CN101799917A - Recovery method of binary image based on bistable state random resonance vibration mechanism - Google Patents

Abstract

The invention relates to a recovery method of a binary image based on a bistable state random resonance vibration mechanism, aiming to solve the problem of poor effect of the traditional image recovery method. The method comprises the following steps of: firstly, obtaining two one-dimensional sequences by respectively scanning a noised binary image according to lines and columns, and subtracting 128 by each pixel value in each one-dimensional sequence; secondly, inputting to a corresponding bistable state system to obtain two columns of corresponding output signals, and regulating parameters of the corresponding bistable state system to enable the output of the bistable state system to reach an optimal random resonance vibration state; thirdly, recovering optimal output signals to two-dimensional signals; and fourthly, inputting the two recovered two-dimensional signals into a discriminator and outputting the recovered binary image by the discriminator after discrimination. The invention has the advantages of favorable robustness and strong noise resistance and is suitable for recovering an image polluted by heavy noise.

Description

Recovery method of binary image based on bistable-state random resonance mechanism

Technical field

The invention belongs to image processing field, relate to a kind of recovery method of binary image based on bistable-state random resonance mechanism.

Background technology

1981, Italian scholar Benzi proposed Stochastic Resonance Theory, disclosed under the synergy of noise, periodic input signal and nonlinear system, and noise energy can be converted into useful signal energy, thereby improved the periodically signal to noise ratio (S/N ratio) of output signal.Accidental resonance utilizes the characteristic of noise enhancing signal, and having broken can only be by eliminating the intrinsic notion that noise come enhancing signal, and obtained using widely in the detection of weak signal with in estimating.Along with going deep into of accidental resonance applied research, its range of application extends to digital image processing field gradually, utilize bistable system to realize the enhancing of image DC component as Ye Qinghua etc., in addition in thresholding system, discoveries such as Marks can make image meet people's visual signature more by the noise that adds suitable intensity to image, and these results provide new thinking for the recovery of image.But at present, accidental resonance is only limited to the single-frequency component of handling in the image in the application in figure image intensifying field, yet in practice, a large amount of images is a nonperiodic signal, these images have comprised abundant frequency content usually, the recovery of single-frequency components far away can not practical requirement, is result on the statistical significance and utilize thresholding system accidental resonance technology institute image restored, and the image after needing to handle in a large number superposes to reach the recovery purpose.

Summary of the invention

The recovery technique that the present invention is directed to conventional images filtering can not be obtained good effect under the low signal-to-noise ratio situation, a kind of bistable-state random resonance recovery method of binary image is provided.

Bistable-state random resonance recovery method of binary image of the present invention may further comprise the steps:

Step (1) obtains two one-dimensional sequence by row and column scanning respectively to noisy bianry image;

Step (2) deducts 128 to each pixel value in each one-dimensional sequence;

Step (3) inputs to corresponding bistable system with each one-dimensional sequence, obtains two corresponding row output signals;

Step (4) is regulated corresponding bistable system parameter, makes the output of bistable system reach best accidental resonance state.Concrete grammar is: calculate the mutual information entropy of bistable system, when the mutual information entropy reached maximum value, the bistable state parameter value of fixing this moment, the output sequence of this moment were the best output of bistable system;

Step (5) restores the optimum output signals that obtains concerning postamble sequence as corresponding, and it is reverted to 2D signal;

Two 2D signal input arbiters that step (6) will be recovered, arbiter is exported the bianry image after restoring after differentiating.The method of discrimination of arbiter is: if the one-dimensional signal that obtains with line scanning is as the input signal of bistable system, thereby the value of output 2D signal on the relevant position that obtains greater than another 2D signal, then this locational pixel value is 255; Otherwise this locational pixel value is 0.

The beneficial effect that the inventive method had is: the present invention is applied to the image restoration field with bistable-state random resonance, and than traditional images filtering restored method, robustness of the present invention is good, and the anti-ability of making an uproar by force is strong, the suitable image that recovers by very noisy polluted.

Description of drawings

Fig. 1 is the inventive method ultimate principle figure.

Embodiment

The present invention is further described below in conjunction with embodiment and accompanying drawing, but the present invention only limits to the embodiment that introduced by no means.

As shown in Figure 1, a kind of bistable-state random resonance recovery method of binary image comprises the steps:

Step (1) is established noisy bianry image 1 the capable M row of N, and then its total pixel value is N * M, and noisy bianry image 1 is obtained one-dimensional sequence S by row and column scanning respectively _Row(m) and S _Col(m) (S wherein _Row(m) and S _Col(m) represent respectively by row with by the pixel value of m the pixel that obtains after the column scan, m=1,2,3 ..., N * M);

The pixel span of the noisy bianry image 1 of step (2) is [0,255].Because it is bipolar signal that bistable system 4 requires input signal, in view of the above, the pixel span of noisy bianry image 1 is mapped to [128,127], promptly the pixel value of each pixel deducts 128, two one-dimensional sequence S that obtain _Row' (m) 2 and S _Col' (m) 3;

Step (3) bistable system 4 expression formulas are as follows:

$\frac{\mathrm{dx}}{\mathrm{dt}}=\mathrm{ax}-b{x}^3+f\left(t\right)+\mathrm{\ξ}\left(t\right)$

In the formula, x is the output signal of system, and a, b are positive number, and f (t) is an input signal, and ξ (t) is a noise item.Suppose that here input noise ξ (t) is 0 white Gaussian noise for average, its autocorrelation function satisfies＜ξ (t) ξ (0) 〉=2D δ (t), D represents noise intensity, δ (t) represents impulse function.

Two one-dimensional sequence S that step (2) is obtained _Row' (m) 2 and S _Col' (m) 3 respectively as the input signal of bistable system, obtains two row output signals of bistable system;

Step (4) is in information theory, and what of information transmission the mutual information entropy described, and can reflect the quality of system's output performance exactly.List entries and output sequence according to step (3), calculate the mutual information entropy of bistable system 4, and with it as measurement index, regulate the bistable system parameter: when the mutual information entropy reaches maximum value, show that bistable system is in best accidental resonance state, the bistable state parameter value of fixing this moment, the output sequence S of this moment _Row" (m) and S _Col" (m) be the best output of bistable system 4, and bistable system 4 outputs of this moment are keeping having obtained fast as far as possible response speed under certain steady-state error prerequisite, it is best that the output performance of system reaches;

Calculate mutual information entropy H (S _In, S _Out) formula be:

$H(S_{\mathrm{in}},S_{\mathrm{out}})=-\underset{i}{\mathrm{\Σ}}{P}_{\mathrm{in}}\left(i\right){\log }_2{P}_{\mathrm{in}}\left(i\right)+\underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}{P}_{\mathrm{out}|\mathrm{in}}\left(j\right|i){\log }_2{P}_{\mathrm{out}|\mathrm{in}}\left(j\right|i)$

Wherein, S _InThe expression input signal, S _OutThe expression output signal, P _In(i) the expression input signal values is the probability of i, P _Out|in(j|i) expression is when bistable system 4 input signal values are i, and the system output signal value is the conditional probability of j.By calculating the mutual information entropy of 4 pairs of signal responses of bistable system, whether the output performance that can weigh system quantitatively reaches optimum state;

The two row optimum output signals S that step (5) obtains step (4) _Row" (m) and S _Col" (m) restore postamble sequence, and it is reverted to two 2D signal I as corresponding _Row(k) (l) and I _Col(k) (l), wherein, k=1,2 ..., N; L=1,2 ..., M;

Step (6) is got two related 2D signal I of step (5) in order to reduce error in judgement _Row(k) (l) and I _Col(k) (l) by arbiter 5, the concrete decision rule of arbiter 5 is as follows: if I _Row(k) (l) greater than I _Col(k) (l), then this locational pixel value is 255; Otherwise this locational pixel value is 0.

$S_{\mathrm{out}}\left(k\right)\left(l\right)=\left\{\begin{array}{c}255,I_{\mathrm{row}}\left(k\right)\left(l\right)>I_{\mathrm{col}}\left(k\right)\left(l\right)\\ 0,I_{\mathrm{row}}\left(k\right)\left(l\right)\≤I_{\mathrm{col}}\left(k\right)\left(l\right)\end{array}\right.$

S wherein _Out(k) (l) be final output result, be the bianry image 6 after the recovery.

Claims (1)

1. based on the recovery method of binary image of bistable-state random resonance mechanism, it is characterized in that this method comprises the steps:

Step (1) obtains two one-dimensional sequence by row and column scanning respectively to noisy bianry image;

Step (2) deducts 128 to each pixel value in each one-dimensional sequence;

Step (3) inputs to corresponding bistable system with each one-dimensional sequence, obtains two corresponding row output signals;

Step (4) is regulated corresponding bistable system parameter, makes the output of bistable system reach best accidental resonance state; Concrete grammar is: calculate the mutual information entropy of bistable system, when the mutual information entropy reached maximum value, the bistable state parameter value of fixing this moment, the output sequence of this moment were the best output of bistable system;

Step (5) restores the optimum output signals that obtains concerning postamble sequence as corresponding, and it is reverted to 2D signal;

Two 2D signal input arbiters that step (6) will be recovered, arbiter is exported the bianry image after restoring after differentiating; The method of discrimination of arbiter is: if the one-dimensional signal that obtains with line scanning is as the input signal of bistable system, thereby the value of output 2D signal on the relevant position that obtains greater than another 2D signal, then this locational pixel value is 255; Otherwise this locational pixel value is 0.

Images