CN104268844A

CN104268844A - Small target infrared image processing method based on weighing local image entropy

Info

Publication number: CN104268844A
Application number: CN201410554115.5A
Authority: CN
Inventors: 周欣; 邓鹤; 孙献平; 叶朝辉; 刘买利
Original assignee: Wuhan Institute of Physics and Mathematics of CAS
Current assignee: Institute of Precision Measurement Science and Technology Innovation of CAS; Wuhan Zhongke Medical Technology Industrial Technology Research Institute Co Ltd
Priority date: 2014-10-17
Filing date: 2014-10-17
Publication date: 2015-01-07
Anticipated expiration: 2034-10-17
Also published as: CN104268844B

Abstract

In order to effectively process small target infrared images under the low signal to noise ratio complex background, the invention discloses a small target infrared image processing method based on the weighing local image entropy and relates to the technical field of digital image processing. Inherent features of the small target infrared images are utilized, a multi-scale gray difference operator and a local image entropy operator are provided, the weighing local image entropy is obtained through the dot product operation, so that the infrared image background and the noise are effectively restrained, a target is enhanced, and finally the signal to noise ratio of the images is greatly improved.

Description

A Small Target Infrared Image Processing Method Based on Weighted Local Image Entropy

技术领域technical field

本发明涉及数字图像处理技术领域，具体是一种基于加权局部图像熵的小目标红外图像处理方法。The invention relates to the technical field of digital image processing, in particular to a small target infrared image processing method based on weighted local image entropy.

背景技术Background technique

小目标红外图像处理技术已在民用领域(如卫星大气红外云图分析、红外医疗图像病理分析、地质分析、海面人员搜救、入侵检测、森林火灾探测)和军事领域(如精确制导、预警探测、战地指挥和侦察、敌我识别)得到广泛应用，其目标检测步骤是红外图像处理领域的重点和难点，其性能好坏直接决定红外系统的有效作用距离及设备的复杂程度，因而该技术的研究受到了国内外众多学者持续而普遍的关注。Small target infrared image processing technology has been used in civilian fields (such as satellite atmospheric infrared cloud image analysis, infrared medical image pathological analysis, geological analysis, sea surface personnel search and rescue, intrusion detection, forest fire detection) and military fields (such as precision guidance, early warning detection, battlefield Command and reconnaissance, friend-or-foe identification) are widely used, and its target detection step is the focus and difficulty in the field of infrared image processing, and its performance directly determines the effective range of the infrared system and the complexity of the equipment, so the research on this technology has received Many scholars at home and abroad have continued to pay close attention to it.

小目标图像中的目标小、强度弱，没有先验的大小、形状及纹理等特征，且目标、背景和噪声混叠在一起，难以直接检测。然而，背景一般认为在空域上具有相关性，在时域上具有稳定性，且在频域上处于图像的低频部分，而目标通常认为在空域上与背景不相关，在频域上处于图像的高频部分。因此，小目标红外图像处理算法主要分为时间域、空间域和变换域三类：时域算法主要用于抑制具有短时平稳性的背景，但对复杂背景的抑制效果不理想。空域算法具有良好的实时性，易于实现。中值滤波只适合消除脉冲宽度小于滤波窗口的随机噪声，无法处理结构化的背景；顶帽变换是一种实用的非线性背景滤波技术，但需要图像的先验知识，自适应性不强；自适应滤波技术如二维最小均方误差滤波等算法，要求背景的统计特性不变或者缓慢变化，所以无法有效抑制复杂背景。变换域算法如基于自适应频率域巴特沃斯高通滤波、小波变换等，但此类算法来源于Fourier变换，受海森堡(Heisenberg)测不准原理的制约(即时间窗口与频率窗口的乘积为一个常数)，并且需要正反两次变换，算法运算量大。The target in the small target image is small, weak in intensity, and has no prior characteristics such as size, shape and texture, and the target, background and noise are mixed together, making it difficult to detect directly. However, the background is generally considered to be correlated in the spatial domain, stable in the time domain, and in the low-frequency part of the image in the frequency domain, while the target is generally considered to be uncorrelated with the background in the spatial domain, and in the low-frequency part of the image in the frequency domain. high frequency part. Therefore, small target infrared image processing algorithms are mainly divided into three categories: time domain, space domain and transform domain. The time domain algorithm is mainly used to suppress the background with short-term stationarity, but the suppression effect on the complex background is not ideal. The airspace algorithm has good real-time performance and is easy to implement. Median filtering is only suitable for eliminating random noise whose pulse width is smaller than the filtering window, and cannot deal with structured background; top-hat transform is a practical nonlinear background filtering technology, but it requires prior knowledge of the image, and its adaptability is not strong; Adaptive filtering techniques such as two-dimensional minimum mean square error filtering and other algorithms require the statistical characteristics of the background to remain unchanged or change slowly, so complex backgrounds cannot be effectively suppressed. Transform domain algorithms are based on adaptive frequency domain Butterworth high-pass filtering, wavelet transform, etc., but such algorithms are derived from Fourier transform and are restricted by the uncertainty principle of Heisenberg (that is, the product of time window and frequency window is a constant), and requires two positive and negative transformations, and the algorithm has a large amount of computation.

虽然小目标红外图像处理领域已取得了很多成果，并且已有很多算法在工程应用中得到了很好的实现，但对于复杂背景下低信噪比小目标红外图像，其目标检测系统工程依然面临很大的难度和复杂性。如何设计出结构简单、滤波效果好、鲁棒性强的小目标红外图像处理算法是目标检测技术研究的关键问题。Although many achievements have been made in the field of small target infrared image processing, and many algorithms have been well realized in engineering applications, the target detection system engineering is still facing challenges for small target infrared images with low signal-to-noise ratio in complex backgrounds. Great difficulty and complexity. How to design a small target infrared image processing algorithm with simple structure, good filtering effect and strong robustness is a key issue in the research of target detection technology.

发明内容Contents of the invention

本发明是针对现有小目标红外图像处理方法存在的上述技术问题，提供了一种基于加权局部图像熵的小目标红外图像处理方法。The present invention aims at the above-mentioned technical problems existing in the existing small target infrared image processing method, and provides a small target infrared image processing method based on weighted local image entropy.

一种基于加权局部图像熵的小目标红外图像处理方法，包括以下步骤：A small target infrared image processing method based on weighted local image entropy, comprising the following steps:

步骤1、求解图像各个像素点(x,y)的多尺度灰度差异D；Step 1. Solve the multi-scale grayscale difference D of each pixel point (x, y) of the image;

步骤2、求解图像各个像素点(x,y)的局部图像熵E；Step 2, solving the local image entropy E of each pixel point (x, y) of the image;

步骤3、通过多尺度灰度差异D和局部图像熵E获得各个像素点(x,y)的加权局部图像熵H；Step 3. Obtain the weighted local image entropy H of each pixel point (x, y) through the multi-scale grayscale difference D and the local image entropy E;

步骤4、根据加权局部图像熵H求解自适应阈值T，并通过自适应阈值T对加权局部图像熵H进行二值化，检测出红外小目标。Step 4. Solve the adaptive threshold T according to the weighted local image entropy H, and perform binarization on the weighted local image entropy H through the adaptive threshold T to detect small infrared targets.

如上所述的步骤1的多尺度灰度差异D通过以下步骤求解：The multi-scale gray difference D of step 1 as described above is solved by the following steps:

步骤1.1、对于红外图像I中每一个像素点(x,y)对应的灰度值为I(x,y)，设置像素点(x,y)的最大邻域空间Ω_max，邻域空间Ω_max的大小为L_max×L_max，其中L_max为大于1的正奇数；Step 1.1. For the grayscale value corresponding to each pixel point (x, y) in the infrared image I (x, y), set the maximum neighborhood space Ω _max of the pixel point (x, y), and the neighborhood space Ω The size of _max is L _max ×L _max , where L _max is a positive odd number greater than 1;

步骤1.2、获得每一个像素点(x,y)的邻域空间集{Ω_k|k＝1,2,…,L}，其中L＝(L_max-1)/2，Ω_k的大小为(2·k+1)×(2·k+1)；Step 1.2. Obtain the neighborhood space set {Ω _k |k=1,2,...,L} of each pixel point (x, y), where L=(L _max -1)/2, and the size of Ω _k is (2·k+1)×(2·k+1);

步骤1.3、利用以下公式计算每一个像素点(x,y)的邻域Ω_k与Ω_max之间的灰度差异D_k(x,y)，k＝1,2,…,L：Step 1.3, use the following formula to calculate the gray level difference D _k (x, y) between the neighborhood Ω _k and Ω _max of each pixel point (x, y), k=1, 2, ..., L:

${D D.}_{k k} ((x x,, y the y)) = = {| | \frac{11}{{N N}_{{Ω Ω}_{k k}}} \underset{((s the s,, t t)) &Element; &Element; {Ω Ω}_{k k}}{Σ Σ} I I ((s the s,, t t)) - - \frac{11}{{N N}_{{Ω Ω}_{max max}}} \underset{((p p,, q q)) &Element; &Element; {Ω Ω}_{max max}}{Σ Σ} I I ((p p,, q q)) | |}^{22},, k k = = 1,2 1,2,, . . . . . .,, L L$

其中，和分别表示邻域Ω_k、Ω_max内像素点的数目，I(s,t)表示邻域Ω_k内的点(s,t)处的灰度值，I(p,q)表示邻域Ω_max内的点(p,q)处的灰度值；in, and Respectively represent the number of pixels in the neighborhood Ω _k and Ω _max , I(s,t) represents the gray value at the point (s,t) in the neighborhood Ω _k , I(p,q) represents the neighborhood Ω The gray value at the point (p,q) within _max ;

步骤1.4、计算每一个像素点(x,y)所对应的的多尺度灰度差异D(x,y)：Step 1.4, calculate the multi-scale grayscale difference D(x,y) corresponding to each pixel point (x,y):

D(x,y)＝max{D₁(x,y),D₂(x,y),...,D_L(x,y)}。D(x,y)=max{D ₁ (x,y), D ₂ (x,y), . . . , D _L (x,y)}.

如上所述的步骤2的局部图像熵E通过以下步骤求解：The local image entropy E of step 2 as described above is solved by the following steps:

设定红外图像I中每一个像素点(x,y)的邻域空间Θ，邻域空间Θ的大小为m×n，计算像素点(x,y)处的局部图像熵:Set the neighborhood space Θ of each pixel point (x, y) in the infrared image I, the size of the neighborhood space Θ is m × n, calculate the local image entropy at the pixel point (x, y):

$E E. ((x x,, y the y)) = = - - {Σ Σ}_{i i = = 00}^{m m - - 11} {Σ Σ}_{j j = = 00}^{n no - - 11} p p ((I I ((i i,, j j)))) \cdot &Center Dot; {log log}_{22} ((p p ((I I ((i i,, j j)))) + + ϵ ϵ)),, p p ((I I ((i i,, j j)))) = = \frac{I I ((i i,, j j))}{{Σ Σ}_{i i = = 00}^{m m - - 11} {Σ Σ}_{j j = = 00}^{n no - - 11} I I ((i i,, j j))}$

其中，ε是设定的正常数，I(i,j)表示邻域Θ内的点(i,j)处的灰度值，遍历红外图像I中每一个像素点，获得红外图像I的局部图像熵E。Among them, ε is a set normal number, I(i, j) represents the gray value at the point (i, j) in the neighborhood Θ, traverses each pixel in the infrared image I, and obtains the local area of the infrared image I Image entropy E.

如上所述的步骤3的加权局部图像熵H通过以下步骤求解：The weighted local image entropy H of step 3 as described above is solved by the following steps:

对每一个像素点(x,y)经过步骤1处理所得到的多尺度灰度差异D与经过步骤2处理所得到的局部图像熵E进行点积运算，获得每一个像素点(x,y)对应的加权局部图像熵H。Perform a dot product operation on the multi-scale grayscale difference D obtained by step 1 for each pixel (x, y) and the local image entropy E obtained by step 2 to obtain each pixel (x, y) The corresponding weighted local image entropy H.

如上所述的自适应阈值T的通过以下公式进行确定：As mentioned above, the adaptive threshold T is determined by the following formula:

T＝c·SNR·σ+mm,SNR＝(H_max-mm)/σT＝c·SNR·σ+mm, SNR＝(H _max -mm)/σ

其中，c为正的常数，σ为加权局部图像熵H的标准差，mm为加权局部图像熵H的均值，H_max为加权局部图像熵H的最大值。Among them, c is a positive constant, σ is the standard deviation of the weighted local image entropy H, mm is the mean value of the weighted local image entropy H, and H _max is the maximum value of the weighted local image entropy H.

本发明与现有技术相比，具有以下优点：Compared with the prior art, the present invention has the following advantages:

1.本发明利用了小目标红外图像中目标和背景的特点，不依赖于红外图像模型和参数选择，能有效地抑制红外图像背景和噪声，提高红外图像的信噪比，从而提高目标的检测概率，降低虚警概率。1. The present invention utilizes the characteristics of the target and background in the small target infrared image, does not depend on the infrared image model and parameter selection, can effectively suppress the infrared image background and noise, improve the signal-to-noise ratio of the infrared image, thereby improving the detection of the target probability, reducing the probability of false alarms.

2.本发明首先构建红外图像的多尺度灰度差异图，能剔除大量的噪声干扰；其次通过点积运算获得加权局部图像熵，所得到的加权局部图像熵图具有很高的信噪比增益，能有效地抑制背景和噪声；然后利用自适应阈值检测目标，避免复杂背景条件下的图像处理不稳定和自适应性等问题。2. The present invention first constructs a multi-scale gray scale difference map of an infrared image, which can eliminate a large amount of noise interference; secondly, obtains a weighted local image entropy through a dot product operation, and the obtained weighted local image entropy map has a very high signal-to-noise ratio gain , which can effectively suppress the background and noise; and then use the adaptive threshold to detect the target, avoiding the instability and adaptiveness of image processing under complex background conditions.

附图说明Description of drawings

图1为本发明的流程框图。Fig. 1 is a flow chart of the present invention.

图2为采用本实施例1方法得到的处理结果示意图与现有技术算法的处理结果示意图的对比图。A为一幅海-空背景的小目标红外原始图像，B为采用多尺度灰度差异算子的滤波结果，C为采用局部图像熵算子的滤波结果，D为加权局部图像熵图，E为采用自适应阈值的检测结果。Fig. 2 is a comparison diagram of the processing result obtained by the method of the first embodiment and the processing result of the prior art algorithm. A is an original infrared image of a small target in the sea-air background, B is the filtering result using the multi-scale gray difference operator, C is the filtering result using the local image entropy operator, D is the weighted local image entropy map, E is the detection result using an adaptive threshold.

图3为采用现有技术和本实施例方法得到的红外图像处理结果示意图。(A_1)，(B_1)，(C_1)，(D_1)：依次为不同背景和噪声程度下的低信噪比小目标红外图像；(A_2)，(B_2)，(C_2)，(D_2)：依次对应于(A_1)，(B_1)，(C_1)，(D_1)的基于最大背景预测模型方法的滤波结果；(A_3)，(B_3)，(C_3)，(D_3)：依次对应于(A_1)，(B_1)，(C_1)，(D_1)的基于顶帽算子的滤波结果；(A_4)，(B_4)，(C_4)，(D_4)：依次对应于(A_1)，(B_1)，(C_1)，(D_1)的采用本实施例方法步骤1～步骤3的滤波结果；(A_5)，(B_5)，(C_5)，(D_5)：依次对应于(A_1)，(B_1)，(C_1)，(D_1)的基于本实施例方法的红外小目标检测结果。Fig. 3 is a schematic diagram of infrared image processing results obtained by using the prior art and the method of this embodiment. (A_1), (B_1), (C_1), (D_1): Infrared images of small targets with low SNR under different background and noise levels; (A_2), (B_2), (C_2), (D_2): Corresponding to (A_1), (B_1), (C_1), (D_1) based on the filtering results of the maximum background prediction model method; (A_3), (B_3), (C_3), (D_3): corresponding to (A_1 ), (B_1), (C_1), (D_1) based on the top-hat operator filtering results; (A_4), (B_4), (C_4), (D_4): corresponding to (A_1), (B_1), (C_1), (D_1) adopt the filtering results of steps 1 to 3 of the method of this embodiment; (A_5), (B_5), (C_5), (D_5): corresponding to (A_1), (B_1), ( C_1), (D_1) the infrared small target detection result based on the method of this embodiment.

具体实施方式Detailed ways

下面通过实施例，并结合附图，对本发明的技术方案作进一步具体的说明。The technical solutions of the present invention will be further specifically described below through the embodiments and in conjunction with the accompanying drawings.

实施例1：Example 1:

图1为本方法主要包括以下步骤：图像输入，多尺度灰度差异算子求解，局部图像熵算子求解，点积运算，自适应阈值求解，二值化。Figure 1 shows that this method mainly includes the following steps: image input, multi-scale gray difference operator solution, local image entropy operator solution, dot product operation, adaptive threshold solution, and binarization.

具体为：Specifically:

步骤1，输入一幅红外图像，求解图像的多尺度灰度差异D：Step 1, input an infrared image, and solve the multi-scale gray difference D of the image:

小目标红外图像一般由目标、背景和噪声三部分组成。小目标的成像尺寸一般小于80个像素，即小于256×256的0.12％，因而目标没有尺寸、形状和纹理等特征，但其在灰度值、频率和相关性等方面与背景、噪声存在差异。多尺度灰度差异算子(D)的核心思想是利用小目标红外图像中的目标区域与目标邻域之间的灰度差异性，通过差异性的度量以抑制背景、增强目标。Infrared images of small targets generally consist of three parts: target, background and noise. The imaging size of a small target is generally less than 80 pixels, that is, less than 0.12% of 256×256, so the target has no characteristics such as size, shape and texture, but it is different from the background and noise in terms of gray value, frequency and correlation. . The core idea of the multi-scale grayscale difference operator (D) is to use the grayscale difference between the target area and the target neighborhood in the small target infrared image to suppress the background and enhance the target through the measurement of the difference.

红外图像I的多尺度灰度差异算子D的求解过程如下：The solution process of multi-scale grayscale difference operator D of infrared image I is as follows:

(1)对于红外图像I中每一个像素点(x,y)，对应的灰度值为I(x,y)，设置像素点(x,y)的最大邻域空间Ω_max，邻域空间Ω_max的大小为L_max×L_max，其中L_max为大于1的正奇数；(1) For each pixel point (x, y) in the infrared image I, the corresponding gray value is I(x, y), set the maximum neighborhood space Ω _max of the pixel point (x, y), the neighborhood space The size of Ω _max is L _max ×L _max , where L _max is a positive odd number greater than 1;

(2)获得每一个像素点(x,y)的邻域空间集{Ω_k|k＝1,2,…,L}，其中L＝(L_max-1)/2，Ω_k的大小为(2·k+1)×(2·k+1)；(2) Obtain the neighborhood space set {Ω _k |k=1,2,...,L} of each pixel (x, y), where L=(L _max -1)/2, and the size of Ω _k is (2·k+1)×(2·k+1);

(3)计算每一个像素点(x,y)的邻域Ω_k与Ω_max之间的灰度差异D_k(x,y)，k＝1,2,…,L：(3) Calculate the gray level difference D _k (x, y) between the neighborhood Ω _k and Ω _max of each pixel (x, y), k=1,2,...,L:

${D D.}_{k k} ((x x,, y the y)) = = {| | \frac{11}{{N N}_{{Ω Ω}_{k k}}} \underset{((s the s,, t t)) &Element; &Element; {Ω Ω}_{k k}}{Σ Σ} I I ((s the s,, t t)) - - \frac{11}{{N N}_{{Ω Ω}_{max max}}} \underset{((p p,, q q)) &Element; &Element; {Ω Ω}_{max max}}{Σ Σ} I I ((p p,, q q)) | |}^{22},, k k = = 1,2 1,2,, . . . . . .,, L L - - - - - - ((11))$

其中，和分别表示邻域Ω_k、Ω_max内像素点的数目，I(s,t)表示邻域Ω_k内的点(s,t)处的灰度值，I(p,q)表示邻域Ω_max内的点(p,q)处的灰度值。in, and Respectively represent the number of pixels in the neighborhood Ω _k and Ω _max , I(s,t) represents the gray value at the point (s,t) in the neighborhood Ω _k , I(p,q) represents the neighborhood Ω The gray value at the point (p,q) within _max .

(4)计算每一个像素点(x,y)所对应的的多尺度灰度差异D(x,y)：(4) Calculate the multi-scale grayscale difference D(x,y) corresponding to each pixel (x,y):

D(x,y)＝max{D₁(x,y),D₂(x,y),...,D_L(x,y)} (2)D(x,y)=max{D ₁ (x,y),D ₂ (x,y),...,D _L (x,y)} (2)

遍历红外图像I中每一个像素点，获得红外图像I的多尺度灰度差异D(如图2的B所示)。从图2的B中可以看出，红外图像I的背景得到抑制，目标得到很好地增强。Every pixel in the infrared image I is traversed to obtain the multi-scale grayscale difference D of the infrared image I (as shown in B of FIG. 2 ). From Figure 2B, it can be seen that the background of the infrared image I is suppressed and the target is well enhanced.

步骤2，求解图像的局部图像熵E:Step 2, solve the local image entropy E of the image:

对于红外图像I的背景而言，纹理特征是确定的，当图像中出现目标时，图像的纹理特征被破坏，而小目标对于整幅图像的熵值贡献较小，但在局部窗口内，小目标的出现会引起局部纹理特征的强烈变化，因而其局部熵值也会发生较大变化。利用目标的出现会导致局部图像熵值发生较大变化这一特性可以抑制背景、增强目标。For the background of the infrared image I, the texture features are definite. When a target appears in the image, the texture feature of the image is destroyed, and the contribution of the small target to the entropy value of the whole image is small, but in the local window, the small target The appearance of the target will cause a strong change in the local texture features, so its local entropy value will also change greatly. Using the feature that the appearance of the target will cause a large change in the local image entropy value can suppress the background and enhance the target.

对于红外图像I中每一个像素点(x,y)，设置其邻域空间Θ，邻域空间Θ的大小为m×n。计算像素点(x,y)处的局部图像熵:For each pixel point (x, y) in the infrared image I, its neighborhood space Θ is set, and the size of the neighborhood space Θ is m×n. Calculate the local image entropy at the pixel point (x,y):

$E E. ((x x,, y the y)) = = - - {Σ Σ}_{i i = = 00}^{m m - - 11} {Σ Σ}_{j j = = 00}^{n no - - 11} p p ((I I ((i i,, j j)))) \cdot &Center Dot; {log log}_{22} ((p p ((I I ((i i,, j j)))) + + ϵ ϵ)),, p p ((I I ((i i,, j j)))) = = \frac{I I ((i i,, j j))}{{Σ Σ}_{i i = = 00}^{m m - - 11} {Σ Σ}_{j j = = 00}^{n no - - 11} I I ((i i,, j j))} - - - - - - ((33))$

其中，ε是预设的正常数，如ε＝10^-6，I(i,j)表示邻域Θ内的点(i,j)处的灰度值。Wherein, ε is a preset normal number, such as ε=10 ⁻⁶ , and I(i, j) represents the gray value at point (i, j) within the neighborhood Θ.

遍历红外图像I中每一个像素点，获得红外图像I的局部图像熵E(如图2的C所示)。图2的A中存在同质区域，根据最大熵原理，该区域的熵值较大，如图2的C所示的白色区域，但目标的出现引起图像局部区域的灰度特征发生变化，该灰度特征变化在图2的C中依然可见。Each pixel in the infrared image I is traversed to obtain the local image entropy E of the infrared image I (as shown in C of FIG. 2 ). There is a homogeneous area in A in Figure 2. According to the principle of maximum entropy, the entropy value of this area is relatively large, such as the white area shown in C in Figure 2, but the appearance of the target causes the grayscale characteristics of the local area of the image to change. Changes in grayscale features are still visible in C of Figure 2.

步骤3，求解图像的加权局部图像熵H:Step 3, solve the weighted local image entropy H of the image:

红外图像I的多尺度灰度差异D(如图2的B所示)和局部图像熵E(如图2的C所示)均可实现对红外图像的背景抑制和目标增强。融合D和E，使得红外图像的背景得到进一步地抑制，目标得到进一步地增强。The multi-scale grayscale difference D (shown in Figure 2 B) and the local image entropy E (shown in Figure 2 C) of the infrared image I can both achieve background suppression and target enhancement of the infrared image. By fusing D and E, the background of the infrared image is further suppressed, and the target is further enhanced.

对每一个像素点(x,y)所对应的经过步骤1处理所得到的多尺度灰度差异D与经过步骤2处理所得到的局部图像熵E进行点积运算，获得每一个像素点(x,y)所对应的加权局部图像熵H，实现对红外图像的背景进一步地抑制和目标进一步地增强，即Perform a dot product operation on the multi-scale grayscale difference D obtained by step 1 and the local image entropy E obtained by step 2 corresponding to each pixel (x, y) to obtain each pixel (x , y) corresponding to the weighted local image entropy H, to further suppress the background of the infrared image and further enhance the target, that is

$H h = = D D. &CircleTimes; &CircleTimes; E E. - - - - - - ((44))$

红外图像I的加权局部图像熵H如图2的D所示。从图2的D中可以看出，红外图像I的背景得到很好地抑制，目标也得到很好地增强。The weighted local image entropy H of the infrared image I is shown in D of Fig. 2 . From Figure 2D, it can be seen that the background of the infrared image I is well suppressed, and the target is also well enhanced.

步骤4，求解自适应阈值T：Step 4, solve the adaptive threshold T:

对经过步骤1、步骤2和步骤3处理所得到的加权局部图像熵H求解自适应阈值T，并通过自适应阈值T对加权局部图像熵H进行二值化，检测出红外小目标(二值化结果如图2的E所示)。自适应阈值T的确定方法为Solve the adaptive threshold T for the weighted local image entropy H obtained through step 1, step 2 and step 3, and binarize the weighted local image entropy H through the adaptive threshold T to detect small infrared targets (binary The results are shown in Figure 2 E). The method of determining the adaptive threshold T is

T＝c·SNR·σ+mm,SNR＝(H_max-mm)/σ (5)T＝c·SNR·σ+mm, SNR＝(H _max -mm)/σ (5)

采用不同红外图像处理方法的处理结果如图3所示，从图3中可以看出，本实施例方法得到的效果最好，其中，最大背景预测模型方法来自于文献(H.Dengand J.G.Liu,Infrared small target detection based on the self-information map,Infrared Physics&Technology,2011,54(2):100-107.)，顶帽算子方法来自于文献(X.Z.Bai and F.G.Zhou,Analysis of new top-hat transformation and theapplication for infrared dim small target detection,Pattern Recognition,2010,43(6):2145-2156.)。Adopt the processing results of different infrared image processing methods as shown in Figure 3, as can be seen from Figure 3, the effect that the present embodiment method obtains is the best, and wherein, the maximum background prediction model method comes from literature (H.Dengand J.G.Liu, Infrared small target detection based on the self-information map, Infrared Physics&Technology, 2011,54(2):100-107.), the top hat operator method comes from the literature (X.Z.Bai and F.G.Zhou, Analysis of new top-hat transformation and the application for infrared dim small target detection, Pattern Recognition, 2010, 43(6):2145-2156.).

采用信噪比(SNR,signal-to-noise ratio)来客观评价不同红外图像处理方法的滤波效果(SNR的表达式参考式(5))。具体数值见表1。The signal-to-noise ratio (SNR, signal-to-noise ratio) is used to objectively evaluate the filtering effects of different infrared image processing methods (the expression of SNR refers to formula (5)). See Table 1 for specific values.

表1 采用不同红外图像处理方法的滤波效果的SNR比较.Table 1 SNR comparison of filtering effects using different infrared image processing methods.

本文中所描述的具体实施例仅仅是对本发明精神作举例说明。本发明所属技术领域的技术人员可以对所描述的具体实施例做各种各样的修改或补充或采用类似的方式替代，但并不会偏离本发明的精神或者超越所附权利要求书所定义的范围。The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which the present invention belongs can make various modifications or supplements to the described specific embodiments or adopt similar methods to replace them, but they will not deviate from the spirit of the present invention or go beyond the definition of the appended claims range.

Claims

1., based on a small target infrared image disposal route for weighting topography entropy, it is characterized in that, comprise the following steps:

Step 1, solve the multiple dimensioned gray difference D of each pixel (x, y) of image;

Step 2, solve the topography entropy E of each pixel (x, y) of image;

Step 3, obtained the weighting topography entropy H of each pixel (x, y) by multiple dimensioned gray difference D and local image entropy E;

Step 4, solve adaptive threshold T according to weighting topography entropy H, and by adaptive threshold T, binaryzation is carried out to weighting topography entropy H, detect infrared small target.

2. a kind of small target infrared image disposal route based on weighting topography entropy according to claim 1, is characterized in that, the multiple dimensioned gray difference D of described step 1 is solved by following steps:

Step 1.1, be I (x, y) for the gray-scale value that each pixel (x, y) in infrared image I is corresponding, the maximum neighborhood space Ω of pixel (x, y) is set _max, neighborhood space Ω _maxsize be L _max× L _max, wherein L _maxfor being greater than the positive odd number of 1;

Step 1.2, obtain the neighborhood space collection { Ω of each pixel (x, y) _k| k=1,2 ..., L}, wherein L=(L _max-1)/2, Ω _ksize be (2k+1) × (2k+1);

Step 1.3, utilize the neighborhood Ω of each pixel (x, y) of following formulae discovery _kwith Ω _maxbetween gray difference D _k(x, y), k=1,2 ..., L:

D_{k} (x, y) = {| \frac{1}{N_{Ω_{k}}} \underset{(s, t) {&Element; Ω}_{k}}{Σ} I (s, t) - \frac{1}{N_{Ω_{\max}}} \underset{(p, q) {&Element; Ω}_{\max}}{Σ} I (p, q) |}^{2}, k = 1,2, . . ., L

Wherein, with represent neighborhood Ω respectively _k, Ω _maxthe number of interior pixel, I (s, t) represents neighborhood Ω _kthe gray-scale value at interior point (s, t) place, I (p, q) represents neighborhood Ω _maxthe gray-scale value at interior point (p, q) place;

Step 1.4, calculate corresponding to each pixel (x, y) multiple dimensioned gray difference D (x, y):

D(x,y)＝max{D ₁(x,y),D ₂(x,y),...,D _L(x,y)}。

3. a kind of small target infrared image disposal route based on weighting topography entropy according to claim 1, is characterized in that, the topography entropy E of described step 2 is solved by following steps:

The neighborhood space Θ of each pixel (x, y) in setting infrared image I, the size of neighborhood space Θ is m × n, calculates topography's entropy at pixel (x, y) place:

E (x, y) = - Σ_{i = 0}^{m - 1} Σ_{j = 0}^{n - 1} p (I (i, j)) \cdot \log_{2} (p (I (i, j)) + ϵ), p (I (i, j)) = \frac{I (i, j)}{Σ_{i = 0}^{m - 1} Σ_{j = 0}^{n - 1} I (i, j)}

Wherein, ε is the normal number of setting, and I (i, j) represents the gray-scale value at point (i, the j) place in neighborhood Θ, and each pixel in traversal infrared image I, obtains the topography entropy E of infrared image I.

4. a kind of small target infrared image disposal route based on weighting topography entropy according to claim 1, is characterized in that, the weighting topography entropy H of described step 3 is solved by following steps:

To each pixel (x, y) process the multiple dimensioned gray difference D that obtains through step 1 and process through step 2 the topography entropy E obtained and carry out dot-product operation, obtain the weighting topography entropy H that each pixel (x, y) is corresponding.

5. a kind of small target infrared image disposal route based on weighting topography entropy according to claim 1, it is characterized in that, described adaptive threshold T is determined by following formula:

T＝c·SNR·σ+mm,SNR＝(H _max-mm)/σ

Wherein, c is positive constant, and σ is the standard deviation of weighting topography entropy H, and mm is the average of weighting topography entropy H, H _maxfor the maximal value of weighting topography entropy H.