于一个严重的气动热环境中。在此状况下,光学窗口所获取的图像会受到热辐射噪声的干扰,从而降低了探测系统对目标探测的信噪比和图像质量。为了获得清晰的探测图像,需要对气动光学热辐射噪声进行校正。其校正的准确性关系到光学探测系统中更高层的图像处理,因此气动光学热辐射噪声校正技术已成为高速飞行器光学探测系统中的一大关键。
现有的气动光学热辐射噪声校正方法主要包括两类:(1)基于光电校正,该方法是根据飞行器飞行弹道特性建立激波辐射及图像退化与工作波长的关系,在此基础上进行导引头工作波长的优化分析,选择合理的工作波长;(2)基于光学窗口设计校正,该方法主要是通过设计适当的窗口结构形式以及致冷方式,以减少光学头罩在湍流中的气动加热程度,从而校正气动热辐射噪声。以上两种方法的缺陷在于:两种方法均采取的是“事前预防”的方式在源头上抑制气动热辐射噪声,然而对于已经产生了热辐射噪声的光学探测图像无法进行复原处理。
[发明内容]
针对现有技术的以上缺陷,本发明提供一种基于多尺度偏差场估计的气动光学热辐射噪声校正方法与系统,其中,本发明直接对退化图像本身进行气动光学热辐射噪声校正,尤其是通过多尺度偏差场估计方法,即以上一尺度估计结果作为下一尺度估计初始值来获得最优解的更新迭代过 程,提升了算法收敛速度,相应可有效解决现有方法中存在的校正效果差、方法复杂度高、无法在图像层面进行热辐射噪声校正的问题。
为实现上述目的,按照本发明的一个方面,提出了一种气动光学热辐射噪声校正方法,包括以下步骤:
(1)对退化图像g0进行预处理,得到尺度由大到小依次排序的多尺度退化图像集合{g0,g1,…,gs};
(2)选取所述集合中最小尺度退化图像gs作为当前尺度退化图像gj,对当前尺度退化图像的偏差场以及复原图像进行估计,得到对应的偏差场估计值
Figure PCTCN2016076616-appb-000001
及复原图像估计值
Figure PCTCN2016076616-appb-000002
(3)根据步骤(1)中的集合对当前尺度的偏差场估计值
Figure PCTCN2016076616-appb-000003
及复原图像估计值
Figure PCTCN2016076616-appb-000004
进行相邻尺度上采样,即令j=j-1,得到进行下一次迭代的当前尺度退化图像gj-1的偏差场bj-1及复原图象ij-1,对当前尺度的偏差场以及复原图像进行估计,继而得到对应的偏差场估计值
Figure PCTCN2016076616-appb-000005
及复原图像估计值
Figure PCTCN2016076616-appb-000006
重复步骤(3),直至得到原始尺度的偏差场估计值
Figure PCTCN2016076616-appb-000007
(4)根据所述原始尺度的偏差场估计值
Figure PCTCN2016076616-appb-000008
对退化图像g0进行复原,得到气动光学热辐射噪声校正图像i0
作为进一步优选的,在步骤(1)中,所述多尺度退化图像集合是通过双线性插值法对退化图像g0进行等比率λ降采样得到的。
作为进一步优选的,对当前尺度的偏差场bj以及复原图像ij进行估计,具体包括:
首先,建立优化求解模型:
Figure PCTCN2016076616-appb-000009
其中,
Figure PCTCN2016076616-appb-000010
为复原图像估计值,
Figure PCTCN2016076616-appb-000011
为偏差场估计值,
Figure PCTCN2016076616-appb-000012
为数据逼近项,
Figure PCTCN2016076616-appb-000013
为数据约束项,α和β为数据约束项系数(也称惩罚系数),α和β的取值可根据实验需求进行调整;
继而,将优化求解模型分解成两个子问题,分别为i-子问题以及b-子问题,
其中,i-子问题的具体形式为:
Figure PCTCN2016076616-appb-000014
其中,
Figure PCTCN2016076616-appb-000015
分别为图像在竖直方向(x轴)和水平方向(y轴)上梯度的数据约束项;
b-子问题的具体形式为:
Figure PCTCN2016076616-appb-000016
最后,分别对i-子问题以及b-子问题求得最优解,即得到偏差场估计值及复原图像估计值。
作为进一步优选的,在对i-子问题进行求解时引入Split Bregman迭代方法。
作为进一步优选的,在步骤(3)中,根据所述降采样比率λ对偏差场估计值及复原图像估计值进行上采样。
另外,按照本发明的另一个方面,提出了一种气动光学热辐射噪声校正系统,包括:
多尺度处理模块包括降采样单元和上采样单元,其中,降采样单元用于对退化图像g0进行预处理,通过双线性插值法对退化图像g0进行等比率 λ降采样得到多尺度退化图像集合{g0,g1,…,gs},上采样单元用于将偏差场估计模块输出的当前尺度的偏差场估计值及复原图像估计值作上采样处理,并反馈至偏差场估计模块;
偏差场估计模块,用于根据多尺度退化图像集合中最小尺度的退化图像,获取当前尺度的偏差场估计值及复原图像估计值,并更新迭代,直至得到原始尺度的偏差场估计值
Figure PCTCN2016076616-appb-000017
并将
Figure PCTCN2016076616-appb-000018
输出至图像复原模块;
图像复原模块,用于根据所述原始尺度的偏差场估计值
Figure PCTCN2016076616-appb-000019
对退化图像g0进行复原,得到气动光学热辐射噪声校正图像i0
作为进一步优选的,所述偏差场估计模块包括图像选取单元和迭代处理单元,所述图像选取单元,用于选取所述集合中最小尺度的退化图像gs并作为迭代处理单元的初始化输入;所述迭代处理单元,用于通过更新迭代获取当前尺度的偏差场估计值及复原图像估计值,直至得到原始尺度的偏差场估计值
Figure PCTCN2016076616-appb-000020
并将
Figure PCTCN2016076616-appb-000021
输出至图像复原模块;
作为进一步优选的,所述多尺度退化图像集合是所述多尺度处理模块通过双线性插值法对退化图像g0进行等比率λ降采样得到的。
作为进一步优选的,所述上采样单元根据所述降采样比率λ对当前尺度的偏差场估计值及复原图像估计值进行上采样。
总体而言,通过本发明所构思的以上技术方案,与现有技术相比,主要具备以下的技术优点:
1、本发明技术方案仅通过光学探测系统获取的退化图像本身即可对气动光学热辐射噪声进行校正,并且能得到较为准确理想的图像复原效果;
2、尤其针对当前尺度偏差场估计时,本发明通过多尺度偏差场估计方法,利用上一尺度的估计结果作为下一尺度估计的输入值,可以使每一次估计更快地得到最优解,显著地提升了算法的收敛速度;
3、本发明中通过Split Bregman迭代进行优化求解,进一步提升了算 法的收敛速度,且更加简洁清楚,因此本发明尤其适用于气动光学热辐射噪声校正领域。
[附图说明]
图1是本发明的气动光学热辐射噪声校正方法总流程图;
图2是通过迭代优化求解的流程图;
图3是本发明的气动光学热辐射噪声校正系统结构图;
图4是本发明气动光学热辐射噪声校正效果示意图。
[具体实施方式]
为了使本发明的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本发明进行进一步详细说明。应当理解,此处所描述的具体实施例仅仅用以解释本发明,并不用于限定本发明。此外,下面所描述的本发明各个实施方式中所涉及到的技术特征只要彼此之间未构成冲突就可以相互组合。
以下首先对本发明用到的术语进行解释和说明。
偏差场:偏差场也被称为非均匀场,表现为亮度的不均匀性,偏差场的概念在医学图像处理中被广泛应用。亮度不均匀是指图像中同一区域像素表现出亮度的变化,这种图像亮度的空间变化由伪影和组织属性内在的不均匀性造成。由偏差场带来的强度或者颜色上的不均匀对许多图像处理算法是个很难解决的问题,例如图像分割、图像检索、目标跟踪与识别等等,都会面临这个问题。比如图像进行分割,通常使用目标图像是分段常数这个假设,如果图像受到了偏差场的影响,图像的强度就会发生渐变,之前的那个假设就用不上了,很多分割算法也不再有效。因为气动光学热辐射噪声与偏差场的性质极其相似,于是本发明将偏差场的概念引入到气动光学热辐射噪声校正中,将气动光学热辐射噪声退化图像g拟合为一个清晰图像i与一个偏差场b相加的模型,即g=i+b。通过得到偏差场的估计值
Figure PCTCN2016076616-appb-000022
后,经过
Figure PCTCN2016076616-appb-000023
即可得到复原图像。
多尺度:多尺度方法通过图像在尺度较小的时候其运算速度较快,得到的估计值经过上采样后作为下一尺度的初始值,这样可以使能量函数更快地收敛到最小值,得到最优解。
如图1所示,为本发明气动光学热辐射噪声校正方法的总流程图,本发明方法具体包括以下步骤:
(1)将退化图像g0进行预处理,通过双线性插值法对退化图像g0进行等比率λ的降采样,得到多尺度退化图像集合{g0,g1,…,gs};
(2)选取所述集合中最小尺度退化图像gs作为当前尺度退化图像gj,对当前尺度退化图像的偏差场bj以及复原图像ij进行估计,得到对应的偏差场估计值
Figure PCTCN2016076616-appb-000024
及复原图像估计值
Figure PCTCN2016076616-appb-000025
如图2所示对于目标函数的优化求解过程。为了能够解决1-范数在优化求解时不连续而导致的求解困难的问题,将引入Bregman变量,在Split Bregman迭代的求解下得到相应的最优解,再通过最小二乘的方法对2-范数进行优化求解,最后得到当前尺度偏差场以及复原图像的估计值,具体方法如下:
根据模型g=i+b可以建立优化求解模型,如公式(1)所示:
Figure PCTCN2016076616-appb-000026
其中,α和β为惩罚系数,其大小分别决定了优化过程中偏差场的光滑性以及对复原图像结构细节信息的保留,α取值越大,估计出的偏差场就越光滑,β取值越大,估计出的复原图像的结构细节信息保留的就越多。α和β的取值可根据实验需求进行调整。式(1)中偏差场估计值
Figure PCTCN2016076616-appb-000027
以及复原图像估计值
Figure PCTCN2016076616-appb-000028
都是需要优化的变量,因此,对其进行优化求解,即固定其中一个变量,对另一变量优化求解,交替进行,所以式(1)就 被分解成了两个子问题,分别为“i-子问题”以及“b-子问题”。
“i-子问题”的具体形式为
Figure PCTCN2016076616-appb-000029
其中,
Figure PCTCN2016076616-appb-000030
Figure PCTCN2016076616-appb-000031
分别为图像在竖直方向(x轴)和水平方向(y轴)上梯度的数据约束项,式(2)中的两个不连续项
Figure PCTCN2016076616-appb-000032
通常是求解的难点,本发明采用Split Bregman迭代进行优化求解。首先引入辅助变量d将式(2)转化成一个约束性问题,如公式(3)所示:
Figure PCTCN2016076616-appb-000033
再引入数据逼近项将式(3)转化为无约束问题,如公式(4)所示:
Figure PCTCN2016076616-appb-000034
其中,μ应设为足够大,这样才能保证式(3)与式(4)在转化时的等价性。μ的取值可根据实验需求进行调整。最后引入Bregman变量r后,得到公式(5):
Figure PCTCN2016076616-appb-000035
式(5)可以通过对变量i,d,r的迭代更新进行优化求解,具体求解过程如公式(6)下:
Figure PCTCN2016076616-appb-000036
通过对式(6)的迭代,即可得到式(2)的最优解。
“b-子问题”的具体形式如公式(7)所示:
Figure PCTCN2016076616-appb-000037
式(7)可以直接通过最小二乘求解的方法求得最优解。
(3)根据步骤(1)中的集合对当前尺度的偏差场估计值
Figure PCTCN2016076616-appb-000038
及复原图像估计值
Figure PCTCN2016076616-appb-000039
进行相邻尺度上采样,即令j=j-1,得到进行下一次迭代的当前尺度gj-1的bj-1及ij-1,对当前尺度的偏差场以及复原图像进行估计,继而得到对应的偏差场估计值
Figure PCTCN2016076616-appb-000040
及复原图像估计值
Figure PCTCN2016076616-appb-000041
重复步骤(3),直至得到原始尺度的偏差场估计值
Figure PCTCN2016076616-appb-000042
(4)根据所述原始尺度的偏差场估计值
Figure PCTCN2016076616-appb-000043
对退化图像g0进行复原,得到气动光学热辐射噪声校正图像i0
相应地,图3为与图1对应的系统结构图,如图3所示,一种气动光学热辐射噪声校正系统,包括:
多尺度处理模块分为图像降采样单元和图像上采样单元,降采样单元用于对退化图像g0进行预处理,通过双线性插值法对退化图像g0进行等比率λ降采样得到多尺度退化图像集合{g0,g1,…,gs},上采样单元用于将偏差场估计模块输出的当前尺度偏差场估计值及复原图像估计值作上采样处理,并反馈至偏差场估计模块;所述上采样单元根据所述降采样比率λ对当前尺度偏差场估计值及复原图像估计值进行上采样。
偏差场估计模块,用于根据多尺度退化图像集合中最小尺度的退化图像,获取当前尺度的偏差场估计值及复原图像估计值,并更新迭代,直至得到原始尺度的偏差场估计值
Figure PCTCN2016076616-appb-000044
并将
Figure PCTCN2016076616-appb-000045
输出至图像复原模块;
所述偏差场估计模块具体包括图像选取单元和迭代处理单元,所述图像选取单元,用于选取所述集合中最小尺度的退化图像gs并作为迭代处理 单元的初始化输入;所述迭代处理单元,用于通过更新迭代获取当前尺度的偏差场估计值及复原图像估计值,直至得到原始尺度的偏差场估计值
Figure PCTCN2016076616-appb-000046
并将
Figure PCTCN2016076616-appb-000047
输出至图像复原模块。
图像复原模块与偏差场估计模块相连,用于根据所述原始尺度的偏差场估计值
Figure PCTCN2016076616-appb-000048
对退化图像g0进行复原,得到气动光学热辐射噪声校正图像i0
图4为气动光学热辐射噪声校正效果示意图,从图4可以看出,通过执行本发明中的气动光学热辐射噪声校正方法,即针对退化图像本身先进行降采样,并通过多尺度偏差场估计,得到原始尺度的偏差场估计值,从而最终对退化图像进行复原。对比图4中的复原图像与原图像,可知本发明技术方案可得到较为准确理想的图像复原效果,且在保证图像复原更加准确的同时,显著提升了算法收敛速度,因而尤其适用于气动光学热辐射噪声校正领域。
本领域的技术人员容易理解,以上所述仅为本发明的较佳实施例而已,并不用以限制本发明,凡在本发明的精神和原则之内所作的任何修改、等同替换和改进等,均应包含在本发明的保护范围之内。

Claims (9)

  1. 一种气动光学热辐射噪声校正方法,其特征在于,包括以下步骤:
    (1)对退化图像g0进行预处理,得到尺度由大到小依次排序的多尺度退化图像集合{g0,g1,…,gs};
    (2)选取所述集合中最小尺度退化图像gs作为当前尺度退化图像gj,对当前尺度的偏差场以及复原图像进行估计,得到对应的偏差场估计值
    Figure PCTCN2016076616-appb-100001
    及复原图像估计值
    Figure PCTCN2016076616-appb-100002
    (3)根据步骤(1)中的集合对当前尺度的偏差场估计值
    Figure PCTCN2016076616-appb-100003
    及复原图像估计值
    Figure PCTCN2016076616-appb-100004
    进行相邻尺度上采样,得到进行下一次迭代的当前尺度退化图像gj-1的偏差场bj-1及复原图象ij-1,对当前尺度的偏差场以及复原图像进行估计,继而得到对应的偏差场估计值
    Figure PCTCN2016076616-appb-100005
    及复原图像估计值
    Figure PCTCN2016076616-appb-100006
    令j=j-1,重复步骤(3),直至得到原始尺度的偏差场估计值
    Figure PCTCN2016076616-appb-100007
    (4)根据所述原始尺度的偏差场估计值
    Figure PCTCN2016076616-appb-100008
    对退化图像g0进行复原,得到气动光学热辐射噪声校正图像i0
  2. 如权利要求1所述的方法,其特征在于在步骤(1)中,所述多尺度退化图像集合是通过双线性插值法对退化图像g0进行等比率λ降采样得到的。
  3. 如权利要求1所述的方法,其特征在于,对当前尺度的偏差场以及复原图像进行估计,具体包括以下步骤:
    首先,建立优化求解模型:
    Figure PCTCN2016076616-appb-100009
    其中,
    Figure PCTCN2016076616-appb-100010
    为复原图像估计值,
    Figure PCTCN2016076616-appb-100011
    为偏差场估计值,
    Figure PCTCN2016076616-appb-100012
    为数据逼近项,
    Figure PCTCN2016076616-appb-100013
    为数据约束项,α和β为惩罚系数,α和β的取值可根据实验需求进行调整;
    继而,将所述优化求解模型分解成两个子问题,分别为i-子问题以及b-子问题,
    其中,i-子问题的具体形式为:
    Figure PCTCN2016076616-appb-100014
    其中,
    Figure PCTCN2016076616-appb-100015
    分别为图像在竖直方向(x轴)和水平方向(y轴)上梯度的数据约束项;
    b-子问题的具体形式为:
    Figure PCTCN2016076616-appb-100016
    最后,分别对i-子问题以及b-子问题求得最优解,即得到偏差场估计值及复原图像估计值。
  4. 如权利要求3所述的方法,其特征在于,在对i-子问题进行求解时引入Split Bregman迭代方法。
  5. 如权利要求1所述的方法,其特征在于,在步骤(3)中,根据所 述等降采样比率λ对偏差场估计值及复原图像估计值进行上采样。
  6. 一种气动光学热辐射噪声校正系统,其特征在于,所述系统包括:
    多尺度处理模块包括降采样单元和上采样单元,其中,降采样单元用于对退化图像g0进行预处理,得到尺度由大到小依次排序的多尺度退化图像集合{g0,g1,…,gs};上采样单元用于偏差场估计模块输出的当前尺度的偏差场估计值及复原图像估计值作上采样处理,并反馈至偏差场估计模块;
    偏差场估计模块,用于根据多尺度退化图像集合中最小尺度的退化图像,获取当前尺度的偏差场估计值及复原图像估计值,并更新迭代,直至得到原始尺度的偏差场估计值
    Figure PCTCN2016076616-appb-100017
    并将
    Figure PCTCN2016076616-appb-100018
    输出至图像复原模块;
    图像复原模块,用于根据所述原始尺度的偏差场估计值
    Figure PCTCN2016076616-appb-100019
    对退化图像g0进行复原,得到气动光学热辐射噪声校正图像i0
  7. 如权利要求6所述的系统,其特征在于,所述偏差场估计模块包括图像选取单元和迭代处理单元,所述图像选取单元,用于选取所述集合中最小尺度的退化图像gs并作为迭代处理单元的初始化输入;所述迭代处理单元,用于通过更新迭代获取当前尺度的偏差场估计值及复原图像估计值,直至得到原始尺度的偏差场估计值
    Figure PCTCN2016076616-appb-100020
    并将
    Figure PCTCN2016076616-appb-100021
    输出至图像复原模块。
  8. 如权利要求6所述的系统,其特征在于,所述多尺度退化图像集合是所述多尺度处理模块通过双线性插值法对退化图像g0进行等比率λ降采样得到的。
  9. 如权利要求6所述的系统,其特征在于,所述上采样单元根据所述降采样比率λ对当前尺度的偏差场估计值及复原图像估计值进行上采样。
PCT/CN2016/076616 2015-10-21 2016-03-17 一种气动光学热辐射噪声校正方法与系统 WO2017067130A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US15/126,271 US9904987B2 (en) 2015-10-21 2016-03-17 Method and system for correcting aero-optical thermal radiation noise

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201510689097.6 2015-10-21
CN201510689097.6A CN105160687B (zh) 2015-10-21 2015-10-21 一种气动光学热辐射噪声校正方法与系统

Publications (1)

Publication Number Publication Date
WO2017067130A1 true WO2017067130A1 (zh) 2017-04-27

Family

ID=54801529

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/076616 WO2017067130A1 (zh) 2015-10-21 2016-03-17 一种气动光学热辐射噪声校正方法与系统

Country Status (3)

Country Link
US (1) US9904987B2 (zh)
CN (1) CN105160687B (zh)
WO (1) WO2017067130A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109389651A (zh) * 2017-08-09 2019-02-26 深圳先进技术研究院 磁共振化学位移编码成像方法与装置
CN109579844A (zh) * 2018-12-04 2019-04-05 电子科技大学 定位方法及系统

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105160687B (zh) * 2015-10-21 2018-01-02 华中科技大学 一种气动光学热辐射噪声校正方法与系统
US10521888B2 (en) * 2015-12-23 2019-12-31 Huazhong University Of Science And Technology Aerothermal radiation effect frequency domain correction method
CN105654430B (zh) 2015-12-24 2017-04-19 华中科技大学 一种反差约束的气动热辐射校正方法
CN111583149B (zh) * 2020-05-08 2023-02-17 西安电子科技大学 一种基于l1-l0范数最小化的气动热辐射图像自动校正方法
CN113421202B (zh) * 2021-06-25 2022-09-02 浙江大华技术股份有限公司 一种红外图像低频非均匀性校正的方法、设备及存储介质
CN114022399B (zh) * 2022-01-07 2022-04-19 武汉工程大学 一种气动光学热辐射校正方法、电子设备及可读存储介质
CN114022398B (zh) * 2022-01-07 2022-05-10 武汉工程大学 基于格林函数双调和样条插值热辐射效应校正方法及装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002300384A (ja) * 2001-03-30 2002-10-11 Minolta Co Ltd 画像復元装置、画像復元方法、プログラム及び記録媒体
US20030039376A1 (en) * 2001-08-27 2003-02-27 John Stach Segmentation in digital watermarking
CN102081737A (zh) * 2010-12-30 2011-06-01 华中科技大学 一种气动热辐射图像多尺度建模方法及其应用
CN102360498A (zh) * 2011-10-27 2012-02-22 江苏省邮电规划设计院有限责任公司 图像超分辨率重建方法
CN103337057A (zh) * 2013-07-05 2013-10-02 西北工业大学 基于多尺度自相似的运动模糊图像盲复原方法
CN103942758A (zh) * 2014-04-04 2014-07-23 中国人民解放军国防科学技术大学 基于多尺度融合的暗通道先验图像去雾方法
CN105160687A (zh) * 2015-10-21 2015-12-16 华中科技大学 一种气动光学热辐射噪声校正方法与系统

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002300384A (ja) * 2001-03-30 2002-10-11 Minolta Co Ltd 画像復元装置、画像復元方法、プログラム及び記録媒体
US20030039376A1 (en) * 2001-08-27 2003-02-27 John Stach Segmentation in digital watermarking
CN102081737A (zh) * 2010-12-30 2011-06-01 华中科技大学 一种气动热辐射图像多尺度建模方法及其应用
CN102360498A (zh) * 2011-10-27 2012-02-22 江苏省邮电规划设计院有限责任公司 图像超分辨率重建方法
CN103337057A (zh) * 2013-07-05 2013-10-02 西北工业大学 基于多尺度自相似的运动模糊图像盲复原方法
CN103942758A (zh) * 2014-04-04 2014-07-23 中国人民解放军国防科学技术大学 基于多尺度融合的暗通道先验图像去雾方法
CN105160687A (zh) * 2015-10-21 2015-12-16 华中科技大学 一种气动光学热辐射噪声校正方法与系统

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109389651A (zh) * 2017-08-09 2019-02-26 深圳先进技术研究院 磁共振化学位移编码成像方法与装置
CN109389651B (zh) * 2017-08-09 2022-12-13 深圳先进技术研究院 磁共振化学位移编码成像方法与装置
CN109579844A (zh) * 2018-12-04 2019-04-05 电子科技大学 定位方法及系统
CN109579844B (zh) * 2018-12-04 2023-11-21 电子科技大学 定位方法及系统

Also Published As

Publication number Publication date
CN105160687B (zh) 2018-01-02
US20170278220A1 (en) 2017-09-28
CN105160687A (zh) 2015-12-16
US9904987B2 (en) 2018-02-27

Similar Documents

Publication Publication Date Title
WO2017067130A1 (zh) 一种气动光学热辐射噪声校正方法与系统
CN105654430B (zh) 一种反差约束的气动热辐射校正方法
WO2019085433A1 (zh) 一种基于全变分的图像复原方法及系统
CN109118446B (zh) 一种水下图像复原及去噪方法
CN111462012A (zh) 一种基于条件生成对抗网络的sar图像仿真方法
WO2017113540A1 (zh) 一种气动光学效应校正识别一体化实时处理的系统和方法
CN106803235B (zh) 基于各向异性时空域全变分非均匀性校正的方法
Zhao et al. Robust and effective mesh denoising using L0 sparse regularization
CN110717956A (zh) 一种有限角投影超像素引导的l0范数最优化重建方法
CN110796616A (zh) 基于分数阶微分算子的l0范数约束和自适应加权梯度的湍流退化图像恢复方法
Shu et al. Alternating minimization algorithm for hybrid regularized variational image dehazing
CN109636722B (zh) 一种基于稀疏表示的在线字典学习超分辨率重建的方法
Liu et al. Single nighttime image dehazing based on unified variational decomposition model and multi-scale contrast enhancement
Chen et al. Hyperspectral imagery denoising using a spatial-spectral domain mixing prior
CN114092353A (zh) 一种基于加权引导滤波的红外图像增强方法
CN113313702A (zh) 基于边界约束与颜色校正的航拍图像去雾方法
Marais et al. Proximal-gradient methods for poisson image reconstruction with bm3d-based regularization
CN108230249A (zh) 基于各向异性的l1范数全变分正则化非均匀性校正方法
CN111583149B (zh) 一种基于l1-l0范数最小化的气动热辐射图像自动校正方法
Wu et al. Image segmentation via Fischer-Burmeister total variation and thresholding
CN108470335B (zh) 一种基于脑源空间分割的多相关源扫描成像方法
CN108876750B (zh) 一种基于非局部相似性低秩矩阵的图像去除泊松噪声的方法
CN106815820B (zh) 一种红外图像条状噪声消除方法
WO2019183843A1 (zh) 基于帧间配准和自适应步长的红外图像非均匀性校正方法
CN111028241B (zh) 一种多尺度血管增强的水平集分割系统与方法

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 15126271

Country of ref document: US

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

Ref document number: 16856577

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: 16856577

Country of ref document: EP

Kind code of ref document: A1