CN102800067A - Fuzzy self-modulation display enhancement method for ISAR (inverse synthetic aperture radar) images - Google Patents

Abstract

The invention relates to a fuzzy self-modulation display enhancement method for ISAR (inverse synthetic aperture radar) images, which mainly comprises the following steps: firstly, carrying out dynamic range estimation on an original ISAR image so as to estimate the currently-acquired dynamic range of the original ISAR image, and carrying out equalized histogram dynamic-compression on the original image by using parameters of the estimated dynamic range; then, carrying out Harr wavelet decomposition on the original ISAR image so as to obtain an original image, an image under three dimensions (one-layer and two-layer wavelet approximation coefficients) and one-layer and two-layer horizontal, vertical and diagonal high-frequency coefficients; respectively carrying out subjection degree coefficient estimation on one-layer and two-layer low-frequency coefficients obtained through carrying out Harr wavelet decomposition on the original image, calculating modulation coefficients, and then calculating edge details; and finally, carrying out Harr wavelet reconstruction by using obtained edge detail results, and then carrying out image enhancement modulation with a result obtained after dynamic compression, thereby obtaining a final enhancement result. The display effect of an ISAR image processed by using the algorithm is greatly strengthened, and the detail part of the ISAR image is obviously improved.

Description

A kind of fuzzy self-modulation of ISAR image shows Enhancement Method

One technical field

The invention belongs to a kind of radar image Enhanced Imaging disposal route, particularly a kind of to the ISAR image enhancement processing method.

Two background technologies

ISAR (Inverse Synthetic Aperture Radar is hereinafter to be referred as the ISAR picture) is exactly the other a kind of radar imagery technology that on the SAR basis, grows up, and is intended to solve the imaging problem of moving target.ISAR can obtain the precise image on non-cooperative moving targets (like aircraft, naval vessel etc.).

As far back as the beginning of the sixties in last century, just proposed the notion of ISAR imaging, and obtained the image of ISAR to the eighties.But there is the problem that dynamic range is big, details is fuzzy in contrary synthetic aperture image; Improving the ISAR image at present mainly is through improving the imaging algorithm aspect of ISAR image; Receive the restriction of inverse synthetic aperture radar imaging mechanism; Improve at present imaging algorithm and receive certain limitation, and the ISAR picture has dynamic range is big, details is fuzzy shortcoming itself, this is insurmountable through improving imaging algorithm.

Traditional passes through research apart from alignment, phase alignment algorithms to improve image quality, but can't change dynamic range, the details visuality of ISAR picture; The present invention utilizes image processing techniques, improves the visuality of ISAR picture through fuzzy membership function, morphologic filtering, the combined method of wavelet transformation.

Three summary of the invention

The main implementation method of this software is: strengthen three main processing flow processs through the ISAR image (ISAR picture) that receives being carried out dynamic range compression processing, index of modulation reconstruct and details.

At first original I SAR picture is carried out dynamic range and estimate, estimate the dynamic range of the current original I SAR picture that obtains, utilize estimated dynamic range parameters, original image is carried out balanced histogram dynamic compression.

Then original I SAR picture is carried out the Harr wavelet decomposition, obtain the high frequency coefficient that original image, one deck and two layers of small echo approach image and one deck and two layers level under 3 yardsticks of coefficient, vertical, diagonal angle; The one deck that respectively original image, Harr wavelet decomposition is obtained, two layers of low frequency coefficient carry out the degree of membership coefficient estimation, calculate the index of modulation, edge calculation details then.

Utilize the edge details result of gained to carry out the Harr wavelet reconstruction at last, then with dynamic compression after the result carry out the image enhanced modulation, finally strengthened the result.

The present invention compared with prior art, its remarkable advantage is:

Adopt the method for real-time statistics ISAR dynamic range of images and signal to noise ratio (S/N ratio); Real-time and Dynamic gain adjustment wave filter through various filtering methods, can keep, when strengthening ISAR as details as far as possible; Solve the problem of ISAR, and its Project Realization is simple as great dynamic range.This method has the advantages that real-time is good, the image reinforced effects is good, and its proposition and Project Realization are haveing highly application value aspect the ISAR picture processing enhancing.

Below in conjunction with accompanying drawing the present invention is described in further detail.

Four description of drawings

Fig. 1 is a workflow diagram of the present invention.

Fig. 2 is an original I SAR image.

Fig. 3 is the ISAR image that obtains after the present invention handles.

Five embodiments

Implementation method practical implementation step of the present invention does, referring to accompanying drawing 1:

1, at first original I SAR picture is carried out dynamic range and estimate, estimate the dynamic range of the current original I SAR picture (as shown in Figure 2) that obtains, utilize estimated dynamic range parameters, original image is carried out balanced histogram dynamic compression, concrete grammar is following:

If ISAR picture size W * H, its grey level histogram uses array representation to be: his [i] (i ∈ [0, N]), and N is the dynamic range of ISAR picture, A is this histogrammic average, promptly (W * H)/N, then threshold value is got T=0.3A.This method is number all is mapped to the nearest gray level greater than threshold value in front less than the gray scale of threshold value, and equidistance is arranged then.Balanced histogram dynamic compression can be described as:

Calculate the labeling function of each gray scale:

$f\left(i\right)=\left\{\begin{array}{c}0,\mathrm{his}\left[i\right]<T\\ 1,\mathrm{his}\left[i\right]\&GreaterEqual;T\end{array}\right.---\left(1\right)$

By formula

(i ∈ [0; 255]) ask for mapping function first; If mapping back number of greyscale levels is L, gray scale is carried out equidistance is arranged as:

${\mathrm{reflect}}^{\&prime;}\left[i\right]=\frac{255}{L}\mathrm{reflect}\left[i\right],(i\&Element;[0,L-1\left]\right)---\left(2\right)$

This method solves not enough this contradiction of dynamic range that the detector output signal dynamic range is excessive and display system is narrow, when improving picture contrast, greatly improve the fault-layer-phenomenon that direct equilibrium brings, and speed is fast slightly, is easier to hardware and realizes.

2, original I SAR picture is carried out the Harr wavelet decomposition, obtain the high frequency coefficient that original image, one deck and two layers of small echo approach image and one deck and two layers level under 3 yardsticks of coefficient, vertical, diagonal angle; The one deck that respectively original image, Harr wavelet decomposition is obtained, two layers of low frequency coefficient carry out the degree of membership coefficient estimation, calculate index of modulation α (i, j), edge calculation details then, method is following:

Original image S is carried out the fast discrete wavelet transformation, utilize the Harr wavelet basis to carry out 2 layers of decomposition, obtain 3 image S under the yardstick, ca1, ca2:

[ca1，ch1，cv1，cd1]＝dwt2(S，se) (3)

[ca2，ch2，cv2，cd2]＝dwt2(ca1，se) (4)

Respectively to the image S under 3 yardsticks, ca1, ca2 by following step, obtains image S, ca1, the high fdrequency component of ca2: S _Hp, ca1 _Hp, ca2 _HpAnd membership function

With

Wherein f is processed image, can be S, ca1 or ca2.

Image f obtains f through high-pass filtering _Hp, the Hi-pass filter that is adopted utilizes the morphology gradient filter, and expression formula is following:

f ⁺＝max{f(m，n)|(m，n)∈W(i，j)} (5)

f ^-＝min{f(m，n)|(m，n)∈W(i，j)} (6)

$f_{\mathrm{hp}}=\left\{\begin{array}{c}f-f^{-},f^{+}-f<f-f^{-}\\ f-f^{+},f^{+}-f>f-f^{-}\\ 0,f^{+}-f=f-f^{-}\end{array}\right.---\left(7\right)$

f ⁺, f ^-Gray scale maximal value and the minimum value of representing this neighborhood respectively, (i is so that (i is the center j), is the neighborhood scope of radius with r, and the value of r is 2 here j) to W.The radio-frequency component f that through type (8)～(10) obtain _HpSet up weak edge f respectively _El, strong edge f _EsWith noise f _nThe membership function μ of three kinds of compositions _El, μ _EsAnd μ _n, with f _HpBe mapped to the fuzzy characteristics plane.

Membership function is selected Gauss's membership function, noise, weak edge and strong edge point (i, j) degree of membership respectively as follows:

${\mathrm{\&mu;}}_{\mathrm{el}}(i,j)=\left\{\begin{array}{cc}\exp (-{(f_{\mathrm{hp}}(i,j)-e_{\mathrm{el}})}^2/2{{\mathrm{\&delta;}}_{\mathrm{el}}}^2)& 0\&le;x\&le;255\\ 0& \mathrm{else}\end{array}\right.---\left(8\right)$

${\mathrm{\&mu;}}_{\mathrm{es}}(i,j)=\left\{\begin{array}{cc}0& \mathrm{else}\\ \exp (-{(f_{\mathrm{hp}}(i,j)-e_{\mathrm{es}})}^2/2{{\mathrm{\&delta;}}_{\mathrm{es}}}^2)& 0\&le;x<e_{\mathrm{es}}\\ 1& e_{\mathrm{es}}\&le;x\&le;255\end{array}\right.---\left(9\right)$

${\mathrm{\&mu;}}_n(i,j)=\left\{\begin{array}{cc}1& 0\&le;x<e_n\\ \exp (-{(f_{\mathrm{hp}}(i,j)-e_n)}^2/2{{\mathrm{\&delta;}}_n}^2)& e_n\&le;x\&le;255\\ 0& \mathrm{else}\end{array}\right.---\left(10\right)$

Wherein, e _El, e _Es, e _nAnd δ _El, δ _Es, δ _nBe respectively the average and the variance of Gaussian function, through membership function with image mapped to the fuzzy characteristics plane.

Utilize weak edge f respectively _El, strong edge f _EsWith noise f _nThe membership function μ of three kinds of compositions _El, μ _EsAnd μ _n, through fuzzy characteristics plane computations index of modulation α (i, j).The f of final output edge details _Eh, expression formula is following:

f _eh(i，j)＝α(i，j)·f _hp(i，j) (11)

α(i，j)＝k ₁·μ _es(f _hp(i，j))-k ₂·max(μ _n(f _hp(i，j))，μ _el(f _hp(i，j))) (12)

Wherein, μ _El, μ _nAnd μ _EsObtain by formula (8)～(10) respectively.k ₁, k ₂Be respectively weighting factor, to S, ca1, ca2 is corresponding three groups of different weighting coefficients respectively.

3, utilize the edge details result of gained to carry out the Harr wavelet reconstruction, then with dynamic compression after the result carry out the image enhanced modulation, finally strengthened the result, reconstructing method is following:

With image S, ca1, the high fdrequency component S that the mapping of the process fuzzy set of ca2 obtains _Eh, ca1 _Eh, ca2 _EhBy the Harr wavelet basis to ca1 _EhCarry out inverse transformation, it is vertical, level and three components of diagonal line are obtained by step 1, by ca2 _Eh, ch2, cv2 and cd2 reconstruct ca1 '.Then by passing through ca1 _Eh+ k ' ca1 ', ch1, cv1 and cd1 reconstruct ca ', expression formula is following:

ca1′＝idwt2[ca2 _eh，ch2，cv2，cd2，se] (13)

ca′＝idwt2[ca1 _eh+k′·ca1′，ch1，cv1，cd1，se] (14)

Behind original image process LPF, with S _EhAnd ca ' carries out weighting summation, obtains final imaging results, and is as shown in Figure 3.

f _out＝S _base+k ₁S _eh+k ₂ca′ (15)。

Claims (1)

1. the fuzzy self-modulation of an ISAR image shows Enhancement Method, it is characterized in that: original I SAR picture is carried out dynamic range estimate, utilize estimated dynamic range parameters, original image is carried out balanced histogram dynamic compression; Then original I SAR picture is carried out wavelet decomposition; The one deck that respectively original image, Harr wavelet decomposition is obtained, two layers of low frequency coefficient carry out the degree of membership coefficient estimation; Calculate the index of modulation; Edge calculation details result then; Utilize the edge details result of gained to carry out the Harr wavelet reconstruction, then with dynamic compression after the result carry out the image enhanced modulation, finally strengthened the result.

Images