CN101706951B - Method, device and system for objectively evaluating pneumatic optical image quality based on feature fusion - Google Patents

Abstract

The invention discloses a method, a device and a system for objectively evaluating the quality of a pneumatic optical image based on feature fusion. The method comprises the steps: acquiring the pneumatic optical image and calculating fuzzy factors, deviation factors and active attenuation factors. The objective evaluation method based on factors feature fusion can effectively avoid complexity and time consumption of a subjective evaluation method, overcomes the defects of the traditional objective evaluation method, has the advantages of simple, fast and reliable calculation, good robustness and strong real-time performance, can be used for carrying out objective evaluation on the quality of the pneumatic optical image, assisting the property comparison of an image restoring and correcting method of a high-speed aircraft, and providing criteria for the frame selection of an iterative restoration algorithm, and iterative stopping operation.

Description

A kind of Pneumatic optical method for objectively evaluating image quality, Apparatus and system based on Feature Fusion

Technical field

The present invention relates to the Pneumatic optical field, relate in particular to a kind of Pneumatic optical method for objectively evaluating image quality, Apparatus and system based on Feature Fusion.

Background technology

When being loaded with optical imagery and surveying the high-speed aircraft of guidance system and fly in the endoatmosphere; Can form complicated flow field between optics head-shield and the incoming flow; Cause pneumatic heat, heat radiation and image transmission to disturb to the optical imagery detection system; Thereby cause skew, the shake and fuzzy of target image, this effect is called aero-optical effect.Aero-optical effect can cause target image smudgy usually, and optical detection and guidance system are brought multiple influence: reduce the target detection signal to noise ratio (S/N ratio), reduce the target detection distance; Reduce target detection identification probability, anti-bait ability drop; Deviation takes place in angle of sight position, line-of-sight rate by line is shaken, surveyed guidance precision and sharply descends.Decrease in image quality has a strong impact on higher level Flame Image Process such as follow-up target detection, tracking, therefore need the Pneumatic optical image be restored, and the quality of recovery effect relates to the problem of how to evaluate picture quality.

The evaluation of picture quality mainly is divided into subjective assessment and objective evaluation.Subjective evaluation method is as image observation person with the people; The picture quality quality is made subjective assessment; Therefore the most effectively, but these class methods are often more consuming time, and complicated operation; Can not satisfy the real-time requirement of high-speed missile imaging detection, therefore not be suitable for the quality assessment of Pneumatic optical image.

Method for objectively evaluating attempts to seek quantitative measuring method from the fidelity and the intelligibility of image.Existing objective evaluation technology mainly is divided into the evaluation method based on statistics and mathematics, and the evaluation method that incorporates human-eye visual characteristic.Square error (Mean-Squared Error, MSE) be adopt the earliest, also be one of the most frequently used method for objectively evaluating, reflected restored image and the global disparity between the degraded image not.Be out of shape the multiple evaluation function that draws by MSE; As the normalization square error (Normalized Mean-SquaredError, NMSE), Y-PSNR (Peak Signal-to-Noise Ratio, PSNR), improved signal to noise ratio (S/N ratio) (Improved Signal-to-Noise Ratio; ISNR), Minkowski tolerance (MinkowskiMetric; MM) etc., because form is simple, calculate advantages such as easy, also by extensive employing.In addition, also has evaluation method, like structural dependence (Structural Correlation based on correlativity; SC), CZenakowski distance (CZenakowski Distance, CZD), the metric analysis method on the frequency domain; As the spectral amplitude distortion (Spectral Magnitude Distortion, SMD), the spectrum phase distortion (SpectralPhase Distortion, SPD); And based on picture quality yardstick (the PictureQuality Scale of human-eye visual characteristic; PQS), disclosed the image order of information, tolerance more accurately and reliably.Yet the degeneration of Pneumatic optical image is not by due to compression losses and the additivity uncorrelated noise, and in view of the complicacy of aero-optical effect, above method for objectively evaluating all is not suitable for the quality assessment of Pneumatic optical image simultaneously.

The objective evaluation of picture quality is the very important and basic research topic of digital image processing field; Different degeneration reasons need be selected different evaluation methods usually, still do not have the general image quality evaluating method at present and can carry out accurate objective evaluation to all images degenerated form.The prior art spininess does not relate to the quality assessment of Pneumatic optical image to compression damaged image and noise degraded image.

Therefore, proposing a kind of method for evaluating objective quality ten minutes necessity that is applicable to the Pneumatic optical image, is one of Pneumatic optical field problem anxious to be solved at present.

Summary of the invention

The embodiment of the invention provides a kind of Pneumatic optical method for objectively evaluating image quality, Apparatus and system based on Feature Fusion; Through obtaining the Pneumatic optical image and calculating the fuzzy factor, displacement factor and activity decay factor; Method for objectively evaluating based on above-mentioned factor Feature Fusion can effectively be avoided the loaded down with trivial details consuming time of subjective evaluation method; Overcome the deficiency of traditional objective evaluation method; Has the advantage that calculating is simple and direct reliably, robustness is good, real-time; The performance that can be used for the quality of Pneumatic optical image is carried out objective evaluation, auxiliary high-speed aircraft image restoration bearing calibration relatively, and the frame selection and the iteration terminating operation that restore algorithm for iteration provide criterion.

The embodiment of the invention provides following technical scheme:

A kind of Pneumatic optical method for objectively evaluating image quality based on Feature Fusion, step comprises:

Step 1, obtain the Pneumatic optical image;

Step 2, obtain the fuzzy factor;

Step 3, the Pneumatic optical degeneration image multiframe is carried out arithmetic mean obtain reference picture;

Step 4, obtain displacement factor;

Step 5, obtain the activity decay factor;

Step 6, the Feature Fusion of carrying out, and according to image degradation type selecting weight coefficient;

The oeverall quality of step 7, acquisition Pneumatic optical image.

Preferably, above-mentioned steps four specifically comprises:

A. the Pneumatic optical image is carried out piecemeal and piece registration operation;

B. obtain centroid motion distance and centroid motion angle;

C. obtain offset distance variance and deviation angle variance;

D. obtain displacement factor.

Preferably, among the above-mentioned steps A, the size of piece is according to image size Dynamic Selection, between 8 * 8 to 64 * 64.Utilize the normalized crosscorrelation method that each piece of Pneumatic optical image to be measured is carried out registration operation respectively.

Preferably, among the above-mentioned steps B, the centroid motion distance of each piece is obtained by following formula with the centroid motion angle:

${\mathrm{\&Delta;}}_{\mathrm{ij}}=\frac{\sqrt{{\mathrm{\&Delta;x}}_{\mathrm{<figure-callout\; id="2"\; label="ij"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">ij</figure-callout>}}^2+{\mathrm{\&Delta;y}}_{\mathrm{ij}}^2}}{\sqrt{{\mathrm{<figure-callout\; id="2"\; label="M"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">M</figure-callout>}}^2+{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">N</figure-callout>}}^2}}=\frac{\sqrt{{({\hat{x}}_{i_2{j}_2}-x_{i_1{j}_1})}^2+{({\hat{y}}_{i_2{j}_2}-y_{i_1{j}_1})}^2}}{\sqrt{{\mathrm{<figure-callout\; id="2"\; label="M"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">M</figure-callout>}}^2+{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">N</figure-callout>}}^2}}$

${\mathrm{\&theta;}}_{\mathrm{ij}}=\frac{1}{\mathrm{\&pi;}}\arctan \left|\frac{{\mathrm{\&Delta;x}}_{\mathrm{ij}}}{{\mathrm{\&Delta;y}}_{\mathrm{ij}}}\right|+\frac{1}{2}=\frac{1}{\mathrm{\&pi;}}\arctan \left|\frac{{\hat{x}}_{i_2{j}_2}-x_{i_1{j}_1}}{{\hat{y}}_{i_2{j}_2}-y_{i_1{j}_1}}\right|+\frac{1}{2}$

Δ x _IjWith Δ y _Ij(i is j) along x axle and the axial side-play amount of y to represent any piece respectively.Δ _IjAnd θ _IjBe the side-play amount of each piece under polar coordinates, wherein Δ _IjThe offset distance of representing each piece, θ _IjRepresent corresponding deviation angle.

Preferably, among the above step C, the offset distance variance and the deviation angle variance of each piece are obtained by following formula:

${\mathrm{\&sigma;}}_{\mathrm{\&Delta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{({\mathrm{\&Delta;}}_{\mathrm{ij}}-\stackrel{\&OverBar;}{\mathrm{\&Delta;}})}^2,\stackrel{\&OverBar;}{\mathrm{\&Delta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{\mathrm{\&Delta;}}_{\mathrm{ij}}$

${\mathrm{\&sigma;}}_{\mathrm{\&theta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{({\mathrm{\&theta;}}_{\mathrm{ij}}-\stackrel{\&OverBar;}{\mathrm{\&theta;}})}^2,\stackrel{\&OverBar;}{\mathrm{\&theta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{\mathrm{ij}}$

Preferably, among the above-mentioned steps D, the displacement factor value of Pneumatic optical image is by CO=(σ _Δ+ σ _θ)/2 obtain.

Preferably, in the above-mentioned steps one, picture quality and image size and picture material are irrelevant, only relevant with image degradation target type and degree of degeneration.

Preferably, above-mentioned target type includes but not limited to circular, little triangle target and four targets.

Preferably, in the above step 2, will be based on the fuzzy factor of Shannon entropy conduct of Pneumatic optical picture material, its formula is:

$\mathrm{SE}=-\frac{1}{\ln \mathrm{MN}}\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}p(x,y)\ln p(x,y),p(x,y)=\frac{f(x,y)}{\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}f(x,y)}$

In the formula, (x is that (M * N is a Pneumatic optical image size to the arrival pixel for x, photon probability density y) y) to p.

Preferably, the size of fuzzy factor values has been reacted the fog-level of Pneumatic optical image, and it is fuzzy more to be worth big more expression, and its span is [0,1].

Preferably, in the above-mentioned steps five, adopt spatial frequency to quantize the activity attenuation degree of Pneumatic optical image.Spatial frequency is obtained by following formula:

$\mathrm{SF}=\sqrt{\frac{1}{2\mathrm{MN}}{\left[\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=2}{\overset{N}{\mathrm{\&Sigma;}}}\right[f(x,y)-f(x,y-1)]}^2+\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{x=2}{\overset{M}{\mathrm{\&Sigma;}}}{[f(x,y)-f(x-1,y)]}^2}$

A kind of Pneumatic optical picture quality objective evaluation device based on Feature Fusion; Above-mentioned objective evaluation device comprises image collection module, fuzzy factor acquisition module, reference picture acquisition module, displacement factor acquisition module, activity decay factor acquisition module, weighting coefficient module, picture quality acquisition module; Said weighting coefficient module; Be used for the fuzzy factor, displacement factor and activity decay factor are carried out Feature Fusion, and according to image degradation type selecting weight coefficient.

Preferably, above-mentioned displacement factor acquisition module also comprises a piecemeal and piece registration module, is used for the Pneumatic optical image is carried out piecemeal and piece registration operation.

Preferably, above-mentioned displacement factor acquisition module also comprises an offset distance and deviation angle acquisition module, is used to obtain the centroid motion distance and centroid motion angle of each piece.

Preferably, above-mentioned displacement factor acquisition module also comprises an offset distance variance and deviation angle variance acquisition module, is used to obtain the offset distance variance and the deviation angle variance of each piece.

Preferably, above-mentioned displacement factor acquisition module also comprises a displacement factor value computing module, is used to calculate the displacement factor value.

A kind of Pneumatic optical picture quality objective evaluation system based on Feature Fusion comprises above-mentioned any described collision detecting device.

A kind of Pneumatic optical method for objectively evaluating image quality, Apparatus and system provided by the invention based on Feature Fusion; Through obtaining the Pneumatic optical image and calculating the fuzzy factor, displacement factor and activity decay factor; Method for objectively evaluating based on above-mentioned factor Feature Fusion can effectively be avoided the loaded down with trivial details consuming time of subjective evaluation method; Overcome the deficiency of traditional objective evaluation method; Has the advantage that calculating is simple and direct reliably, robustness is good, real-time; The performance that can be used for the quality of Pneumatic optical image is carried out objective evaluation, auxiliary high-speed aircraft image restoration bearing calibration relatively, and the frame selection and the iteration terminating operation that restore algorithm for iteration provide criterion.

Description of drawings

In order to be illustrated more clearly in the embodiment of the invention or technical scheme of the prior art; To do to introduce simply to the accompanying drawing of required use in embodiment or the description of the Prior Art below; Obviously, the accompanying drawing in describing below only is some embodiments of the present invention, for those of ordinary skills; Under the prerequisite of not paying creative work property, can also obtain other accompanying drawing according to these accompanying drawings.

Fig. 1 is the process flow diagram based on the Pneumatic optical method for objectively evaluating image quality of Feature Fusion that the embodiment of the invention provides;

Fig. 2 is infrared wind-tunnel image synoptic diagram;

Fig. 3 is the Pneumatic optical picture quality objective evaluation device synoptic diagram based on Feature Fusion that the embodiment of the invention provides;

The inner structure synoptic diagram of the displacement factor acquisition module that Fig. 4 embodiment of the invention provides.

Embodiment

The embodiment of the invention provides a kind of Pneumatic optical method for objectively evaluating image quality, Apparatus and system based on Feature Fusion; Through obtaining the Pneumatic optical image and calculating the fuzzy factor, displacement factor and activity decay factor; Method for objectively evaluating based on above-mentioned factor Feature Fusion can effectively be avoided the loaded down with trivial details consuming time of subjective evaluation method; Overcome the deficiency of traditional objective evaluation method; Has the advantage that calculating is simple and direct reliably, robustness is good, real-time; The performance that can be used for the quality of Pneumatic optical image is carried out objective evaluation, auxiliary high-speed aircraft image restoration bearing calibration relatively, and the frame selection and the iteration terminating operation that restore algorithm for iteration provide criterion.For making the object of the invention, technical scheme and advantage clearer, the embodiment that develops simultaneously with reference to the accompanying drawings is to further explain of the present invention.

The embodiment of the invention provides a kind of Pneumatic optical method for objectively evaluating image quality based on Feature Fusion, and as shown in Figure 1, concrete steps comprise:

Step 1, obtain the Pneumatic optical image;

The Pneumatic optical image obtains through ultrasonic arc tunnel, and wherein, ultrasonic arc tunnel target type comprises three kinds of circular, little triangle target and four targets, and the Pneumatic optical image is taken through high frame frequency infrared thermoviewer and obtained.Be illustrated in figure 2 as infrared wind-tunnel image.(a) and (b) with (c) be respectively the frozen frozen mass of circular, little triangle target and four target image sequences; (d) and (g) be respectively the 4th and the 9th frame of circular target sequence; (e) and (h) be respectively the 4th and the 8th frame of little triangular day mark sequence; (f) and (i) be respectively the 2nd and the 14th frame of four target sequences.Technical indicator: Frames representes the totalframes of each sequence; Aperture representes the aperture; λ representes wavelength; τ representes integral time; F representes focal length; F representes frame frequency; FPA representes the detector pixel; Resolution representes test macro resolution.

Picture quality and image size and picture material are irrelevant, only relevant with image degradation target type and degree of degeneration.

Step 2, obtain the fuzzy factor;

The embodiment of the invention is analyzed multiple characteristics of image (comprising resemblance, unchangeability characteristic, entropy characteristic, gradient characteristic etc.) on the formation mechanism of research aero-optical effect, introduce the quality of fuzzy factor pair Pneumatic optical image and measure.

Particularly, because when the Pneumatic optical image took place to blur, it is unordered that its bulk strength will sharply descend, the edge is extended, the Pneumatic optical image becomes.Entropy is an index of describing the unordered degree of Pneumatic optical image, and the embodiment of the invention adopts the fuzzy factor of Shannon entropy conduct based on the Pneumatic optical picture material, and its formula is:

$\mathrm{SE}=-\frac{1}{\ln \mathrm{MN}}\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}p(x,y)\ln p(x,y),p(x,y)=\frac{f(x,y)}{\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}f(x,y)}$

In the formula, (x is that (M * N is a Pneumatic optical image size to the arrival pixel for x, photon probability density y) y) to p.Particularly, the size of fuzzy factor values has been reacted the fog-level of Pneumatic optical image, and it is fuzzy more to be worth big more expression, and its span is [0,1].

Step 3, the Pneumatic optical degeneration image multiframe is carried out arithmetic mean obtain reference picture;

Particularly; The Pneumatic optical degeneration image multiframe carried out arithmetic mean is approximate to be obtained reference picture and be: receive the influence of aero-optical effect; Each pixel of Pneumatic optical image is near a random fluctuation zero-mean position all; And Gaussian distributed, so the distribution center of pixel can be approximately reference picture respective pixel point position.We select the arithmetic mean of multiframe sequence image to be similar to reference picture.Each piece in the Pneumatic optical degeneration image travels through whole reference picture respectively, and according to computes go out corresponding normalization correlation matrix γ (u, v):

$\mathrm{\&gamma;}(u,v)=\frac{{\mathrm{\&Sigma;}}_{x,y}[f(x,y)-{\stackrel{\&OverBar;}{f}}_{u,v}][t(x-u,y-v)-\stackrel{\&OverBar;}{t}]}{{\left\{{\mathrm{\&Sigma;}}_{x,y}{[f(x,y)-{\stackrel{\&OverBar;}{f}}_{u,v}]}^2{\mathrm{\&Sigma;}}_{x,y}{[t(x-u,y-v)-\stackrel{\&OverBar;}{t}]}^{0.2}\right\}}^{0.5}}$

In the formula, ((x, y) expression is used for the characteristics of image that the piece registration is extracted, f to t to f for x, y) expression reference picture _{U, v}Expression selects the average of piece, the average of t representation feature t according to characteristic t in reference picture.Suppose that (u, the coordinate of peak value loca v) is (y to γ _Peak, x _Peak), in the reference picture

Obtain by following formula:

$x_{i_1{j}_1}=\frac{\underset{x=x_{\mathrm{peak}}-m+1}{\overset{x_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\underset{y=y_{\mathrm{peak}}-n+1}{\overset{y_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\mathrm{xf}(x,y)}{\underset{x=x_{\mathrm{peak}}-m+1}{\overset{x_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\underset{y=y_{\mathrm{peak}}-n+1}{\overset{y_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}f(x,y)},$ $y_{i_1{j}_1}=\frac{\underset{x=x_{\mathrm{peak}}-m+1}{\overset{x_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\underset{y=y_{\mathrm{peak}}-n+1}{\overset{y_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\mathrm{yf}(x,y)}{\underset{x=x_{\mathrm{peak}}-m+1}{\overset{x_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\underset{y=y_{\mathrm{peak}}-n+1}{\overset{y_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}f(x,y)}$

Step 4, obtain displacement factor;

When skew takes place in the Pneumatic optical image, its original position of will squinting, the part of image, thus cause the distortion that is distorted of Pneumatic optical image.Particularly, obtaining displacement factor further comprises the steps:

A. the Pneumatic optical image is carried out piecemeal and piece registration operation: suppose that Pneumatic optical image size to be measured is M * N, at first carry out the branch block operations, be divided into i * j piece, every block size is m * n.For any piece (i, j), its barycenter is obtained by following formula:

$x_{\mathrm{ij}}=\frac{\underset{x=(i-1)m+1}{\overset{\mathrm{im}}{\mathrm{\&Sigma;}}}\underset{y=(j-1)n+1}{\overset{\mathrm{jn}}{\mathrm{\&Sigma;}}}\mathrm{xf}(x,y)}{\underset{x=(i-1)m+1}{\overset{\mathrm{im}}{\mathrm{\&Sigma;}}}\underset{y=(j-1)n+1}{\overset{\mathrm{jn}}{\mathrm{\&Sigma;}}}f(x,y)},$ $y_{\mathrm{ij}}=\frac{\underset{x=(i-1)m+1}{\overset{\mathrm{im}}{\mathrm{\&Sigma;}}}\underset{y=(j-1)n+1}{\overset{\mathrm{jn}}{\mathrm{\&Sigma;}}}\mathrm{yf}(x,y)}{\underset{x=(i-1)m+1}{\overset{\mathrm{im}}{\mathrm{\&Sigma;}}}\underset{y=(j-1)n+1}{\overset{\mathrm{jn}}{\mathrm{\&Sigma;}}}f(x,y)},$

Order

With

Be respectively the piece (i of reference picture ₁, j ₁) and the piece (i of degraded image ₂, j ₂) barycenter.Particularly, the Pneumatic optical image is carried out the branch block operations, the size of piece is according to image size Dynamic Selection, between 8 * 8 to 64 * 64.Utilize the normalized crosscorrelation method that each piece of Pneumatic optical image to be measured is carried out registration operation respectively.

B. obtain centroid motion distance and centroid motion angle: the centroid motion distance and centroid motion angle of calculating each piece in the Pneumatic optical degeneration image according to following two formulas respectively:

${\mathrm{\&Delta;}}_{\mathrm{ij}}=\frac{\sqrt{{\mathrm{\&Delta;<figure-callout\; id="2"\; label="x\; ij"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">x</figure-callout>}}_{\mathrm{ij}}^{}+{\mathrm{\&Delta;y}}_{\mathrm{ij}}^2}}{\sqrt{{\mathrm{<figure-callout\; id="2"\; label="M"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">M</figure-callout>}}^2+{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">N</figure-callout>}}^2}}=\frac{\sqrt{{({\hat{x}}_{i_2{j}_2}-x_{i_1{j}_1})}^2+{({\hat{y}}_{i_2{j}_2}-y_{i_1{j}_1})}^2}}{\sqrt{{\mathrm{<figure-callout\; id="2"\; label="M"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">M</figure-callout>}}^2+{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">N</figure-callout>}}^2}}$

${\mathrm{\&theta;}}_{\mathrm{ij}}=\frac{1}{\mathrm{\&pi;}}\arctan \left|\frac{{\mathrm{\&Delta;x}}_{\mathrm{ij}}}{{\mathrm{\&Delta;y}}_{\mathrm{ij}}}\right|+\frac{1}{2}=\frac{1}{\mathrm{\&pi;}}\arctan \left|\frac{{\hat{x}}_{i_2{j}_2}-x_{i_1{j}_1}}{{\hat{y}}_{i_2{j}_2}-y_{i_1{j}_1}}\right|+\frac{1}{2}$

Δ x _IjWith Δ y _Ij(i is j) along x axle and the axial side-play amount of y to represent any piece respectively.Δ _IjAnd θ _IjBe the side-play amount of each piece under polar coordinates, wherein Δ _IjThe offset distance of representing each piece, θ _IjRepresent corresponding deviation angle.

C. obtain offset distance variance and deviation angle variance: to the offset distance variance and the deviation angle variance of each piece of view picture testing image statistics, its computing method are following:

${\mathrm{\&sigma;}}_{\mathrm{\&Delta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{({\mathrm{\&Delta;}}_{\mathrm{ij}}-\stackrel{\&OverBar;}{\mathrm{\&Delta;}})}^2,$ $\stackrel{\&OverBar;}{\mathrm{\&Delta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{\mathrm{\&Delta;}}_{\mathrm{ij}}$

${\mathrm{\&sigma;}}_{\mathrm{\&theta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{({\mathrm{\&theta;}}_{\mathrm{ij}}-\stackrel{\&OverBar;}{\mathrm{\&theta;}})}^2,$ $\stackrel{\&OverBar;}{\mathrm{\&theta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{\mathrm{ij}}$

D. obtain displacement factor: the displacement factor value of Pneumatic optical image to be detected is by CO=(σ _Δ+ σ _θ)/2 obtain.

Step 5, obtain the activity decay factor;

When light is propagated in High Speed Flow Field, because the refractive index inhomogeneity of random medium can cause wavefront shape and intensity to distort, thereby make the decay of target image activity, it is unclear etc. to be embodied in image energy reduction, texture.In embodiments of the present invention, adopt spatial frequency to describe the average mean square error between neighbor, prove that spatial frequency can be used for describing the activity level of image.Therefore, adopt spatial frequency to quantize the activity attenuation degree of Pneumatic optical image.Spatial frequency is obtained by following formula:

$\mathrm{SF}=\sqrt{\frac{1}{2\mathrm{MN}}{\left[\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=2}{\overset{N}{\mathrm{\&Sigma;}}}\right[f(x,y)-f(x,y-1)]}^2+\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{x=2}{\overset{M}{\mathrm{\&Sigma;}}}{[f(x,y)-f(x-1,y)]}^2}$

Step 6, the Feature Fusion of carrying out, and according to image degradation type selecting weight coefficient;

Above-mentioned three factors (the fuzzy factor, displacement factor and activity decay factor) are independent, uncorrelated mutually in twos.The Pneumatic optical image quality evaluating method is obtained by following formula:

BOA＝SE ^α·CO ^β·(1-SF)γ

In the formula, α, β and γ are weight coefficient.Particularly, rule of thumb be worth circular objective weight factor beta=0.4; Weight coefficient α=0.4 of little triangular day mark; Article four, weight coefficient β=γ=0.2 of target.

The oeverall quality of step 7, acquisition Pneumatic optical image.

The present invention also provides a specific embodiment following:

The high temperature blackbody radiation source forms required circular, little triangle target or four targets, and by collimating light pipe target is calibrated to the infinite distance.The infrared target radiation gets into arc tunnel through incident ZnS window, receives flow disturbance, is received by thermal infrared imager through Ge material outgoing window then.IR target generator, it is made up of high temperature black matrix, infrared collimating optical system, target target disc and control combination, is used for simulation and produces infrared target signal.The embodiment of the invention adopts infrared wind-tunnel image sequence, gathers circular, little triangle target and three sequences of four targets altogether.From each image sequence, respectively select 3 frames (comprising frozen frozen mass), be illustrated in figure 2 as infrared wind-tunnel image.Utilize BOA that picture quality is estimated, and compare, carry out as follows with existing evaluation method:

(1) obtain the Pneumatic optical image, the image size is 512 * 512, gathers from ultrasonic arc tunnel.Fig. 2 (a) and (b) with (c) be respectively the frozen frozen mass of circular, little triangle target and four target image sequences.(d) and (g) be respectively the 4th and the 9th frame of circular target sequence; (e) and (h) be respectively the 4th, 8 frames of little triangle target sequence; (f) and (i) be respectively the 2nd, 14 frames of four target sequences.

(2), ask the multiframe arithmetic mean to be similar to its corresponding reference picture respectively for three image sequences.

(3), calculate its fuzzy factor values, displacement factor value and activity decay factor value respectively, and acquisition BOA value is separately multiplied each other in weighting for circular, little triangle target and four target images.Rule of thumb be worth circular objective weight factor beta=0.4; Weight coefficient α=0.4 of little triangular day mark; Article four, weight coefficient β=γ=0.2 of target.

(4) Si Telieer that calculates testing image than (Strehl Ratio, SR).Si Telieer likens the image quality that is used for estimating thermal infrared imager for reference index to, i.e. the actual mass of testing image.

(5) calculate other classical evaluation index value of testing image respectively, like L ₂Norm, structurally associated (Structural Correlation; SC), correlated quality (Correlation Quality; CQ), CZenakowski distance (CZenakowski Distance, CZD), spectrum phase distortion (Spectral PhaseDistortion, SPD), spectral amplitude distortion (Spectral Magnitude Distortion; SMD) and structural similarity tolerance (Structural SIMilarity, SSIM).

Following table has shown of the present invention based on the Pneumatic optical method for objectively evaluating image quality of Feature Fusion and the comparative result of the method for objectively evaluating of classics.Wherein, BOA has well distinguished the mass discrepancy of image between image and group on the same group, conform to the evaluation result of reference index SR, and other classical evaluation index can not correctly be estimated the quality of Pneumatic optical image.

In addition, the embodiment of the invention also provides a kind of Pneumatic optical picture quality objective evaluation device based on Feature Fusion.A kind of Pneumatic optical picture quality objective evaluation device synoptic diagram as shown in Figure 3, as to provide for the embodiment of the invention based on Feature Fusion.

A kind of OO Parallel Collision pick-up unit comprises image collection module 11, fuzzy factor acquisition module 22, reference picture acquisition module 33, displacement factor acquisition module 44, activity decay factor acquisition module 55, weighting coefficient module 66, picture quality acquisition module 77.

Image collection module 11 is used for obtaining Pneumatic optical image to be detected through ultrasonic arc tunnel;

Particularly, the Pneumatic optical image obtains through ultrasonic arc tunnel, and wherein, ultrasonic arc tunnel target type comprises three kinds of circular, little triangle target and four targets, and the Pneumatic optical image is taken through high frame frequency infrared thermoviewer and obtained.Be illustrated in figure 2 as infrared wind-tunnel image.(a) and (b) with (c) be respectively the frozen frozen mass of circular, little triangle target and four target image sequences; (d) and (g) be respectively the 4th and the 9th frame of circular target sequence; (e) and (h) be respectively the 4th and the 8th frame of little triangular day mark sequence; (f) and (i) be respectively the 2nd and the 14th frame of four target sequences.Technical indicator: Frames representes the totalframes of each sequence; Aperture representes the aperture; λ representes wavelength; τ representes integral time; F representes focal length; F representes frame frequency; FPA representes the detector pixel; Resolution representes test macro resolution.Picture quality and image size and picture material are irrelevant, only relevant with image degradation target type and degree of degeneration.

Fuzzy factor acquisition module 22, the Shannon entropy that is used for the Pneumatic optical picture material is as the fuzzy factor;

Particularly, the embodiment of the invention is analyzed multiple characteristics of image (comprising resemblance, unchangeability characteristic, entropy characteristic, gradient characteristic etc.) on the formation mechanism of research aero-optical effect, introduces the quality of fuzzy factor pair Pneumatic optical image and measures.

Because when the Pneumatic optical image took place to blur, it is unordered that its bulk strength will sharply descend, the edge is extended, the Pneumatic optical image becomes.Entropy is an index of describing the unordered degree of Pneumatic optical image, and the embodiment of the invention adopts the fuzzy factor of Shannon entropy conduct based on the Pneumatic optical picture material, and its formula is:

$\mathrm{SE}=-\frac{1}{\ln \mathrm{MN}}\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}p(x,y)\ln p(x,y),p(x,y)=\frac{f(x,y)}{\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}f(x,y)}$

In the formula, (x is that (M * N is a Pneumatic optical image size to the arrival pixel for x, photon probability density y) y) to p.Particularly, the size of fuzzy factor values has been reacted the fog-level of Pneumatic optical image, and it is fuzzy more to be worth big more expression, and its span is [0,1].

Reference picture acquisition module 33 is used for that the Pneumatic optical degeneration image multiframe is carried out arithmetic mean and obtains reference picture;

Particularly; The Pneumatic optical degeneration image multiframe carried out arithmetic mean is approximate to be obtained reference picture and be: receive the influence of aero-optical effect; Each pixel of Pneumatic optical image is near a random fluctuation zero-mean position all; And Gaussian distributed, so the distribution center of pixel can be approximately reference picture respective pixel point position.We select the arithmetic mean of multiframe sequence image to be similar to reference picture.Each piece in the Pneumatic optical degeneration image travels through whole reference picture respectively, and according to computes go out corresponding normalization correlation matrix γ (u, v):

$\mathrm{\&gamma;}(u,v)=\frac{{\mathrm{\&Sigma;}}_{x,y}[f(x,y)-{\stackrel{\&OverBar;}{f}}_{u,v}][t(x-u,y-v)-\stackrel{\&OverBar;}{t}]}{{\left\{{\mathrm{\&Sigma;}}_{x,y}{[f(x,y)-{\stackrel{\&OverBar;}{f}}_{u,v}]}^2{\mathrm{\&Sigma;}}_{x,y}{[t(x-u,y-v)-\stackrel{\&OverBar;}{t}]}^{0.2}\right\}}^{0.5}}$

In the formula, ((x, y) expression is used for the characteristics of image that the piece registration is extracted, f to t to f for x, y) expression reference picture _{U, v}Expression selects the average of piece, the average of t representation feature t according to characteristic t in reference picture.Suppose that (u, the coordinate of peak value loca v) is (y to γ _Peak, x _Peak), in the reference picture

Obtain by following formula:

$x_{i_1{j}_1}=\frac{\underset{x=x_{\mathrm{peak}}-m+1}{\overset{x_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\underset{y=y_{\mathrm{peak}}-n+1}{\overset{y_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\mathrm{xf}(x,y)}{\underset{x=x_{\mathrm{peak}}-m+1}{\overset{x_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\underset{y=y_{\mathrm{peak}}-n+1}{\overset{y_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}f(x,y)},$ $y_{i_1{j}_1}=\frac{\underset{x=x_{\mathrm{peak}}-m+1}{\overset{x_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\underset{y=y_{\mathrm{peak}}-n+1}{\overset{y_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\mathrm{yf}(x,y)}{\underset{x=x_{\mathrm{peak}}-m+1}{\overset{x_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}\underset{y=y_{\mathrm{peak}}-n+1}{\overset{y_{\mathrm{peak}}}{\mathrm{\&Sigma;}}}f(x,y)}$

Displacement factor acquisition module 44 is used to obtain deviation distance variance and deviation angle variance to obtain the displacement factor value;

As shown in Figure 4, further, said displacement factor acquisition module 44 comprises piecemeal and piece registration module 441, offset distance and deviation angle acquisition module 442, offset distance variance and deviation angle variance acquisition module 443, displacement factor value computing module 444.

Piecemeal and piece registration module 441 are used for the Pneumatic optical image is carried out piecemeal and piece registration operation;

Particularly, suppose that Pneumatic optical image size to be measured is M * N, at first carry out the branch block operations, be divided into i * j piece, every block size is m * n.For any piece (i, j), its barycenter is obtained by following formula:

$x_{\mathrm{ij}}=\frac{\underset{x=(i-1)m+1}{\overset{\mathrm{im}}{\mathrm{\&Sigma;}}}\underset{y=(j-1)n+1}{\overset{\mathrm{jn}}{\mathrm{\&Sigma;}}}\mathrm{xf}(x,y)}{\underset{x=(i-1)m+1}{\overset{\mathrm{im}}{\mathrm{\&Sigma;}}}\underset{y=(j-1)n+1}{\overset{\mathrm{jn}}{\mathrm{\&Sigma;}}}f(x,y)},$ $y_{\mathrm{ij}}=\frac{\underset{x=(i-1)m+1}{\overset{\mathrm{im}}{\mathrm{\&Sigma;}}}\underset{y=(j-1)n+1}{\overset{\mathrm{jn}}{\mathrm{\&Sigma;}}}\mathrm{yf}(x,y)}{\underset{x=(i-1)m+1}{\overset{\mathrm{im}}{\mathrm{\&Sigma;}}}\underset{y=(j-1)n+1}{\overset{\mathrm{jn}}{\mathrm{\&Sigma;}}}f(x,y)}$

Order With

Be respectively the piece (i of reference picture ₁, j ₁) and the piece (i of degraded image ₂, j ₂) barycenter.Particularly, the Pneumatic optical image is carried out the branch block operations, the size of piece is according to image size Dynamic Selection, between 8 * 8 to 64 * 64.Utilize the normalized crosscorrelation method that each piece of Pneumatic optical image to be measured is carried out registration operation respectively.

Offset distance and deviation angle acquisition module 442 are used to obtain the centroid motion distance and centroid motion angle of each piece;

Particularly, calculate the centroid motion distance and centroid motion angle of each piece in the Pneumatic optical degeneration image respectively according to following two formulas:

${\mathrm{\&Delta;}}_{\mathrm{ij}}=\frac{\sqrt{{\mathrm{\&Delta;<figure-callout\; id="2"\; label="x\; ij"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">x</figure-callout>}}_{\mathrm{ij}}^{}+{\mathrm{\&Delta;y}}_{\mathrm{ij}}^2}}{\sqrt{{\mathrm{<figure-callout\; id="2"\; label="M"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">M</figure-callout>}}^2+{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">N</figure-callout>}}^2}}=\frac{\sqrt{{({\hat{x}}_{i_2{j}_2}-x_{i_1{j}_1})}^2+{({\hat{y}}_{i_2{j}_2}-y_{i_1{j}_1})}^2}}{\sqrt{{\mathrm{<figure-callout\; id="2"\; label="M"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">M</figure-callout>}}^2+{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H2009101992060E00021.png"\; state="\{\{state\}\}">N</figure-callout>}}^2}}$

${\mathrm{\&theta;}}_{\mathrm{ij}}=\frac{1}{\mathrm{\&pi;}}\arctan \left|\frac{{\mathrm{\&Delta;x}}_{\mathrm{ij}}}{{\mathrm{\&Delta;y}}_{\mathrm{ij}}}\right|+\frac{1}{2}=\frac{1}{\mathrm{\&pi;}}\arctan \left|\frac{{\hat{x}}_{i_2{j}_2}-x_{i_1{j}_1}}{{\hat{y}}_{i_2{j}_2}-y_{i_1{j}_1}}\right|+\frac{1}{2}$

Δ x _IjWith Δ y _Ij(i is j) along x axle and the axial side-play amount of y to represent any piece respectively.Δ _IjAnd θ _IjBe the side-play amount of each piece under polar coordinates, wherein Δ _IjThe offset distance of representing each piece, θ _IjRepresent corresponding deviation angle.

Offset distance variance and deviation angle variance acquisition module 443 are used to obtain the offset distance variance and the deviation angle variance of each piece;

Particularly, to the offset distance variance and the deviation angle variance of all pieces of view picture testing image statistics, its computing method are following:

${\mathrm{\&sigma;}}_{\mathrm{\&theta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{({\mathrm{\&theta;}}_{\mathrm{ij}}-\stackrel{\&OverBar;}{\mathrm{\&theta;}})}^2,$ $\stackrel{\&OverBar;}{\mathrm{\&Delta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{\mathrm{\&Delta;}}_{\mathrm{ij}}$

${\mathrm{\&sigma;}}_{\mathrm{\&theta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{({\mathrm{\&theta;}}_{\mathrm{ij}}-\stackrel{\&OverBar;}{\mathrm{\&theta;}})}^2,$ $\stackrel{\&OverBar;}{\mathrm{\&theta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{\mathrm{ij}}$

Displacement factor value computing module 444 is used to calculate the displacement factor value;

Particularly, the displacement factor value of Pneumatic optical image to be detected is by CO=(σ _Δ+ σ _θ)/2 obtain.

Activity decay factor acquisition module 55 is used to obtain the activity decay factor;

Particularly, when light is propagated, because the refractive index inhomogeneity of random medium can cause wavefront shape and intensity to distort, thereby make the decay of target image activity in High Speed Flow Field, it is unclear etc. to be embodied in image energy reduction, texture.In embodiments of the present invention, adopt spatial frequency to describe the average mean square error between neighbor, prove that spatial frequency can be used for describing the activity level of image.Therefore, adopt spatial frequency to quantize the activity attenuation degree of Pneumatic optical image.Spatial frequency is obtained by following formula:

$\mathrm{SF}=\sqrt{\frac{1}{2\mathrm{MN}}{\left[\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=2}{\overset{N}{\mathrm{\&Sigma;}}}\right[f(x,y)-f(x,y-1)]}^2+\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{x=2}{\overset{M}{\mathrm{\&Sigma;}}}{[f(x,y)-f(x-1,y)]}^2}$

Weighting coefficient module 66 is used for the fuzzy factor, displacement factor and activity decay factor are carried out Feature Fusion, and according to image degradation type selecting weight coefficient;

The fuzzy factor in the embodiment of the invention, displacement factor and activity decay factor are independent, uncorrelated mutually in twos.The Pneumatic optical image quality evaluating method is obtained by following formula:

BOA＝SE ^α·CO ^β·(1-SF)γ

In the formula, α, β and γ are weight coefficient.Particularly, rule of thumb be worth circular objective weight factor beta=0.4; Weight coefficient α=0.4 of little triangular day mark; Article four, weight coefficient β=γ=0.2 of target.

Picture quality acquisition module 77 is used to obtain the oeverall quality of pneumatic image.

The embodiment of the invention also further provides a kind of Pneumatic optical picture quality objective evaluation system based on Feature Fusion; Comprise the described Pneumatic optical picture quality objective evaluation device of the foregoing description based on Feature Fusion; Specify and see the foregoing description for details, repeat no more here.

One of ordinary skill in the art will appreciate that and realize that all or part of step that the foregoing description method is carried is to instruct relevant hardware to accomplish through program; Described program can be stored in a kind of computer-readable recording medium; This program comprises one of step or its combination of method embodiment when carrying out.

In addition, each functional unit in each embodiment of the present invention can be integrated in the processing module, also can be that the independent physics in each unit exists, and also can be integrated in the module two or more unit.Above-mentioned integrated module both can adopt the form of hardware to realize, also can adopt the form of software function module to realize.If said integrated module realizes with the form of software function module and during as independently production marketing or use, also can be stored in the computer read/write memory medium.

In sum; This paper provides a kind of Pneumatic optical method for objectively evaluating image quality, Apparatus and system based on Feature Fusion; Through obtaining the Pneumatic optical image and calculating the fuzzy factor, displacement factor and activity decay factor; Method for objectively evaluating based on above-mentioned factor Feature Fusion can effectively be avoided the loaded down with trivial details consuming time of subjective evaluation method, has overcome the deficiency of traditional objective evaluation method, has the advantage that calculating is simple and direct reliably, robustness is good, real-time.

More than a kind of Pneumatic optical method for objectively evaluating image quality based on Feature Fusion provided by the present invention, Apparatus and system have been carried out detailed introduction; Used concrete example among this paper principle of the present invention and embodiment are set forth, the explanation of above embodiment just is used for helping to understand scheme of the present invention; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, the part that on embodiment and range of application, all can change, in sum, this description should not be construed as limitation of the present invention.

Claims (17)

1. Pneumatic optical method for objectively evaluating image quality based on Feature Fusion is characterized in that said method for objectively evaluating comprises:

Step 1, obtain the Pneumatic optical image;

Step 2, obtain the fuzzy factor;

Step 3, the Pneumatic optical degeneration image multiframe is carried out arithmetic mean obtain reference picture;

Step 4, obtain displacement factor;

Step 5, obtain the activity decay factor;

Step 6, the Feature Fusion of carrying out, and according to image degradation type selecting weight coefficient;

The oeverall quality of step 7, acquisition Pneumatic optical image.

2. method for objectively evaluating according to claim 1 is characterized in that, said step 4 specifically comprises:

A. the Pneumatic optical image is carried out piecemeal and piece registration operation;

B. obtain centroid motion distance and centroid motion angle;

C. obtain offset distance variance and deviation angle variance;

D. obtain displacement factor.

3. method for objectively evaluating according to claim 2; It is characterized in that in said steps A, the size of piece is according to image size Dynamic Selection; Between 8 * 8 to 64 * 64, utilize the normalized crosscorrelation method that each piece of Pneumatic optical image to be measured is carried out registration operation respectively.

4. method for objectively evaluating according to claim 2 is characterized in that, in said step B, the centroid motion distance of each piece is obtained by following formula with the centroid motion angle:

${\mathrm{\&Delta;}}_{\mathrm{ij}}=\frac{\sqrt{{\mathrm{\&Delta;x}}_{\mathrm{ij}}^2+{\mathrm{\&Delta;y}}_{\mathrm{ij}}^2}}{\sqrt{M^2+N^2}}=\frac{\sqrt{{({\hat{x}}_{i_2{j}_2}-x_{i_1{j}_1})}^2+{({\hat{y}}_{i_2{j}_2}-y_{i_1{j}_1})}^2}}{\sqrt{M^2+N^2}}$

${\mathrm{\&theta;}}_{\mathrm{ij}}=\frac{1}{\mathrm{\&pi;}}\arctan \left|\frac{\mathrm{\&Delta;}{x}_{\mathrm{ij}}}{\mathrm{\&Delta;}{y}_{\mathrm{ij}}}\right|+\frac{1}{2}=\frac{1}{\mathrm{\&pi;}}\arctan \left|\frac{{\hat{x}}_{i_2{j}_2}-x_{i_1{j}_1}}{{\hat{y}}_{i_2{j}_2}-y_{i_1{j}_1}}\right|+\frac{1}{2}$

Δ x _IjWith Δ y _Ij(i is j) along x axle and the axial side-play amount of y, Δ to represent any piece respectively _IjAnd θ _IjBe the side-play amount of each piece under polar coordinates, wherein Δ _IjThe offset distance of representing each piece, θ _IjRepresent corresponding deviation angle,

With

Be respectively the piece (i of reference picture ₁, j ₁) and the piece (i of degraded image ₂, j ₂) barycenter, M * N is a Pneumatic optical image size.

5. method for objectively evaluating according to claim 4 is characterized in that, in said step C, the offset distance variance and the deviation angle variance of each piece are obtained by following formula:

${\mathrm{\&sigma;}}_{\mathrm{\&Delta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{({\mathrm{\&Delta;}}_{\mathrm{ij}}-\stackrel{\&OverBar;}{\mathrm{\&Delta;}})}^2,\stackrel{\&OverBar;}{\mathrm{\&Delta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{\mathrm{\&Delta;}}_{\mathrm{ij}}$

${\mathrm{\&sigma;}}_{\mathrm{\&theta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{({\mathrm{\&theta;}}_{\mathrm{ij}}-\stackrel{\&OverBar;}{\mathrm{\&theta;}})}^2,\stackrel{\&OverBar;}{\mathrm{\&theta;}}=\frac{\mathrm{mn}}{\mathrm{MN}}\underset{i=1}{\overset{\frac{M}{m}}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{\frac{N}{n}}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{\mathrm{ij}}$

M * N is a Pneumatic optical image size, and said Pneumatic optical image is divided into i * j piece, and every block size is m * n.

6. method for objectively evaluating according to claim 5 is characterized in that, in said step D, the displacement factor value of Pneumatic optical image is by CO=(σ _Δ+ σ _θ)/2 obtain.

7. method for objectively evaluating according to claim 1 is characterized in that, in said step 1, picture quality and image size and picture material are irrelevant, only relevant with image degradation target type and degree of degeneration.

8. method for objectively evaluating according to claim 7 is characterized in that, said target type comprises circular, little triangle target and four targets.

9. method for objectively evaluating according to claim 1 is characterized in that, in said step 2, will be based on the fuzzy factor of Shannon entropy conduct of Pneumatic optical picture material, and its formula is:

$\mathrm{SE}=-\frac{1}{\ln \mathrm{MN}}\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}p(x,y)\ln p(x,y),p(x,y)=\frac{f(x,y)}{\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}f(x,y)}$

In the formula, (x is that (M * N is a Pneumatic optical image size to the arrival pixel for x, photon probability density y), f (x, y) expression reference picture y) to p.

10. method for objectively evaluating according to claim 9 is characterized in that, the size of fuzzy factor values has been reacted the fog-level of Pneumatic optical image, and it is fuzzy more to be worth big more expression, and its span does ^[0,1]

11. method for objectively evaluating according to claim 1 is characterized in that, in said step 5, adopts spatial frequency to quantize the activity attenuation degree of Pneumatic optical image, spatial frequency is obtained by following formula:

$\mathrm{SF}=\sqrt{\frac{1}{2\mathrm{MN}}[\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=2}{\overset{N}{\mathrm{\&Sigma;}}}{[f(x,y)-f(x,y-1)]}^2+\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{x=2}{\overset{M}{\mathrm{\&Sigma;}}}{[f(x,y)-f(x-1,y)]}^2}$

M * N is a Pneumatic optical image size, f (x, y) expression reference picture.

12. the Pneumatic optical picture quality objective evaluation device based on Feature Fusion is characterized in that said objective evaluation device comprises

Image collection module is used for obtaining Pneumatic optical image to be detected through ultrasonic arc tunnel;

Fuzzy factor acquisition module, the Shannon entropy that is used for the Pneumatic optical picture material is as the fuzzy factor;

The reference picture acquisition module is used for that the Pneumatic optical degeneration image multiframe is carried out arithmetic mean and obtains reference picture;

The displacement factor acquisition module is used to obtain deviation distance variance and deviation angle variance to obtain the displacement factor value;

Activity decay factor acquisition module is used to obtain the activity decay factor;

The weighting coefficient module is used for the fuzzy factor, displacement factor and activity decay factor are carried out Feature Fusion, and according to image degradation type selecting weight coefficient;

The picture quality acquisition module is used to obtain the oeverall quality of pneumatic image.

13. objective evaluation device according to claim 12 is characterized in that, said displacement factor acquisition module also comprises a piecemeal and piece registration module, is used for the Pneumatic optical image is carried out piecemeal and piece registration operation.

14. objective evaluation device according to claim 12 is characterized in that, said displacement factor acquisition module also comprises an offset distance and deviation angle acquisition module, is used to obtain the centroid motion distance and centroid motion angle of each piece.

15. objective evaluation device according to claim 12 is characterized in that, said displacement factor acquisition module also comprises an offset distance variance and deviation angle variance acquisition module, is used to obtain the offset distance variance and the deviation angle variance of each piece.

16. objective evaluation device according to claim 12 is characterized in that, said displacement factor acquisition module also comprises a displacement factor value computing module, is used to calculate the displacement factor value.

17. the Pneumatic optical picture quality objective evaluation system based on Feature Fusion is characterized in that, comprises like any described objective evaluation device of claim 12 to 16.

Images