CN104574301A - Terahertz image reconstruction method and system - Google Patents

Abstract

The invention relates to a terahertz image reconstruction method and system. The method comprises steps as follows: image denoising: denoising a received terahertz image signal though median filtering and average filtering; image reconstruction: performing continuous Fourier transformation on the terahertz image signal to acquire multiple displacement images of unknown scenes, performing periodic sampling on the displacement images to acquire low-resolution images, performing mixed discrete Fourier transformation on multiple frames of low-resolution images, acquiring an original scene frequency domain coefficient according to the relationship between the mixed discrete Fourier transformation of multiple frames of observation images and the continuous Fourier transformation of the unknown scenes, and then performing inverse Fourier transformation to acquire a reconstructed image; non-local filtering, edge processing and overlapping processing. According to the terahertz image reconstruction method and system, image reconstruction is mainly performed on the basis of Fourier transformation and inverse transformation, the terahertz image with the higher resolution is acquired, the imaging time is shortened, and the problem of the low image resolution is solved.

Description

A kind of Terahertz image reconstructing method and system

Technical field

The present invention relates to a kind of Terahertz image reconstructing method and system, particularly relate to a kind of Terahertz image reconstructing method based on frequency domain method and system.

Background technology

Along with the development of technology, terahertz imaging is applied to production and the life of people more and more.Terahertz emission imaging technique makes it have a good application prospect in safety inspection field because of distinctive spectral characteristic, this technology can implement undamaged detection to the hidden contraband goods of carrying of human body, potential danger can be found early, therefore can be widely used in safety inspection field.At present, be pulse or the continuous all more difficult acquisition high-definition picture of the non-Near-Field Radar Imaging of Terahertz.Obtain high-resolution and can use less scanning step, but can increase the scanning imagery time like this, and scanning step is owing to being subject to diffraction limit restriction, can not infinitely reduce, therefore, its image resolution ratio is low, and imaging effect is not good.

Summary of the invention

The technical matters that the present invention solves is: build a kind of Terahertz image reconstructing method and system, the Terahertz image resolution ratio overcoming prior art is low, the technical matters that imaging is not good.

Technical scheme of the present invention is: provide a kind of Terahertz image reconstructing method, step is as follows:

Image denoising: medium filtering and mean filter removing signal noise are adopted to the Terahertz picture signal received, the value of any in Terahertz image is replaced with the Mesophyticum in a field by described medium filtering, and described mean filter replaces original pixel value by Terahertz image by each neighborhood of pixels average;

Image Reconstruction: the displacement diagram picture that continuous fourier transform obtains several unknown scenes is carried out to Terahertz picture signal, periodic sampling is carried out to described displacement diagram picture and obtains low-resolution image, multiframe low-resolution image is carried out to the discrete Fourier transformation of mixing, observe the relation of image between the discrete Fourier transformation of mixing and the continuous Fourier of unknown scene change according to multiframe and obtain original scene frequency domain coefficient, then carry out the image that inverse Fourier transform obtains reconstructing;

Non local filtering: the estimated value of being tried to achieve pixel by the weighted mean value of total space territory pixel, obtains the similarity between two pixels, is then weighted on average to it;

Edge treated: adopt horizontal and vertical operator to carry out edge treated to the image after non-local filtering process;

Overlap-add procedure: superposed by the image of the image after edge treated with non local filtering process, carries out image sharpening by the millimeter-wave image after superposition and is finally processed image.

Further technical scheme of the present invention is: when carrying out discrete Fourier transformation conversion, comprise the registration to multiframe low-resolution image and estimation.

Further technical scheme of the present invention is: in acquisition displacement image process, carry out global displacement to image.

Further technical scheme of the present invention is: obtain multiframe low-resolution image to displacement image with different cycles sampling.

Further technical scheme of the present invention is: in non local filter step, comprises and determines search window, similarity window and filtering depth parameter.

Further technical scheme of the present invention is: in non local filter step, and the similarity between two pixels is according to the similar retrieval between gray scale vector.

Further technical scheme of the present invention is: described similarity window is to be with centered by noise pixel, the square field of fixed size.

Technical scheme of the present invention is: build a kind of Terahertz image reconstruction system, comprise image denoising module, Image Reconstruction module, non local filtration module, edge treated module, overlap-add procedure module, described image denoising module adopts medium filtering and mean filter removing signal noise to the Terahertz picture signal received, the value of any in Terahertz image is replaced with the Mesophyticum in a field by described medium filtering, and described mean filter replaces original pixel value by Terahertz image by each neighborhood of pixels average; Described Image Reconstruction module carries out to Terahertz picture signal the displacement diagram picture that continuous fourier transform obtains several unknown scenes, periodic sampling is carried out to described displacement diagram picture and obtains low-resolution image, multiframe low-resolution image is carried out to the discrete Fourier transformation of mixing, described Image Reconstruction module is observed the relation of image between the discrete Fourier transformation of mixing and the continuous Fourier of unknown scene change according to multiframe and is obtained original scene frequency domain coefficient, then carries out the image that inverse Fourier transform obtains reconstructing; Described non local filtration module tries to achieve the estimated value of pixel by the weighted mean value of total space territory pixel, obtains the similarity between two pixels, is then weighted on average to it; Described edge treated module adopts horizontal and vertical operator to carry out edge treated to the image after non-local filtering process, the image of image after edge treated with non local filtering process superposes by described overlap-add procedure module, the millimeter-wave image after superposition is carried out image sharpening and is finally processed image.

Further technical scheme of the present invention is: described non local filtration module comprises determines search window, similarity window and filtering depth parameter, and described similarity window is to be with centered by noise pixel, the square field of fixed size.

Further technical scheme of the present invention is: in described non local filtration module, and the similarity between two pixels is according to the similar retrieval between gray scale vector.

Technique effect of the present invention is: provide a kind of Terahertz image reconstructing method and system, comprise image denoising: medium filtering and mean filter removing signal noise are adopted to the Terahertz picture signal received, the value of any in Terahertz image is replaced with the Mesophyticum in a field by described medium filtering, and described mean filter replaces original pixel value by Terahertz image by each neighborhood of pixels average; Image Reconstruction: the displacement diagram picture that continuous fourier transform obtains several unknown scenes is carried out to Terahertz picture signal, periodic sampling is carried out to described displacement diagram picture and obtains low-resolution image, multiframe low-resolution image is carried out to the discrete Fourier transformation of mixing, observe the relation of image between the discrete Fourier transformation of mixing and the continuous Fourier of unknown scene change according to multiframe and obtain original scene frequency domain coefficient, then carry out the image that inverse Fourier transform obtains reconstructing; Non local filtering: the estimated value of being tried to achieve pixel by the weighted mean value of total space territory pixel, obtains the similarity between two pixels, is then weighted on average to it; Edge treated: adopt horizontal and vertical operator to carry out edge treated to the image after non-local filtering process; Overlap-add procedure: superposed by the image of the image after edge treated with non local filtering process, carries out image sharpening by the millimeter-wave image after superposition and is finally processed image.A kind of Terahertz image reconstructing method of the present invention and system, mainly bring the image restoration carried out, obtain more high-resolution Terahertz image, and shorten imaging time, solve the problem that image resolution ratio is low based on Fourier transform and contravariant.

Accompanying drawing explanation

Fig. 1 is process flow diagram of the present invention.

Fig. 2 is Image Reconstruction process flow diagram of the present invention.

Fig. 3 is structural representation of the present invention.

Embodiment

Below in conjunction with specific embodiment, technical solution of the present invention is further illustrated.

As shown in Figure 1, the specific embodiment of the present invention is: provide a kind of Terahertz image reconstructing method, step is as follows:

Image denoising: medium filtering and mean filter removing signal noise are adopted to the Terahertz picture signal received, the value of any in Terahertz image is replaced with the Mesophyticum in a field by described medium filtering, and described mean filter replaces original pixel value by Terahertz image by each neighborhood of pixels average.

Specific implementation process is as follows: first eliminate isolated noise with medium filtering, then use mean filter smoothed image noise.

Medium filtering: it is a kind of conventional nonlinear smoothing filtering, replaces the value of any in digital picture with the Mesophyticum of each point value in a field of this point.Two dimension median filter exports:

f(x,y)＝med{p(x-k,y-l)},(k,l∈W) (1)

P (x, y), f (x, y) are respectively original image and filtered image, and W is two dimension pattern plate.

Mean filter: each neighborhood of pixels average of mean filter replaces original pixel value,

$f(x,y)=\mathrm{\Σ}\frac{p(x-k,y-l)}{n},(k,1\∈W)---\left(2\right)$

N is the total number of two dimension pattern plate pixel.

Image Reconstruction: the displacement diagram picture that continuous fourier transform obtains several unknown scenes is carried out to Terahertz picture signal, periodic sampling is carried out to described displacement diagram picture and obtains low-resolution image, multiframe low-resolution image is carried out to the discrete Fourier transformation of mixing, observe the relation of image between the discrete Fourier transformation of mixing and the continuous Fourier of unknown scene change according to multiframe and obtain original scene frequency domain coefficient, then carry out the image that inverse Fourier transform obtains reconstructing.

As shown in Figure 2, specific implementation process is as follows: suppose that signal bandwidth is limited in original scene, if x is (t ₁, t ₂) be a secondary high-definition picture continuously, X (w ₁, w ₂) be its continuous fourier transform, only consider global displacement in frequency domain method, produce a kth displacement diagram as x _k(t ₁, t ₂)=x (t ₁+ μ _k1, t ₂+ μ _k2), wherein, μ _k1and μ _k2for any given value, k=1,2 ..., p, by the displacement property of continuous fourier transform, displacement diagram is as X _k(w ₁, w ₂) continuous fourier transform be:

X _k(w ₁,w ₂)＝exp[j2π(μ _k1w ₁+μ _k2w ₂)]X(w ₁,w ₂) (3)

To displacement image x _k(t ₁, t ₂) respectively with T ₁and T ₂for periodic sampling, obtain low-resolution image y _k[n ₁, n ₂].According to spectral aliasing character, and X (w ₁, w ₂) broadband finiteness, the continuous fourier transform of high-definition picture and kth observe the pass between the discrete Fourier transformation of LR image be:

$Y_k[w_1,w_2]=\frac{1}{T_1{T}_2}\underset{n_1=0}{\overset{L_1-1}{\mathrm{\Σ}}}\underset{n_2=0}{\overset{L_2-1}{\mathrm{\Σ}}}{X}_k\×(\frac{2\mathrm{\π}}{T_1}(\frac{w_1}{T_1}+n_1)\·\frac{2\mathrm{\π}}{T_2}\left(\frac{w_2}{T_2}{+n}_2\right))---\left(4\right)$

Above formula is expressed in matrix as: Y=Ф X.Wherein, Y is the column vector of p × 1, and a kth element is y _k[n ₁, n ₂] DFT coefficient, X is a L ₁l ₂the column vector of × 1 is x (t ₁, t ₂) continuous fourier transform coefficient, Ф is a p × L ₁l ₂matrix.Like this, multiframe is observed the discrete Fourier transformation of image through mixing and the continuous fourier transform of unknown scene, connect with the form of system of equations, obtained the frequency domain coefficient of original scene by solving equations.

Inverse Fourier transform is also the usual algorithm in Fourier transform:

$f\left(t\right)=\frac{1}{2\mathrm{\π}}{\∫}_{-\∞}^{\∞}F\left(\mathrm{\ω}\right)e^{\mathrm{jet}}\mathrm{d\ω}$

The recovery that inverse Fourier transform obtains original scene is carried out according to the frequency domain coefficient of original scene.

Eliminate spectral aliasing by frequency field and improve the spatial resolution of image, because the details of image shows by high-frequency information, and by eliminating spectral aliasing, just can obtain the high-frequency information be more submerged, therefore relying on frequency-domain solution spectral aliasing is exactly the details increasing image, improves resolution.

Non local filtering: the estimated value of being tried to achieve pixel by the weighted mean value of total space territory pixel, obtains the similarity between two pixels, is then weighted on average it.

Specific implementation process is as follows: refer to that the gray-scale value of current pixel point is obtained by the gray-scale value weighted mean of the total space territory pixel similar to its structure, weight depends on structural similarity degree.Suppose given discrete by digital picture v={v (the i) ∣ i ∈ I} of noise pollution, can be tried to achieve by the weighted mean of total space territory pixel the estimated value NL [v] (i) of pixel i:

NL[v](i)＝Σw(i，j)v(j) (5)

Weight { w (i, j) } _jdepend on the similarity of pixel i and j, and meet:

0≤w(i，j)≤1；

Σ _jw(i，j)＝1. (6)

Similarity between two pixel i and j depends on gray scale vector v (N _i) and v (N _j) between similarity.N _krepresent the square field being centrally located at the fixed size of k.This similarity is by weighted euclidean distance ‖ v (N _i)-v (N _j) ‖ ² _{2, a}decreasing function represent.Wherein a is the standard deviation of gaussian kernel.Euclidean distance expectation value between the noise pixel point of image can be tried to achieve by following formula:

$E{\left|\right|v\left(N_i\right)-v\left(N_j\right)\left|\right|}_{2,a}^2={\left|\right|u\left(N_i\right)-u\left(N_j\right)\left|\right|}_{2,a}^2+{2\mathrm{\σ}}^2:---\left(7\right)$

The pass of v and u is: v=u+n, v are image pixel observed readings, and u is image actual value, and n is the noise of superposition.σ is the standard deviation of the spacing of two gray scale vectors.The expectation of this Euclidean distance maintains the similarity between different pixels point.With v (N _i) pixel in similar gray scale field has larger weight generally, defined by following formula:

$w(i,j)=\frac{1}{Z\left(i\right)}{e}^{-\frac{{\left|\right|v\left(N_i\right)-v\left(N_j\right)\left|\right|}_{2,a}^2}{h^2}}---\left(8\right)$

Normaliztion constant factor Z (i) is defined as:

$Z\left(i\right)={\underset{j}{\mathrm{\Σ}}e}^{-\frac{{\left|\right|v\left(N_i\right)-v\left(N_j\right)\left|\right|}_{2,a}^2}{h^2}}---\left(9\right)$

Wherein h represents filter strength, the decay of control characteristic function, or the rate of decay of the further control weight factor.

Be generally convenience of calculation, N _iget centered by pixel i, the square field of fixed size (2m+1) × (2m+1), w (i, j) and Z (i) can be expressed as:

$w(i,j)=\frac{1}{G\left(i\right)}\exp \left[\frac{\underset{n_i\∈N_i,n_j\∈N_j,k_i\∈k}{\mathrm{\Σ}}{k}_i{(n_i-n_j)}^2}{h^2}\right]---\left(10\right)$

$Z\left(i\right)=\underset{j}{\mathrm{\Σ}}\exp \left[\frac{\underset{n_i\∈N_i,n_j\∈N_j,k_i\∈k}{\mathrm{\Σ}}{k}_i{(n_i-n_j)}^2}{h^2}\right]---\left(11\right)$

$k_i=\frac{1}{m}\underset{{d=d}_i}{\overset{m}{\mathrm{\Σ}}}\frac{1}{{(2d+1)}^2}---\left(12\right)$

The geometry in what non local filtering was compared the is whole field of two single-points, so have more repellence to noise, and the part leached contains less geometry information.

Edge treated: adopt Roberts edge detection operator to realize horizontal and vertical direction and edge treated is carried out to the image after non-local filtering process.If f (x, y) is gradation of image distribution function, then its Reberts edge detection operator is

$g(x,y)={\{{[\sqrt{f(x,y)}-\sqrt{f(x+1,y+1)}]}^2+{[\sqrt{f(x+1,y)}-\sqrt{f(x,y+1)}]}^2\}}^{\frac{1}{2}}$

Roberts edge detection operator realizes the rim detection in horizontal and vertical direction respectively, and operational form is:

$\left\{\begin{array}{c}{\mathrm{\Δ}}_{x}f(x,y)=f(x,y)-f(x-1,y-1)\\ {\mathrm{\Δ}}_{y}f(x,y)=f(x-1,y)-f(x,y-1)\end{array}\right.$

Overlap-add procedure: the image of the image after edge treated with second order high-pass filtering process superposes by the image registration based on half-tone information method, concrete grammar is definition benchmark image I (x, y) with template image T (x, y), make template image move on benchmark image, and calculate similarity degree between the two, namely the place that peak value occurs is registration position, calculating formula of similarity on each displacement point (i, j) determined is

$D(i,j)=\frac{{\mathrm{\Σ}}_x{\mathrm{\Σ}}_{y}T(x,y)I(x-i,y-j)}{\sqrt{{\mathrm{\Σ}}_x{\mathrm{\Σ}}_y{I}^2(x-i,y-j)}}$

Image sharpening: the millimeter-wave image after superposition is carried out image sharpening, and utilize Roberts operator to carry out sharpening, Roberts operator template is the template of a 2*2, and for current pending pixel f (x, y), Roberts operator definitions is as follows:

$\▿f=|f(x+1.y+1)-f(x,y)|+|f(x+1,y)-f(x,y+1)|$

Specifically being expressed as of template

$D_1=\left[\begin{array}{cc}-1& 0\\ 0& 1\end{array}\right]D_2=\left[\begin{array}{cc}0& -1\\ 1& 0\end{array}\right]$

ξ ₁＝D ₁(f(x,y)) ξ ₂＝D ₂(f(x,y))

$\▿f(x,y)=\left|{\mathrm{\ξ}}_1\right|+\left|{\mathrm{\ξ}}_2\right|$

Finally processed image.

The preferred embodiment of the present invention is: when carrying out discrete Fourier transformation conversion, comprise the registration to multiframe low-resolution image and estimation.For multiframe low-resolution image, first carry out registration by multipoint positioning in use, carry out follow-up Fourier transform by after multiframe low-resolution image, to make the better effects if of reconstructed image.In acquisition displacement image process, further method needs to consider estimation, by carrying out estimation, can obtain better displacement diagram picture.

As shown in Figure 3, the specific embodiment of the present invention is: build a kind of Terahertz image reconstruction system, comprise image denoising module 1, Image Reconstruction module 2, non local filtration module 3, edge treated module 4, overlap-add procedure module 5, described image denoising module 1 adopts medium filtering and mean filter removing signal noise to the Terahertz picture signal received, the value of any in Terahertz image is replaced with the Mesophyticum in a field by described medium filtering, and described mean filter replaces original pixel value by Terahertz image by each neighborhood of pixels average; Described Image Reconstruction module 2 pairs of Terahertz picture signals carry out the displacement diagram picture that continuous fourier transform obtains several unknown scenes, periodic sampling is carried out to described displacement diagram picture and obtains low-resolution image, multiframe low-resolution image is carried out to the discrete Fourier transformation of mixing, described Image Reconstruction module 2 is observed the relation of image between the discrete Fourier transformation of mixing and the continuous Fourier of unknown scene change according to multiframe and is obtained original scene frequency domain coefficient, then carries out the image that inverse Fourier transform obtains reconstructing; Described non local filtration module 3 tries to achieve the estimated value of pixel by the weighted mean value of total space territory pixel, obtains the similarity between two pixels, is then weighted on average to it; Described edge treated module 4 adopts horizontal and vertical operator to carry out edge treated to the image after non-local filtering process, the image of image after edge treated with non local filtering process superposes by described overlap-add procedure module 5, the millimeter-wave image after superposition is carried out image sharpening and is finally processed image.

As shown in Figure 3, specific embodiment of the invention process is: image denoising module 1 adopts medium filtering and mean filter removing signal noise to the Terahertz picture signal received, the value of any in Terahertz image is replaced with the Mesophyticum in a field by described medium filtering, and described mean filter replaces original pixel value by Terahertz image by each neighborhood of pixels average.

Specific implementation process is as follows: first eliminate isolated noise with medium filtering, then use mean filter smoothed image noise.

Medium filtering: it is a kind of conventional nonlinear smoothing filtering, this point of the value of any in digital picture

A field in the Mesophyticum of each point value replace.Two dimension median filter exports:

f(x,y)＝med{p(x-k,y-l)},(k,l∈W) (1)

P (x, y), f (x, y) are respectively original image and filtered image, and W is two dimension pattern plate.

Mean filter: each neighborhood of pixels average of mean filter replaces original pixel value,

$f(x,y)=\mathrm{\Σ}\frac{p(x-k,y-l)}{n},(k,1\∈W)---\left(2\right)$

N is the total number of two dimension pattern plate pixel.

Image Reconstruction module 2 pairs of Terahertz picture signals carry out the displacement diagram picture that continuous fourier transform obtains several unknown scenes, periodic sampling is carried out to described displacement diagram picture and obtains low-resolution image, multiframe low-resolution image is carried out to the discrete Fourier transformation of mixing, observe the relation of image between the discrete Fourier transformation of mixing and the continuous Fourier of unknown scene change according to multiframe and obtain original scene frequency domain coefficient, Image Reconstruction module 2 carries out the image that inverse Fourier transform obtains reconstructing.

As shown in Figure 2, specific implementation process is as follows: suppose that signal bandwidth is limited in original scene, if x is (t ₁, t ₂) be a secondary high-definition picture continuously, X (w ₁, w ₂) be its continuous fourier transform, only consider global displacement in frequency domain method, produce a kth displacement diagram as x _k(t ₁, t ₂)=x (t ₁+ μ _k1, t ₂+ μ _k2), wherein, μ _k1and μ _k2for any given value, k=1,2 ..., p, by the displacement property of continuous fourier transform, displacement diagram is as X _k(w ₁, w ₂) continuous fourier transform be:

X _k(w ₁,w ₂)＝exp[j2π(μ _k1w ₁+μ _k2w ₂)]X(w ₁,w ₂) (3)

To displacement image x _k(t ₁, t ₂) respectively with T ₁and T ₂for periodic sampling, obtain low-resolution image y _k[n ₁, n ₂].According to spectral aliasing character, and X (w ₁, w ₂) broadband finiteness, the continuous fourier transform of high-definition picture and kth observe the pass between the discrete Fourier transformation of LR image be:

$Y_k[w_1,w_2]=\frac{1}{T_1{T}_2}\underset{n_1=0}{\overset{L_1-1}{\mathrm{\Σ}}}\underset{n_2=0}{\overset{L_2-1}{\mathrm{\Σ}}}{X}_k\×(\frac{2\mathrm{\π}}{T_1}(\frac{w_1}{T_1}+n_1)\·\frac{2\mathrm{\π}}{T_2}\left(\frac{w_2}{T_2}{+n}_2\right))---\left(4\right)$

Above formula is expressed in matrix as: Y=Ф X.Wherein, Y is the column vector of p × 1, and a kth element is y _k[n ₁, n ₂] DFT coefficient, X is a L ₁l ₂the column vector of × 1 is x (t ₁, t ₂) continuous fourier transform coefficient, Ф is a p × L ₁l ₂matrix.Like this, multiframe is observed the discrete Fourier transformation of image through mixing and the continuous fourier transform of unknown scene, connect with the form of system of equations, obtained the frequency domain coefficient of original scene by solving equations.

Inverse Fourier transform is also the usual algorithm in Fourier transform:

$f\left(t\right)=\frac{1}{2\mathrm{\π}}{\∫}_{-\∞}^{\∞}F\left(\mathrm{\ω}\right)e^{\mathrm{jet}}\mathrm{d\ω}$

The recovery that inverse Fourier transform obtains original scene is carried out according to the frequency domain coefficient of original scene.

Eliminate spectral aliasing by frequency field and improve the spatial resolution of image, because the details of image shows by high-frequency information, and by eliminating spectral aliasing, just can obtain the high-frequency information be more submerged, therefore relying on frequency-domain solution spectral aliasing is exactly the details increasing image, improves resolution.

Non local filtration module 3 tries to achieve the estimated value of pixel by the weighted mean value of total space territory pixel, obtains the similarity between two pixels, is then weighted on average it.

Specific implementation process is as follows: refer to that the gray-scale value of current pixel point is obtained by the gray-scale value weighted mean of the total space territory pixel similar to its structure, weight depends on structural similarity degree.Suppose given discrete by digital picture v={v (the i) ∣ i ∈ I} of noise pollution, can be tried to achieve by the weighted mean of total space territory pixel the estimated value NL [v] (i) of pixel i:

NL[v](i)＝Σw(i，j)v(j) (5)

Weight { w (i, j) } _jdepend on the similarity of pixel i and j, and meet:

0≤w(i，j)≤1；

Σ _jw(i，j)＝1. (6)

Similarity between two pixel i and j depends on gray scale vector v (N _i) and v (N _j) between similarity.N _krepresent the square field being centrally located at the fixed size of k.This similarity is by weighted euclidean distance ‖ v (N _i)-v (N _j) ‖ ² _{2, a}decreasing function represent.Wherein a is the standard deviation of gaussian kernel.Euclidean distance expectation value between the noise pixel point of image can be tried to achieve by following formula:

$E{\left|\right|v\left(N_i\right)-v\left(N_j\right)\left|\right|}_{2,a}^2={\left|\right|u\left(N_i\right)-u\left(N_j\right)\left|\right|}_{2,a}^2+{2\mathrm{\σ}}^2:---\left(7\right)$

The pass of v and u is: v=u+n, v are image pixel observed readings, and u is image actual value, and n is the noise of superposition.σ is the standard deviation of the spacing of two gray scale vectors.The expectation of this Euclidean distance maintains the similarity between different pixels point.With v (N _i) pixel in similar gray scale field has larger weight generally, defined by following formula:

$w(i,j)=\frac{1}{Z\left(i\right)}{e}^{-\frac{{\left|\right|v\left(N_i\right)-v\left(N_j\right)\left|\right|}_{2,a}^2}{h^2}}---\left(8\right)$

Normaliztion constant factor Z (i) is defined as:

$Z\left(i\right)={\underset{j}{\mathrm{\Σ}}e}^{-\frac{{\left|\right|v\left(N_i\right)-v\left(N_j\right)\left|\right|}_{2,a}^2}{h^2}}---\left(9\right)$

Wherein h represents filter strength, the decay of control characteristic function, or the rate of decay of the further control weight factor.

Be generally convenience of calculation, N _iget centered by pixel i, the square field of fixed size (2m+1) × (2m+1), w (i, j) and Z (i) can be expressed as:

$w(i,j)=\frac{1}{G\left(i\right)}\exp \left[\frac{\underset{n_i\∈N_i,n_j\∈N_j,k_i\∈k}{\mathrm{\Σ}}{k}_i{(n_i-n_j)}^2}{h^2}\right]---\left(10\right)$

$Z\left(i\right)=\underset{j}{\mathrm{\Σ}}\exp \left[\frac{\underset{n_i\∈N_i,n_j\∈N_j,k_i\∈k}{\mathrm{\Σ}}{k}_i{(n_i-n_j)}^2}{h^2}\right]---\left(11\right)$

$k_i=\frac{1}{m}\underset{{d=d}_i}{\overset{m}{\mathrm{\Σ}}}\frac{1}{{(2d+1)}^2}---\left(12\right)$

The geometry in what non local filtering was compared the is whole field of two single-points, so have more repellence to noise, and the part leached contains less geometry information.

Edge treated module 4 adopts Roberts edge detection operator to realize horizontal and vertical direction to carry out edge treated to the image after non-local filtering process.If f (x, y) is gradation of image distribution function, then its Reberts edge detection operator is

$g(x,y)={\{{[\sqrt{f(x,y)}-\sqrt{f(x+1,y+1)}]}^2+{[\sqrt{f(x+1,y)}-\sqrt{f(x,y+1)}]}^2\}}^{\frac{1}{2}}$

Roberts edge detection operator realizes the rim detection in horizontal and vertical direction respectively, and operational form is:

$\left\{\begin{array}{c}{\mathrm{\Δ}}_{x}f(x,y)=f(x,y)-f(x-1,y-1)\\ {\mathrm{\Δ}}_{y}f(x,y)=f(x-1,y)-f(x,y-1)\end{array}\right.$

The image of image after edge treated with second order high-pass filtering process superposes based on the image registration of half-tone information method by overlap-add procedure module 5, concrete grammar is definition benchmark image I (x, y) with template image T (x, y), make template image move on benchmark image, and calculate similarity degree between the two, namely the place that peak value occurs is registration position, calculating formula of similarity on each displacement point (i, j) determined is

$D(i,j)=\frac{{\mathrm{\Σ}}_x{\mathrm{\Σ}}_{y}T(x,y)I(x-i,y-j)}{\sqrt{{\mathrm{\Σ}}_x{\mathrm{\Σ}}_y{I}^2(x-i,y-j)}}$

Image sharpening: the millimeter-wave image after superposition is carried out image sharpening, and utilize Roberts operator to carry out sharpening, Roberts operator template is the template of a 2*2, and for current pending pixel f (x, y), Roberts operator definitions is as follows:

$\▿f=|f(x+1.y+1)-f(x,y)|+|f(x+1,y)-f(x,y+1)|$

Specifically being expressed as of template

$D_1=\left[\begin{array}{cc}-1& 0\\ 0& 1\end{array}\right]D_2=\left[\begin{array}{cc}0& -1\\ 1& 0\end{array}\right]$

ξ ₁＝D ₁(f(x,y)) ξ ₂＝D ₂(f(x,y))

$\▿f(x,y)=\left|{\mathrm{\ξ}}_1\right|+\left|{\mathrm{\ξ}}_2\right|$

Finally processed image.

Technique effect of the present invention is: provide a kind of Terahertz image reconstructing method and system, comprise image denoising: medium filtering and mean filter removing signal noise are adopted to the Terahertz picture signal received, the value of any in Terahertz image is replaced with the Mesophyticum in a field by described medium filtering, and described mean filter replaces original pixel value by Terahertz image by each neighborhood of pixels average; Image Reconstruction: the displacement diagram picture that continuous fourier transform obtains several unknown scenes is carried out to Terahertz picture signal, periodic sampling is carried out to described displacement diagram picture and obtains low-resolution image, multiframe low-resolution image is carried out to the discrete Fourier transformation of mixing, observe the relation of image between the discrete Fourier transformation of mixing and the continuous Fourier of unknown scene change according to multiframe and obtain original scene frequency domain coefficient, then carry out the image that inverse Fourier transform obtains reconstructing; Non local filtering: the estimated value of being tried to achieve pixel by the weighted mean value of total space territory pixel, obtains the similarity between two pixels, is then weighted on average to it; Edge treated: adopt horizontal and vertical operator to carry out edge treated to the image after non-local filtering process; Overlap-add procedure: superposed by the image of the image after edge treated with non local filtering process, carries out image sharpening by the millimeter-wave image after superposition and is finally processed image.A kind of Terahertz image reconstructing method of the present invention and system, mainly bring the image restoration carried out, obtain more high-resolution Terahertz image, and shorten imaging time, solve the problem that image resolution ratio is low based on Fourier transform and contravariant.

Above content is in conjunction with concrete preferred implementation further description made for the present invention, can not assert that specific embodiment of the invention is confined to these explanations.For general technical staff of the technical field of the invention, without departing from the inventive concept of the premise, some simple deduction or replace can also be made, all should be considered as belonging to protection scope of the present invention.

Claims (10)

1. a Terahertz image reconstructing method, step is as follows:

Image denoising: medium filtering and mean filter removing signal noise are adopted to the Terahertz picture signal received, the value of any in Terahertz image is replaced with the Mesophyticum in a field by described medium filtering, and described mean filter replaces original pixel value by Terahertz image by each neighborhood of pixels average;

Image Reconstruction: the displacement diagram picture that continuous fourier transform obtains several unknown scenes is carried out to Terahertz picture signal, periodic sampling is carried out to described displacement diagram picture and obtains low-resolution image, multiframe low-resolution image is carried out to the discrete Fourier transformation of mixing, observe the relation of image between the discrete Fourier transformation of mixing and the continuous Fourier of unknown scene change according to multiframe and obtain original scene frequency domain coefficient, then carry out the image that inverse Fourier transform obtains reconstructing;

Non local filtering: the estimated value of being tried to achieve pixel by the weighted mean value of total space territory pixel, obtains the similarity between two pixels, is then weighted on average to it;

Edge treated: adopt horizontal and vertical operator to carry out edge treated to the image after non-local filtering process;

Overlap-add procedure: superposed by the image of the image after edge treated with non local filtering process, carries out image sharpening by the millimeter-wave image after superposition and is finally processed image.

2. Terahertz image reconstructing method according to claim 1, is characterized in that, when carrying out discrete Fourier transformation conversion, comprises the registration to multiframe low-resolution image and estimation.

3. Terahertz image reconstructing method according to claim 1, is characterized in that, in acquisition displacement image process, carries out global displacement to image.

4. Terahertz image reconstructing method according to claim 1, is characterized in that, obtains multiframe low-resolution image to displacement image with different cycles sampling.

5. Terahertz image reconstructing method according to claim 1, is characterized in that, in non local filter step, comprise and determine search window, similarity window and filtering depth parameter.

6. Terahertz image reconstructing method according to claim 1, is characterized in that, in non local filter step, the similarity between two pixels is according to the similar retrieval between gray scale vector.

7. Terahertz image reconstructing method according to claim 5, is characterized in that, described similarity window is to be with centered by noise pixel, the square field of fixed size.

8. a Terahertz image reconstruction system, it is characterized in that, comprise image denoising module, Image Reconstruction module, non local filtration module, edge treated module, overlap-add procedure module, described image denoising module adopts medium filtering and mean filter removing signal noise to the Terahertz picture signal received, the value of any in Terahertz image is replaced with the Mesophyticum in a field by described medium filtering, and described mean filter replaces original pixel value by Terahertz image by each neighborhood of pixels average; Described Image Reconstruction module carries out to Terahertz picture signal the displacement diagram picture that continuous fourier transform obtains several unknown scenes, periodic sampling is carried out to described displacement diagram picture and obtains low-resolution image, multiframe low-resolution image is carried out to the discrete Fourier transformation of mixing, described Image Reconstruction module is observed the relation of image between the discrete Fourier transformation of mixing and the continuous Fourier of unknown scene change according to multiframe and is obtained original scene frequency domain coefficient, then carries out the image that inverse Fourier transform obtains reconstructing; Described non local filtration module tries to achieve the estimated value of pixel by the weighted mean value of total space territory pixel, obtains the similarity between two pixels, is then weighted on average to it; Described edge treated module adopts horizontal and vertical operator to carry out edge treated to the image after non-local filtering process, the image of image after edge treated with non local filtering process superposes by described overlap-add procedure module, the millimeter-wave image after superposition is carried out image sharpening and is finally processed image.

9. Terahertz image reconstruction system according to claim 8, it is characterized in that, described non local filtration module comprises determines search window, similarity window and filtering depth parameter, and described similarity window is to be with centered by noise pixel, the square field of fixed size.

10. Terahertz image reconstruction system according to claim 1, is characterized in that, in described non local filtration module, the similarity between two pixels is according to the similar retrieval between gray scale vector.