CN104077744A

CN104077744A - Image enhancing method and device

Info

Publication number: CN104077744A
Application number: CN201310103337.0A
Authority: CN
Inventors: 王绪四
Original assignee: Samsung Guangzhou Mobile R&D Center; Samsung Electronics Co Ltd
Current assignee: Samsung Guangzhou Mobile R&D Center; Samsung Electronics Co Ltd
Priority date: 2013-03-27
Filing date: 2013-03-27
Publication date: 2014-10-01
Anticipated expiration: 2033-03-27
Also published as: CN104077744B

Abstract

An image enhancement method and device are provided. The image enhancement method includes: performing frequency-domain transformation on the three components R, G, and B of the color image respectively to obtain frequency-domain transformation coefficients; setting the frequency-domain transformation coefficients whose absolute value is smaller than a predetermined threshold to zero, and considering the color image The continuity of the weak detail signal and the undistortion of the strong detail signal set the frequency-domain transformation coefficient whose absolute value is greater than or equal to the predetermined threshold to a value related to itself; increase the difference by extremizing the set frequency-domain transformation coefficient; And performing inverse frequency-domain transform on the extremalized frequency-domain transform coefficients to obtain three components Rn, Gn, and Bn.

Description

Image enhancement method and device

技术领域technical field

本发明涉及图像处理领域，更具体地，涉及一种图像增强方法和装置。The present invention relates to the field of image processing, and more particularly, to an image enhancement method and device.

背景技术Background technique

图像增强的目的是通过对图像进行处理来改善图像的质量和视觉效果，或者将图像转换成更适合于人眼观看或机器分析、识别的形式，以便更有效地从图像中获取有用信息。The purpose of image enhancement is to improve the quality and visual effect of the image by processing the image, or convert the image into a form that is more suitable for human viewing or machine analysis and recognition, so as to obtain useful information from the image more effectively.

通常来说，图像增强包括噪声滤波、对比度增强、饱和度增强和图像锐化等几个方面的处理。具体地，图像在获取、传输和存储过程中常常会受到各种噪声的干扰和影响而使图像降质，噪声滤波用于最大限度地降低噪声给图像带来的影响；对比度增强用于提高图像的可视度，将目标信息或隐藏的信息凸现出来；饱和度增强主要用于提高图像的层次感，让图像色彩更加艳丽；图像锐化用于增强图像的轮廓纹理，提高目标物体的清晰度，使其更加易于被检测和识别。Generally speaking, image enhancement includes noise filtering, contrast enhancement, saturation enhancement and image sharpening. Specifically, images are often disturbed and affected by various noises during the process of acquisition, transmission, and storage, which degrades the image quality. Noise filtering is used to minimize the impact of noise on the image; contrast enhancement is used to improve image quality. Visibility, to highlight the target information or hidden information; Saturation enhancement is mainly used to improve the layering of the image, making the image color more colorful; Image sharpening is used to enhance the contour texture of the image and improve the clarity of the target object , making it easier to detect and identify.

根据处理空间的不同，图像增强可分为空间域图像增强和变换域(诸如，频率域)图像增强两种。具体地，空间域图像增强主要包括直方图均衡、直方图修改等，变换域图像增强中的变换主要采用傅立叶变换、小波变换等。然而，由于它们本身的局限性或理论的缺陷，上述图像增强处理效果不佳，诸如，会造成图像模糊，丢失细节轮廓信息，图像失真严重等。According to different processing spaces, image enhancement can be divided into two types: spatial domain image enhancement and transform domain (such as frequency domain) image enhancement. Specifically, image enhancement in the spatial domain mainly includes histogram equalization, histogram modification, etc., and transformation in image enhancement in the transform domain mainly uses Fourier transform, wavelet transform, and the like. However, due to their own limitations or theoretical flaws, the above-mentioned image enhancement processes are not effective, such as blurring the image, losing details and contour information, and serious image distortion.

此外，通过便携式终端拍摄图像时，由于受到硬件或者周围环境等各方面的影响，图像质量可能会较差，诸如，图像中存在噪声、对比度较差以及图像模糊的现象，而上述常用图像增强技术均无法进行有效处理。In addition, when taking an image through a portable terminal, the image quality may be poor due to the influence of various aspects such as hardware or the surrounding environment, such as noise, poor contrast, and image blur in the image, while the above-mentioned common image enhancement technology cannot be processed effectively.

发明内容Contents of the invention

根据本发明的示例性实施例，提供了一种图像增强方法，包括：对彩色图像的三个分量R、G、B分别进行频域变换以获得频域变换系数；将绝对值小于预定阈值的频域变换系数设置为零，并考虑彩色图像弱细节信号的连续性以及强细节信号不失真将绝对值大于或等于预定阈值的频域变换系数设置为与其自身相关的值；通过将设置的频域变换系数极值化以增大差异性；以及对极值化的频域变换系数进行频域逆变换以获得三个分量Rn、Gn、Bn。According to an exemplary embodiment of the present invention, an image enhancement method is provided, including: performing frequency-domain transformation on three components R, G, and B of a color image respectively to obtain frequency-domain transformation coefficients; The frequency domain transformation coefficient is set to zero, and considering the continuity of the weak detail signal of the color image and the strong detail signal without distortion, the frequency domain transformation coefficient whose absolute value is greater than or equal to the predetermined threshold is set to a value related to itself; by setting the frequency domain extremizing the domain transform coefficients to increase the difference; and performing inverse frequency domain transform on the extremalized frequency domain transform coefficients to obtain three components Rn, Gn, Bn.

所述图像增强方法还可包括：在对彩色图像的三个分量R、G、B分别进行频域变换之前，对三个分量R、G、B进行归一化处理。The image enhancement method may further include: performing normalization processing on the three components R, G, B of the color image before performing frequency domain transformation on the three components R, G, B respectively.

所述频域变换可以是轮廓波(Contourlet)变换，所述频域变换系数可以是Contourlet变换系数。The frequency domain transform may be a contourlet (Contourlet) transform, and the frequency domain transform coefficients may be Contourlet transform coefficients.

设置频域变换系数的步骤可包括：通过以下等式来设置频域变换系数，The step of setting the frequency-domain transform coefficient may include: setting the frequency-domain transform coefficient by the following equation,

$Coeffδ Coeffδ = = \{\begin{matrix} sgn sgn ((Coeff Coeff)) ((| | Coeff Coeff | | - - ((\frac{max max ((| | Coeff Coeff | |)) - - | | Coeff Coeff | |}{max max ((| | Coeff Coeff | |)) - - δ δ})) ((\frac{22^{\frac{{((mean mean 22 ((\frac{R R + + G G + + B B}{33}))))}^{22}}{22 \cdot \cdot ((std std 22 ((\frac{R R + + G G + + B B}{33})))) \cdot \cdot σ σ} \cdot \cdot δ δ}}{22^{\frac{{((mean mean 22 ((\frac{R R + + G G + + B B}{33}))))}^{22}}{22 \cdot \cdot ((std std 22 ((\frac{R R + + G G + + B B}{33})))) \cdot \cdot σ σ} \cdot &Center Dot; | | Coeff Coeff | |}})) δ δ)),, & | | Coeff Coeff | | &GreaterEqual; &Greater Equal; δ δ \\ 00,, & | | Coeff Coeff | | < < δ δ \end{matrix}$

其中，δ表示预定阈值，Coeff表示Contourlet变换系数，Coeffδ表示设置后的Coeff，sgn表示取符号运算，max表示取最大值运算，mean2表示取均值运算，std2表示取标准差运算，σ表示R、G、B的噪声标准方差估计值，其值通过等式进行计算，其中，Median表示取中值运算，Coeff1表示Coeff的第一层系数，r的值为0.6745。Among them, δ represents the predetermined threshold, Coeff represents the Contourlet transformation coefficient, Coeffδ represents the Coeff after setting, sgn represents the symbol operation, max represents the maximum value operation, mean2 represents the mean value operation, std2 represents the standard deviation operation, σ represents R, The estimated value of the noise standard deviation of G and B, whose value is given by the equation Perform calculations, where Median represents the median value operation, Coeff1 represents the coefficient of the first layer of Coeff, and the value of r is 0.6745.

δ可采用贝叶斯收缩(BayesShrink)阈值。δ can adopt Bayesian Shrink (BayesShrink) threshold.

δ可采用自适应贝叶斯收缩(BayesShrink)阈值。δ can adopt an adaptive Bayesian shrinkage (BayesShrink) threshold.

将设置的频域变换系数极值化的步骤可包括：将设置的Contourlet变换系数中的第一层系数极值化。The step of extremizing the set frequency-domain transform coefficients may include: extremizing the first layer coefficients among the set Contourlet transform coefficients.

将设置的Contourlet变换系数中的第一层系数极值化的步骤可包括：通过预定矩阵对由Coeffδ的转换到二维的第一层系数中相同方向的系数构成的多个子图的每一个选择二维区域；通过均值矩阵对每个选择的二维区域进行卷积处理以获得卷积值；对于每个选择的二维区域，当选择的二维区域的中心点的值大于或等于所述卷积值时，中心点的值取选择的二维区域中的最大值，当选择的二维区域的中心点的值小于所述卷积值时，中心点的值取选择的二维区域中的最小值。The step of extremizing the first layer coefficients in the set Contourlet transform coefficients may include: selecting each of a plurality of subgraphs composed of coefficients in the same direction in the first layer coefficients converted from Coeff δ to two dimensions by a predetermined matrix A two-dimensional area; each selected two-dimensional area is convoluted by a mean matrix to obtain a convolution value; for each selected two-dimensional area, when the value of the center point of the selected two-dimensional area is greater than or equal to the When convolving the value, the value of the center point takes the maximum value in the selected two-dimensional area. When the value of the center point of the selected two-dimensional area is smaller than the convolution value, the value of the center point takes the value of the selected two-dimensional area minimum value.

所述图像增强方法还可包括：通过使用三个分量Rn、Gn、Bn进行空间域图像增强。The image enhancement method may further include: performing spatial domain image enhancement by using three components Rn, Gn, Bn.

通过使用三个分量Rn、Gn、Bn进行空间域图像增强的步骤可包括：通过金字塔形式的矩阵分别对三个分量Rn、Gn、Bn进行卷积处理以获得RnF、GnF、BnF，并通过以下等式来获得空间域图像增强的三个分量RNS、GNS、BNS，The step of performing space-domain image enhancement by using three components Rn, Gn, Bn may include: respectively performing convolution processing on the three components Rn, Gn, Bn through a matrix in the form of a pyramid to obtain RnF, GnF, BnF, and through the following Equation to obtain the three components RNS, GNS, BNS of spatial domain image enhancement,

$\{\begin{matrix} RNS RNS ((i i,, j j)) = = Rn n ((i i,, j j)) \cdot \cdot ((11 + + t t \cdot \cdot ((11 - - \frac{Rn n ((i i,, j j)) + + Gn Gn ((i i,, j j)) + + Bn Bn ((i i,, j j))}{33 \cdot \cdot Rn n ((i i,, j j))})))) \cdot \cdot {((\sqrt{\frac{Rn n ((i i,, j j))}{RnF nF ((i i,, j j))}}))}^{11 &PlusMinus; &PlusMinus; ζ ζ} \\ GNS GPS ((i i,, j j)) = = Gn Gn ((i i,, j j)) \cdot \cdot ((11 + + t t \cdot &Center Dot; ((11 - - \frac{Rn Rn ((i i,, j j)) + + Gn Gn ((i i,, j j)) + + Bn Bn ((i i,, j j))}{33 \cdot &Center Dot; Gn Gn ((i i,, j j))})))) \cdot \cdot {((\sqrt{\frac{Gn Gn ((i i,, j j))}{GnF GnF ((i i,, j j))}}))}^{11 &PlusMinus; &PlusMinus; ζ ζ} \\ BNS BNS ((i i,, j j)) = = Bn Bn ((i i,, j j)) \cdot &Center Dot; ((11 + + t t \cdot &Center Dot; ((11 - - \frac{Rn n ((i i,, j j)) + + Gn Gn ((i i,, j j)) + + Bn Bn ((i i,, j j))}{33 \cdot &Center Dot; Bn Bn ((i i,, j j))})))) \cdot &Center Dot; {((\sqrt{\frac{Bn Bn ((i i,, j j))}{BnF BnF ((i i,, j j))}}))}^{11 &PlusMinus; &PlusMinus; ζ ζ} \end{matrix}$

其中，Rn(i，j)、Gn(i，j)和Bn(i，j)分别表示在Rn、Gn、Bn中坐标(i，j)上的值，RnF(i，j)、GnF(i，j)、BnF(i，j)分别表示在RnF、GnF、BnF中坐标(i，j)上的值，RNS(i，j)、GNS(i，j)、BNS(i，j)分别表示在RNS、GNS、BNS中坐标(i，j)上的值，t表示整体饱和度调节因子，ζ表示局部对比度调节因子，并且当RNS(i，j)、GNS(i，j)、BNS(i，j)超出预定范围时，将RNS(i，j)、GNS(i，j)、BNS(i，j)限制为预定范围的相应端点值。Among them, Rn(i, j), Gn(i, j) and Bn(i, j) respectively represent the values on coordinates (i, j) in Rn, Gn, Bn, RnF(i, j), GnF( i, j), BnF(i, j) represent the value on coordinate (i, j) in RnF, GnF, BnF respectively, RNS(i, j), GNS(i, j), BNS(i, j) Respectively represent the value on the coordinate (i, j) in RNS, GNS, BNS, t represents the overall saturation adjustment factor, ζ represents the local contrast adjustment factor, and when RNS (i, j), GNS (i, j), When BNS(i, j) exceeds the predetermined range, limit RNS(i, j), GNS(i, j), and BNS(i, j) to corresponding endpoint values of the predetermined range.

彩色图像可通过便携式终端获得。Color images are available via portable terminals.

根据本发明的示例性实施例，提供了一种图像增强装置，包括：频域变换单元，对彩色图像的三个分量R、G、B分别进行频域变换以获得频域变换系数；阈值处理单元，将绝对值小于预定阈值的频域变换系数设置为零，并考虑彩色图像弱细节信号的连续性以及强细节信号不失真将绝对值大于或等于预定阈值的频域变换系数设置为与其自身相关的值；极大极小处理单元，通过将设置的频域变换系数极值化以增大差异性；以及频域逆变换单元，对极值化的频域变换系数进行频域逆变换以获得三个分量Rn、Gn、Bn。According to an exemplary embodiment of the present invention, an image enhancement device is provided, including: a frequency domain transformation unit, which performs frequency domain transformation on the three components R, G, and B of the color image respectively to obtain frequency domain transformation coefficients; threshold value processing The unit is to set the frequency-domain transformation coefficient whose absolute value is less than the predetermined threshold to zero, and consider the continuity of the weak detail signal of the color image and the undistorted strong detail signal to set the frequency-domain transformation coefficient whose absolute value is greater than or equal to the predetermined threshold to be equal to itself related values; the maxima-minimum processing unit, which increases the difference by extremizing the set frequency domain transform coefficients; and the frequency domain inverse transform unit, which performs frequency domain inverse transform on the extremized frequency domain transform coefficients to Three components Rn, Gn, Bn are obtained.

所述图像增强装置还可包括：空间域图像增强单元，通过使用三个分量Rn、Gn、Bn进行空间域图像增强。The image enhancement device may further include: a spatial domain image enhancement unit, which performs spatial domain image enhancement by using three components Rn, Gn, and Bn.

将在接下来的描述中部分阐述本发明另外的方面和/或优点，还有一部分通过描述将是清楚的，或者可以经过本发明的实施而得知。Additional aspects and/or advantages of the present invention will be set forth in part in the following description, and some will be clear from the description, or can be learned through practice of the present invention.

附图说明Description of drawings

通过下面结合附图进行的详细描述，本发明的上述和其它目的和特点将会变得更加清楚，其中：The above-mentioned and other objects and features of the present invention will become clearer through the following detailed description in conjunction with the accompanying drawings, wherein:

图1是示出根据本发明示例性实施例的图像增强方法的流程图；FIG. 1 is a flowchart illustrating an image enhancement method according to an exemplary embodiment of the present invention;

图2是示出根据本发明示例性实施例的图1中的极值化步骤的流程图；FIG. 2 is a flow chart illustrating the extremization step in FIG. 1 according to an exemplary embodiment of the present invention;

图3是示出根据本发明示例性实施例的图像增强装置的框图。FIG. 3 is a block diagram illustrating an image enhancement device according to an exemplary embodiment of the present invention.

具体实施方式Detailed ways

现在，详细描述本发明的示例性实施例，其示例在附图中表示，其中，相同的标号始终表示相同的部件。Exemplary embodiments of the present invention will now be described in detail, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to like parts throughout.

图1是示出根据本发明示例性实施例的图像增强方法的流程图。FIG. 1 is a flowchart illustrating an image enhancement method according to an exemplary embodiment of the present invention.

参照图1，在步骤S110，可对彩色图像I(三维矩阵m×n×3，其中，m、n为正整数)进行RGB空间分解，以获得三个分量R、G和B，所述三个分量R、G和B均为二维矩阵。具体地，可根据以下等式1来进行RGB空间分解。Referring to Fig. 1, in step S110, the color image I (three-dimensional matrix m×n×3, wherein m and n are positive integers) can be decomposed into RGB space to obtain three components R, G and B, the three The components R, G and B are all two-dimensional matrices. Specifically, RGB space decomposition may be performed according to Equation 1 below.

$\{\begin{matrix} R R ((i i,, j j)) = = I I ((i i,, j j,, 11)) \\ G G ((i i,, j j)) = = I I ((i i,, j j,, 22)) \\ B B ((i i,, j j)) = = I I ((i i,, j j,, 33)) \end{matrix} - - - - - - ((11))$

其中，R(i，j)、G(i，j)和B(i，j)均为整数，分别表示在二维矩阵R、G和B中坐标(i，j)上的值，且取值范围均为[0，255]，i为整数，取值范围为[0，m-1]，j为整数，取值范围为[0，n-1]。Among them, R(i, j), G(i, j) and B(i, j) are all integers, representing the values on the coordinates (i, j) in the two-dimensional matrices R, G and B respectively, and take The value range is [0, 255], i is an integer, the value range is [0, m-1], j is an integer, the value range is [0, n-1].

在步骤S120，可分别对三个分量R、G、B进行归一化处理。具体地，可分别将三个分量R、G、B中的每个元素除以一个预定值来进行归一化处理，这里，所述预定值可以但不限于是255。In step S120, normalization processing may be performed on the three components R, G, and B respectively. Specifically, each element of the three components R, G, and B may be divided by a predetermined value to perform normalization processing. Here, the predetermined value may be but not limited to 255.

应该了解，上述步骤S110、S120并是不必要步骤，诸如，在已提前获知彩色图像的三个分量R、G、B的情况下，可省略上述步骤S110，并且本领域技术人员可根据实际需要而将仅为了使计算方便而进行的步骤S120省略。It should be understood that the above-mentioned steps S110 and S120 are unnecessary steps. For example, in the case where the three components R, G, and B of the color image have been known in advance, the above-mentioned step S110 can be omitted, and those skilled in the art can The step S120 performed only for the convenience of calculation is omitted.

在步骤S130，对彩色图像的三个分量R、G、B分别进行频域变换以获得频域变换系数。这里，仅作为示例，所述频域变换可以是轮廓波(Contourlet)变换，所述频域变换系数即为Contourlet变换系数，相应地，对彩色图像的三个分量R、G、B分别进行Contourlet变换而获得的Contourlet变换系数可分别为Rcoeff、Gcoeff和Bcoeff，它们均为一维向量。以下，为描述方便，将频域变换描述为Contourlet变换，但应该了解，本发明还可采用其它的频域变换，诸如，小波变换、非采样轮廓波变换(NSCT)或其它多分辨率几何分析工具等。In step S130, three components R, G, and B of the color image are respectively subjected to frequency-domain transformation to obtain frequency-domain transformation coefficients. Here, as an example only, the frequency-domain transform may be a contourlet (Contourlet) transform, and the frequency-domain transform coefficients are Contourlet transform coefficients. Correspondingly, the three components R, G, and B of the color image are respectively subjected to Contourlet The Contourlet transformation coefficients obtained by the transformation can be respectively Rcoeff, Gcoeff and Bcoeff, all of which are one-dimensional vectors. Hereinafter, for the convenience of description, the frequency domain transform is described as Contourlet transform, but it should be understood that the present invention can also adopt other frequency domain transforms, such as wavelet transform, non-sampling contourlet transform (NSCT) or other multi-resolution geometric analysis tools etc.

在步骤S140，将绝对值小于预定阈值的Contourlet变换系数设置为零，并考虑彩色图像弱细节信号的连续性以及强细节信号不失真将绝对值大于或等于预定阈值的Contourlet变换系数设置为与其自身相关的值。具体地，可通过以下等式2来设置Contourlet变换系数。In step S140, the Contourlet transform coefficients whose absolute values are smaller than the predetermined threshold are set to zero, and the Contourlet transform coefficients whose absolute values are greater than or equal to the predetermined threshold are set to be equal to themselves in consideration of the continuity of the weak detail signal of the color image and the non-distortion of the strong detail signal associated value. Specifically, the Contourlet transform coefficient can be set by Equation 2 below.

$Coeffδ Coeffδ = = \{\begin{matrix} sgn sgn ((Coeff Coeff)) ((| | Coeff Coeff | | - - ((\frac{max max ((| | Coeff Coeff | |)) - - | | Coeff Coeff | |}{max max ((| | Coeff Coeff | |)) - - δ δ})) ((\frac{22^{\frac{{((mean mean 22 ((\frac{R R + + G G + + B B}{33}))))}^{22}}{22 \cdot &Center Dot; ((std std 22 ((\frac{R R + + G G + + B B}{33})))) \cdot &Center Dot; σ σ} \cdot &Center Dot; δ δ}}{22^{\frac{{((mean mean 22 ((\frac{R R + + G G + + B B}{33}))))}^{22}}{22 \cdot &Center Dot; ((std std 22 ((\frac{R R + + G G + + B B}{33})))) \cdot &Center Dot; σ σ} \cdot &Center Dot; | | Coeff Coeff | |}})) δ δ)),, & | | Coeff Coeff | | &GreaterEqual; &Greater Equal; δ δ \\ 00,, & | | Coeff Coeff | | < < δ δ \end{matrix} - - - - - - ((22))$

其中，δ表示预定阈值，Coeff(即，Rcoeff、Gcoeff和Bcoeff的统一表达)表示Contourlet变换系数，Coeffδ(即，Rcoeffδ、Gcoeffδ和Bcoeffδ的统一表达)表示设置后的Coeff，sgn表示取符号运算，max表示取最大值运算，mean2表示取均值运算，std2表示取标准差运算，σ表示R、G、B的噪声标准方差估计值，其值通过等式进行计算，其中，Median表示取中值运算，Coeff1表示Coeff的第一层系数，r的值为0.6745。Among them, δ represents the predetermined threshold, Coeff (that is, the unified expression of Rcoeff, Gcoeff and Bcoeff) represents the Contourlet transformation coefficient, Coeffδ (that is, the unified expression of Rcoeffδ, Gcoeffδ and Bcoeffδ) represents the set Coeff, and sgn represents the sign operation, max means to take the maximum value operation, mean2 means to take the mean value operation, std2 means to take the standard deviation operation, σ means the estimated value of the noise standard deviation of R, G, and B, and its value is passed through the equation Perform calculations, where Median represents the median value operation, Coeff1 represents the coefficient of the first layer of Coeff, and the value of r is 0.6745.

应该理解，上述等式2仅是设置Contourlet变换系数的示例，本领域技术人员完全可采用其它方案来设置Contourlet变换系数，诸如，可采用硬阈值函数、软阈值函数、折中阈值函数和半软阈值函数等，分别如下等式所示：It should be understood that the above equation 2 is only an example of setting the Contourlet transform coefficients, and those skilled in the art can completely adopt other schemes to set the Contourlet transform coefficients, such as hard threshold function, soft threshold function, compromise threshold function and semi-soft threshold function. Threshold function, etc., respectively, as shown in the following equations:

${w w}_{δ δ} = = \{\begin{matrix} w w,, & | | w w | | &GreaterEqual; &Greater Equal; δ δ \\ 00,, & | | w w | | < < δ δ \end{matrix}$

${w w}_{δ δ} = = \{\begin{matrix} sgn sgn ((w w)) ((| | w w | | - - δ δ)),, & | | w w | | &GreaterEqual; &Greater Equal; δ δ \\ 00,, & | | w w | | < < δ δ \end{matrix}$

${w w}_{δ δ} = = \{\begin{matrix} sgn sgn ((w w)) ((| | w w | | - - γ γ \cdot \cdot δ δ)),, & | | w w | | &GreaterEqual; &Greater Equal; δ δ \\ 00,, & | | w w | | < < δ δ \end{matrix}$

${w w}_{δ δ} = = \{\begin{matrix} w w,, & | | w w | | &GreaterEqual; &Greater Equal; {δ δ}_{22} \\ sgn sgn ((w w)) \frac{{δ δ}_{22} ((| | w w | | - - {δ δ}_{11}))}{{δ δ}_{22} - - {δ δ}_{11}},, & {δ δ}_{11} \leq \leq | | w w | | < < {δ δ}_{22} \\ 00,, & | | w w | | < < {δ δ}_{11} \end{matrix}$

其中，δ、δ₁、δ₂分别表示阈值，w表示Contourlet系数，w_δ表示设置后的Contourlet系数，γ的取值范围为[0，1]。Among them, δ, δ ₁ , and δ ₂ represent thresholds respectively, w represents the Contourlet coefficient, w _δ represents the set Contourlet coefficient, and the value range of γ is [0, 1].

此外，对于等式2中的预定阈值δ，可采用贝叶斯收缩(BayesShrink)阈值或自适应BayesShrink阈值。BayesShrink阈值的等式为其中，为R、G、B的噪声标准方差估计值，σ_s为R、G、B的标准差估计值；自适应BayesShrink阈值的等式为其中，表示通过BayesShrink阈值估计的Contourlet变换的第1层j方向的阈值，其值通过等式进行计算，其中，Median表示取中值运算，表示Contourlet变换的第1层j方向的高频系数，r的值为0.6745；表示第1层j方向的自适应阈值，J₁表示第1层的总方向数，表示第1层j方向的能量值，η通常为min(J₁)，min表示取最小值运算。In addition, for the predetermined threshold δ in Equation 2, a Bayesian shrinkage (BayesShrink) threshold or an adaptive BayesShrink threshold may be used. The equation for the BayesShrink threshold is in, is the estimated value of noise standard deviation of R, G, B, σ _s is the estimated value of standard deviation of R, G, B; the equation of adaptive BayesShrink threshold is in, Represents the threshold in the j-direction of layer 1 of the Contourlet transform estimated by the BayesShrink threshold, whose value is passed by the equation Calculate, where Median represents the median operation, Represents the high-frequency coefficient of the first layer j direction of the Contourlet transform, and the value of r is 0.6745; Indicates the adaptive threshold of the first layer j direction, J ₁ represents the total number of directions in the first layer, Indicates the energy value in the j-direction of the first layer, η is usually min(J ₁ ), and min represents the minimum value operation.

上述BayesShrink阈值和自适应BayesShrink阈值均为现有技术中的阈值，因此这里将不作详细描述。此外，应该了解，除了使用BayesShrink阈值或自适应BayesShrink阈值之外，本领域技术人员还可根据实际需要采用其它阈值，诸如，统一(VisuShrink)阈值、基于零均值正态分布的置信区间阈值等。Both the aforementioned BayesShrink threshold and the adaptive BayesShrink threshold are thresholds in the prior art, and thus will not be described in detail here. In addition, it should be understood that, in addition to using the BayesShrink threshold or the adaptive BayesShrink threshold, those skilled in the art can also adopt other thresholds according to actual needs, such as a unified (VisuShrink) threshold, a confidence interval threshold based on a zero-mean normal distribution, and the like.

在步骤S150，通过将设置的Contourlet变换系数极值化以增大差异性。下面将参照图2对将设置的Contourlet变换系数极值化的步骤进行更详细地描述。In step S150, the difference is increased by extremizing the set Contourlet transform coefficients. The step of extremalizing the set Contourlet transform coefficients will be described in more detail below with reference to FIG. 2 .

图2是示出根据本发明示例性实施例的图1中的极值化步骤的流程图。FIG. 2 is a flowchart illustrating an extremization step in FIG. 1 according to an exemplary embodiment of the present invention.

参照图2，在步骤S151，通过预定矩阵对由Coeffδ的转换到二维的第一层系数中相同方向的系数构成的多个子图的每一个中的每个点选择以所述每个点为中心的二维区域。With reference to Fig. 2, in step S151, select each point in each of the plurality of subgraphs that are formed by the coefficients of the same direction in the coefficients of the first layer of coefficients in the two-dimensional conversion of Coeff δ by a predetermined matrix to select each point as A two-dimensional region at the center.

在步骤S152，通过均值矩阵对每个选择的二维区域进行卷积处理以获得卷积值，其中，所述均值矩阵可以为奇数乘奇数形式的二维矩阵，其中的元素取值可相同，并且所有元素的和可以为1。In step S152, each selected two-dimensional region is convoluted by means of a mean matrix to obtain a convolution value, wherein the mean value matrix can be a two-dimensional matrix in the form of an odd number multiplied by an odd number, and the values of the elements therein can be the same, And the sum of all elements can be 1.

在步骤S153，对于每个选择的二维区域，当选择的二维区域的中心点的值大于或等于所述卷积值时，中心点的值取选择的二维区域中的最大值，当选择的二维区域的中心点的值小于所述卷积值时，中心点的值取选择的二维区域中的最小值。In step S153, for each selected two-dimensional area, when the value of the center point of the selected two-dimensional area is greater than or equal to the convolution value, the value of the center point takes the maximum value in the selected two-dimensional area, when When the value of the center point of the selected two-dimensional area is smaller than the convolution value, the value of the center point takes the minimum value in the selected two-dimensional area.

诸如，假设由Rcoeffδ的转换到二维的第一层系数中相同方向的系数构成的一个子图为 $[\begin{matrix} 0.1 & 0.1 & 0.1 & 0.1 & 0.1 \\ 0.2 & 0.2 & 0.2 & 0.2 & 0.2 \\ 0.3 & 0.3 & 0.6 & 0.3 & 0.3 \\ 0.4 & 0.4 & 0.4 & 0.4 & 0.4 \\ 0.5 & 0.5 & 0.5 & 0.5 & 0.5 \end{matrix}],$ 预定矩阵为3×3形式矩阵，如果对于在位置(3，3)的点0.6，则选择的二维区域为 $[\begin{matrix} 0.2 & 0.2 & 0.2 \\ 0.3 & 0.6 & 0.3 \\ 0.4 & 0.4 & 0.4 \end{matrix}],$ 如果对于在位置(1，1)的点0.1，则选择的二维区域为 $[\begin{matrix} 0 & 0 & 0 \\ 0 & 0.1 & 0.1 \\ 0 & 0.2 & 0.2 \end{matrix}],$ 等等。假设均值矩阵为 $[\begin{matrix} \frac{1}{9} & \frac{1}{9} & \frac{1}{9} \\ \frac{1}{9} & \frac{1}{9} & \frac{1}{9} \\ \frac{1}{9} & \frac{1}{9} & \frac{1}{9} \end{matrix}],$ 则对所述位置(3，3)的点0.6的二维区域进行卷积0.2*1/9+0.2*1/9+0.2*1/9+0.3*1/9+0.3*1/9+0.4*1/9+0.4*1/9+0.4*1/9+0.6*1/9以获得卷积值1/3。由于所述二维区域的中心点的值0.6大于1/3，故中心点的值取选择的二维区域中的最大值，即，0.6。For example, assuming that a subgraph composed of coefficients in the same direction in the first layer of coefficients transformed from Rcoeffδ to two-dimensional is $[\begin{matrix} 0.1 & 0.1 & 0.1 & 0.1 & 0.1 \\ 0.2 & 0.2 & 0.2 & 0.2 & 0.2 \\ 0.3 & 0.3 & 0.6 & 0.3 & 0.3 \\ 0.4 & 0.4 & 0.4 & 0.4 & 0.4 \\ 0.5 & 0.5 & 0.5 & 0.5 & 0.5 \end{matrix}],$ The predetermined matrix is a matrix in the form of 3×3. If the point at position (3, 3) is 0.6, the selected two-dimensional area is $[\begin{matrix} 0.2 & 0.2 & 0.2 \\ 0.3 & 0.6 & 0.3 \\ 0.4 & 0.4 & 0.4 \end{matrix}],$ If for a point 0.1 at position (1, 1), the selected two-dimensional area is $[\begin{matrix} 0 & 0 & 0 \\ 0 & 0.1 & 0.1 \\ 0 & 0.2 & 0.2 \end{matrix}],$ etc. Suppose the mean matrix is $[\begin{matrix} \frac{1}{9} & \frac{1}{9} & \frac{1}{9} \\ \frac{1}{9} & \frac{1}{9} & \frac{1}{9} \\ \frac{1}{9} & \frac{1}{9} & \frac{1}{9} \end{matrix}],$ Then perform convolution on the two-dimensional area of point 0.6 at the position (3, 3) 0.2*1/9+0.2*1/9+0.2*1/9+0.3*1/9+0.3*1/9+ 0.4*1/9+0.4*1/9+0.4*1/9+0.6*1/9 to get the convolution value 1/3. Since the value 0.6 of the center point of the two-dimensional area is greater than 1/3, the value of the center point is the maximum value in the selected two-dimensional area, ie, 0.6.

应该理解，上述极值化处理仅是示例性的，本领域技术人员完全可根据实际需要进行其它算法或方案的极值化处理。It should be understood that the foregoing extremization processing is only exemplary, and those skilled in the art may perform extremization processing with other algorithms or schemes according to actual needs.

参照回图1，在步骤S160，对极值化的Contourlet变换系数进行Contourlet逆变换以获得三个分量Rn、Gn、Bn，所述三个分量Rn、Gn、Bn均为二维矩阵。Referring back to FIG. 1 , in step S160 , inverse Contourlet transform is performed on the extremalized Contourlet transform coefficients to obtain three components Rn, Gn, Bn, all of which are two-dimensional matrices.

在步骤S170，可通过使用三个分量Rn、Gn、Bn进行空间域图像增强。具体地，可通过金字塔形式的矩阵分别对三个分量Rn、Gn、Bn进行卷积处理以获得三个分量RnF、GnF、BnF，并进而通过以下等式3来获得空间域图像增强的三个分量RNS、GNS、BNS，其中，所述金字塔形式的矩阵可按照金字塔形式选定元素的值，即，矩阵中间的值最大，且离中心越远的值越小，并且所有元素的和可以为1，诸如， $[\begin{matrix} \frac{1}{16} & \frac{1}{8} & \frac{1}{16} \\ \frac{1}{8} & \frac{1}{4} & \frac{1}{8} \\ \frac{1}{16} & \frac{1}{8} & \frac{1}{16} \end{matrix}] .$ In step S170, spatial domain image enhancement may be performed by using three components Rn, Gn, Bn. Specifically, the three components Rn, Gn, and Bn can be convoluted through a matrix in the form of a pyramid to obtain the three components RnF, GnF, and BnF, and then the three components of spatial domain image enhancement can be obtained by the following Equation 3: Component RNS, GNS, BNS, wherein, the matrix of the pyramid form can select the value of the element according to the pyramid form, that is, the value in the middle of the matrix is the largest, and the value farther away from the center is smaller, and the sum of all elements can be 1, such as, $[\begin{matrix} \frac{1}{16} & \frac{1}{8} & \frac{1}{16} \\ \frac{1}{8} & \frac{1}{4} & \frac{1}{8} \\ \frac{1}{16} & \frac{1}{8} & \frac{1}{16} \end{matrix}] .$

$\{\begin{matrix} RNS RNS ((i i,, j j)) = = Rn n ((i i,, j j)) \cdot &Center Dot; ((11 + + t t \cdot \cdot ((11 - - \frac{Rn n ((i i,, j j)) + + Gn Gn ((i i,, j j)) + + Bn Bn ((i i,, j j))}{33 \cdot \cdot Rn n ((i i,, j j))})))) \cdot &Center Dot; {((\sqrt{\frac{Rn n ((i i,, j j))}{RnF nF ((i i,, j j))}}))}^{11 &PlusMinus; &PlusMinus; ζ ζ} \\ GNS GPS ((i i,, j j)) = = Gn Gn ((i i,, j j)) \cdot \cdot ((11 + + t t \cdot \cdot ((11 - - \frac{Rn n ((i i,, j j)) + + Gn Gn ((i i,, j j)) + + Bn Bn ((i i,, j j))}{33 \cdot \cdot Gn Gn ((i i,, j j))})))) \cdot \cdot {((\sqrt{\frac{Gn Gn ((i i,, j j))}{GnF GnF ((i i,, j j))}}))}^{11 &PlusMinus; &PlusMinus; ζ ζ} \\ BNS BNS ((i i,, j j)) = = Bn Bn ((i i,, j j)) \cdot \cdot ((11 + + t t \cdot \cdot ((11 - - \frac{Rn n ((i i,, j j)) + + Gn Gn ((i i,, j j)) + + Bn Bn ((i i,, j j))}{33 \cdot &Center Dot; Bn Bn ((i i,, j j))})))) \cdot \cdot {((\sqrt{\frac{Bn Bn ((i i,, j j))}{BnF BnF ((i i,, j j))}}))}^{11 &PlusMinus; &PlusMinus; ζ ζ} \end{matrix} - - - - - - ((33))$

其中，Rn(i，j)、Gn(i，j)、Bn(i，j)分别表示在二维矩阵Rn、Gn、Bn中坐标(i，j)上的值；相应地，RnF(i，j)、GnF(i，j)、BnF(i，j)分别表示在二维矩阵RnF、GnF、BnF中坐标(i，j)上的值；RNS、GNS、BNS分别表示空间域图像增强后的三个分量Rn、Gn、Bn，RNS(i，j)、GNS(i，j)、BNS(i，j)分别表示在二维矩阵RNS、GNS、BNS中坐标(i，j)上的值；t表示整体饱和度调节因子，通常可根据实际需要取值为较小的非负数，诸如0.1；ζ表示局部对比度调节因子，取值范围为[0，1]，在对其进行设置时，可通过判断彩色图像中心点与邻近区域的相似度(或差异度)来自行调整，以实现彩色图像局部增强，而对于特殊图像或特殊要求，可调节ζ控制彩色图像局部增强的程度，保证彩色图像不失真；当RNS(i，j)、GNS(i，j)、BNS(i，j)超出预定范围时，可将RNS(i，j)、GNS(i，j)、BNS(i，j)限制为所述预定范围的相应端点值。Among them, Rn(i, j), Gn(i, j), Bn(i, j) respectively represent the values on the coordinates (i, j) in the two-dimensional matrix Rn, Gn, Bn; correspondingly, RnF(i , j), GnF(i, j), BnF(i, j) represent the values on the coordinates (i, j) in the two-dimensional matrix RnF, GnF, BnF respectively; RNS, GNS, BNS respectively represent the spatial domain image enhancement The last three components Rn, Gn, Bn, RNS(i, j), GNS(i, j), BNS(i, j) are respectively represented on the coordinates (i, j) in the two-dimensional matrix RNS, GNS, BNS The value of ; t represents the overall saturation adjustment factor, which can usually be a small non-negative number according to actual needs, such as 0.1; ζ represents the local contrast adjustment factor, the value range is [0, 1], it is set , it can be adjusted by judging the similarity (or difference) between the center point of the color image and the adjacent area to achieve local enhancement of the color image. For special images or special requirements, you can adjust the degree of local enhancement of the color image. Ensure that the color image is not distorted; when RNS(i, j), GNS(i, j), BNS(i, j) exceeds the predetermined range, RNS(i, j), GNS(i, j), BNS( i, j) are limited to respective endpoint values of said predetermined range.

应该了解，上述空间域图像增强处理仅是示例性的，本领域技术人员完全可根据实际需要进行其它算法或方案的空间域图像增强处理，诸如，饱和度提升、对比度增强等。It should be understood that the above spatial domain image enhancement processing is only exemplary, and those skilled in the art may perform other algorithms or schemes of spatial domain image enhancement processing according to actual needs, such as saturation enhancement, contrast enhancement, and the like.

此外，如果前面进行了步骤S110和S120，则可在步骤S180，将RNS、GNS、BNS恢复为彩色图像Inew(三维矩阵m×n×3，其中，m、n为正整数)。具体地，可根据以下等式4来恢复为彩色图像Inew。In addition, if steps S110 and S120 have been performed before, then in step S180, the RNS, GNS, and BNS can be restored to a color image Inew (three-dimensional matrix m×n×3, where m and n are positive integers). Specifically, the color image Inew can be restored according to Equation 4 below.

$\{\begin{matrix} Inew Inew ((i i,, j j,, 11)) = = RNS RNS ((i i,, j j)) \times \times 255255 \\ Inew Inew ((i i,, j j,, 22)) = = GNS GPS ((i i,, j j)) \times \times 255255 \\ Inew Inew ((i i,, j j,, 33)) = = BNS BNS ((i i,, j j)) \times \times 255255 \end{matrix} - - - - - - ((44))$

其中，i为整数，取值范围[0，m-1]，j为整数，取值范围[0，n-1]。Wherein, i is an integer with a value range of [0, m-1], and j is an integer with a value range of [0, n-1].

参照图3，根据本发明示例性实施例的图像增强装置300可包括频域变换单元310、阈值处理单元320、极大极小处理单元330和频域逆变换单元340。Referring to FIG. 3 , an image enhancement device 300 according to an exemplary embodiment of the present invention may include a frequency domain transformation unit 310 , a threshold processing unit 320 , a maximin processing unit 330 and a frequency domain inverse transformation unit 340 .

频域变换单元310可对彩色图像的三个分量R、G、B分别进行频域变换以获得频域变换系数。The frequency-domain transformation unit 310 may perform frequency-domain transformation on the three components R, G, and B of the color image respectively to obtain frequency-domain transformation coefficients.

阈值处理单元320可将绝对值小于预定阈值的频域变换系数设置为零，并考虑彩色图像弱细节信号的连续性以及强细节信号不失真将绝对值大于或等于预定阈值的频域变换系数设置为与其自身相关的值。The threshold processing unit 320 may set the frequency-domain transform coefficients whose absolute value is smaller than a predetermined threshold to zero, and consider the continuity of the weak detail signal of the color image and the non-distortion of the strong detail signal to set the frequency-domain transform coefficient whose absolute value is greater than or equal to the predetermined threshold to zero. is the value associated with itself.

极大极小处理单元330可通过将设置的频域变换系数极值化以增大差异性。The maximin processing unit 330 can extremize the set frequency-domain transform coefficients to increase the difference.

频域逆变换单元340可对极值化的频域变换系数进行频域逆变换以获得三个分量Rn、Gn、Bn。The frequency domain inverse transform unit 340 may perform frequency domain inverse transform on the extremalized frequency domain transform coefficients to obtain three components Rn, Gn, Bn.

此外，所述图像增强装置300还可选地包括空间域图像增强单元350，其可通过使用三个分量Rn、Gn、Bn进行空间域图像增强；所述图像增强装置300还可选地包括RGB空间分解单元、归一化单元和彩色图像恢复单元，以分别执行彩色图像的RGB空间分解、三个分量R、G、B的归一化以及彩色图像的恢复；频域变换单元310、阈值处理单元320、极大极小处理单元330和频域逆变换单元340还可构成为一个单独的频率域图像增强单元以实现相同的功能。In addition, the image enhancement device 300 also optionally includes a spatial domain image enhancement unit 350, which can perform spatial domain image enhancement by using three components Rn, Gn, and Bn; the image enhancement device 300 also optionally includes RGB Space decomposition unit, normalization unit and color image recovery unit, to respectively perform the RGB space decomposition of color image, the normalization of three components R, G, B and the restoration of color image; frequency domain transformation unit 310, threshold value processing The unit 320, the maximin processing unit 330 and the frequency domain inverse transform unit 340 can also be configured as a single frequency domain image enhancement unit to achieve the same function.

根据本发明的示例性实施例，在彩色图像频域增强处理中，考虑彩色图像、噪声以及频域本身的影响而提出具有自适应性和普适性的阈值设置方案，从而有效地去除彩色图像的噪声；通过增加最外层变换系数的差异性，提升了图像的边缘差异并尤其是减小Contourlet变换产生的纹刷效应。在彩色图像空间域增强处理中，根据彩色图像中心点与邻域的差异性，提高了彩色图像的清晰度、对比度和鲜艳度。此外，本发明尤其适用于通过便携式终端(诸如，移动电话、平板电脑、便携式数字助理等)拍摄的质量较差的图像。According to an exemplary embodiment of the present invention, in the color image frequency domain enhancement process, an adaptive and universal threshold setting scheme is proposed considering the influence of the color image, noise and the frequency domain itself, so as to effectively remove the color image noise; by increasing the difference of the outermost transformation coefficients, the edge difference of the image is improved and especially the texture brush effect produced by the Contourlet transformation is reduced. In the enhancement process of the color image space domain, according to the difference between the center point of the color image and the neighborhood, the definition, contrast and vividness of the color image are improved. Furthermore, the present invention is especially applicable to poor quality images captured by portable terminals such as mobile phones, tablets, portable digital assistants, and the like.

虽然已经参照特定示例性实施例示出和描述了本发明，但是本领域的技术人员将理解，在不脱离范围由权利要求及其等同物限定的本发明的精神和范围的情况下可作出形式和细节上的各种改变。While the invention has been shown and described with reference to certain exemplary embodiments, it will be understood by those skilled in the art that forms and modifications may be made without departing from the spirit and scope of the invention, the scope of which is defined by the claims and their equivalents. Various changes in details.

Claims

1. An image enhancement method, comprising:

frequency domain transforming the three components R, G, B of the color image to obtain frequency domain transform coefficients, respectively;

setting the frequency domain transform coefficients having absolute values less than a predetermined threshold to zero, and setting the frequency domain transform coefficients having absolute values greater than or equal to the predetermined threshold to values related to themselves in consideration of continuity of the weak detail signal and undistorted strong detail signal of the color image;

increasing the diversity by polarizing the set frequency domain transform coefficients; and

the polarized frequency domain transform coefficients are inverse frequency domain transformed to obtain three components Rn, Gn, Bn.

2. The image enhancement method of claim 1, further comprising:

the three components R, G, B are normalized before the three components R, G, B of the color image are frequency domain transformed, respectively.

3. The image enhancement method of claim 1, wherein the frequency domain transform is a Contourlet (Contourlet) transform and the frequency domain transform coefficients are Contourlet transform coefficients.

4. The image enhancement method according to claim 3, wherein the step of setting the frequency domain transform coefficients comprises: the frequency domain transform coefficients are set by the following equation,

<math> <mrow> <mi>Coeffδ</mi> <mo>=</mo> <mfenced open='{' close=''> <mtable> <mtr> <mtd> <mi>sgn</mi> <mrow> <mo>(</mo> <mi>Coeff</mi> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <mo>|</mo> <mi>Coeff</mi> <mo>|</mo> <mo>-</mo> <mrow> <mo>(</mo> <mfrac> <mrow> <mi>max</mi> <mrow> <mo>(</mo> <mo>|</mo> <mi>Coeff</mi> <mo>|</mo> <mo>)</mo> </mrow> <mo>-</mo> <mo>|</mo> <mi>Coeff</mi> <mo>|</mo> </mrow> <mrow> <mi>max</mi> <mrow> <mo>(</mo> <mo>|</mo> <mi>Coeff</mi> <mo>|</mo> <mo>)</mo> </mrow> <mo>-</mo> <mi>δ</mi> </mrow> </mfrac> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <mfrac> <msup> <mn>2</mn> <mrow> <mfrac> <msup> <mrow> <mo>(</mo> <mi>mean</mi> <mn>2</mn> <mrow> <mo>(</mo> <mfrac> <mrow> <mi>R</mi> <mo>+</mo> <mi>G</mi> <mo>+</mo> <mi>B</mi> </mrow> <mn>3</mn> </mfrac> <mo>)</mo> </mrow> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mrow> <mn>2</mn> <mo>·</mo> <mrow> <mo>(</mo> <mi>std</mi> <mn>2</mn> <mrow> <mo>(</mo> <mfrac> <mrow> <mi>R</mi> <mo>+</mo> <mi>G</mi> <mo>+</mo> <mi>B</mi> </mrow> <mn>3</mn> </mfrac> <mo>)</mo> </mrow> <mo>)</mo> </mrow> <mo>·</mo> <mi>σ</mi> </mrow> </mfrac> <mo>·</mo> <mi>δ</mi> </mrow> </msup> <msup> <mn>2</mn> <mrow> <mfrac> <msup> <mrow> <mo>(</mo> <mi>mean</mi> <mn>2</mn> <mrow> <mo>(</mo> <mfrac> <mrow> <mi>R</mi> <mo>+</mo> <mi>G</mi> <mo>+</mo> <mi>B</mi> </mrow> <mn>3</mn> </mfrac> <mo>)</mo> </mrow> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mrow> <mn>2</mn> <mo>·</mo> <mrow> <mo>(</mo> <mi>std</mi> <mn>2</mn> <mrow> <mo>(</mo> <mfrac> <mrow> <mi>R</mi> <mo>+</mo> <mi>G</mi> <mo>+</mo> <mi>B</mi> </mrow> <mn>3</mn> </mfrac> <mo>)</mo> </mrow> <mo>)</mo> </mrow> <mo>·</mo> <mi>σ</mi> </mrow> </mfrac> <mo>·</mo> <mo>|</mo> <mi>Coeff</mi> <mo>|</mo> </mrow> </msup> </mfrac> <mo>)</mo> <mi>δ</mi> <mo>)</mo> <mo>,</mo> </mrow> </mrow> </mtd> <mtd> <mo>|</mo> <mi>Coeff</mi> <mo>|</mo> <mo>&GreaterEqual;</mo> <mi>δ</mi> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> <mo>,</mo> </mtd> <mtd> <mo>|</mo> <mi>Coeff</mi> <mo>|</mo> <mo><</mo> <mi>δ</mi> </mtd> </mtr> </mtable> </mfenced> </mrow> </math>

where δ represents a predetermined threshold, Coeff represents a Contourlet transform coefficient, Coeff δ represents Coeff after setting, sgn represents a sign-taking operation, max represents a maximum-value-taking operation, mean2 represents a mean-value-taking operation, std2 represents a standard deviation-taking operation, and σ represents an estimated noise standard deviation value of R, G, B, the value of which is obtained by the equationA calculation is performed in which Median denotes the Median operation, Coeff1 denotes the first layer coefficient of Coeff, and r has a value of 0.6745.

5. The image enhancement method of claim 4, wherein δ employs a Bayesian shrinkage (Bayesian shrinkage) threshold.

6. The image enhancement method of claim 5, wherein δ employs an adaptive Bayesian shrinkage (Bayesian shrinkage) threshold.

7. The image enhancement method of claim 6, wherein the step of quantizing the set frequency domain transform coefficients comprises: the first layer coefficient in the set Contourlet transform coefficients is quantized.

8. The image enhancement method of claim 7, wherein the step of quantizing the first layer coefficient in the set Contourlet transform coefficients comprises:

selecting a two-dimensional area for each point in each of a plurality of subgraphs constituted by coefficients of Coeff δ converted to the same direction in the two-dimensional first-layer coefficients by a predetermined matrix;

performing convolution processing on each selected two-dimensional area through a mean matrix to obtain a convolution value;

for each selected two-dimensional region, the value of the center point takes the maximum value in the selected two-dimensional region when the value of the center point of the selected two-dimensional region is greater than or equal to the convolution value, and the value of the center point takes the minimum value in the selected two-dimensional region when the value of the center point of the selected two-dimensional region is less than the convolution value.

9. The image enhancement method of claim 1, further comprising:

spatial domain image enhancement is performed by using the three components Rn, Gn, Bn.

10. The image enhancement method of claim 9, wherein the step of performing spatial domain image enhancement by using the three components Rn, Gn, Bn comprises:

convolution processing is respectively carried out on the three components Rn, Gn and Bn through a matrix in a pyramid form to obtain RnF, GnF and BnF, the three components RNS, GNS and BNS of the spatial domain image enhancement are obtained through the following equations,

<math> <mfenced open='{' close=''> <mtable> <mtr> <mtd> <mi>RNS</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>Rn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>·</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>+</mo> <mi>t</mi> <mo>·</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <mfrac> <mrow> <mi>Rn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>+</mo> <mi>Gn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>+</mo> <mi>Bn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mn>3</mn> <mo>·</mo> <mi>Rn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>)</mo> </mrow> <mo>)</mo> </mrow> <mo>·</mo> <msup> <mrow> <mo>(</mo> <msqrt> <mfrac> <mrow> <mi>Rn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mi>RnF</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> </mfrac> </msqrt> <mo>)</mo> </mrow> <mrow> <mn>1</mn> <mo>&PlusMinus;</mo> <mi>ζ</mi> </mrow> </msup> </mtd> </mtr> <mtr> <mtd> <mi>GNS</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>Gn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>·</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>+</mo> <mi>t</mi> <mo>·</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <mfrac> <mrow> <mi>Rn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>+</mo> <mi>Gn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>+</mo> <mi>Bn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mn>3</mn> <mo>·</mo> <mi>Gn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>)</mo> </mrow> <mo>)</mo> </mrow> <mo>·</mo> <msup> <mrow> <mo>(</mo> <msqrt> <mfrac> <mrow> <mi>Gn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mi>GnF</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> </mfrac> </msqrt> <mo>)</mo> </mrow> <mrow> <mn>1</mn> <mo>&PlusMinus;</mo> <mi>ζ</mi> </mrow> </msup> </mtd> </mtr> <mtr> <mtd> <mi>BNS</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>Bn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>·</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>+</mo> <mi>t</mi> <mo>·</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <mfrac> <mrow> <mi>Rn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>+</mo> <mi>Gn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>+</mo> <mi>Bn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mn>3</mn> <mo>·</mo> <mi>Bn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>)</mo> </mrow> <mo>)</mo> </mrow> <mo>·</mo> <msup> <mrow> <mo>(</mo> <msqrt> <mfrac> <mrow> <mi>Bn</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mi>BnF</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> </mfrac> </msqrt> <mo>)</mo> </mrow> <mrow> <mn>1</mn> <mo>&PlusMinus;</mo> <mi>ζ</mi> </mrow> </msup> </mtd> </mtr> </mtable> </mfenced> </math>

where Rn (i, j), Gn (i, j), and Bn (i, j) respectively represent values at coordinates (i, j) in Rn, Gn, Bn, RnF (i, j), GnF (i, j), BnF (i, j) respectively represent values at coordinates (i, j) in RnF, GnF, BnF, RNS, GNS (i, j), BNS (i, j) respectively represent values at coordinates (i, j) in RNS, GNS, BNS, t represents an overall saturation adjustment factor, ζ represents a local contrast adjustment factor, and RNS (i, j), GNS (i, j), BNS (i, j) are limited to corresponding end point values of a predetermined range when RNS (i, j), GNS (i, j), BNS (i, j) exceed the predetermined range.

11. The image enhancement method according to one of claims 1 to 10, wherein the color image is obtained by a portable terminal.

12. An image enhancement apparatus comprising:

a frequency domain transform unit that frequency domain transforms the three components R, G, B of the color image, respectively, to obtain frequency domain transform coefficients;

a threshold processing unit that sets the frequency domain transform coefficient whose absolute value is smaller than a predetermined threshold to zero, and sets the frequency domain transform coefficient whose absolute value is greater than or equal to the predetermined threshold to a value related to itself in consideration of continuity of the color image weak detail signal and undistorted strong detail signal;

a maximum and minimum processing unit, which increases the difference by polarizing the set frequency domain transformation coefficient; and

and a frequency domain inverse transformation unit which performs frequency domain inverse transformation on the polarized frequency domain transformation coefficients to obtain three components Rn, Gn and Bn.

13. The image enhancement apparatus of claim 12, further comprising:

and a spatial domain image enhancement unit for performing spatial domain image enhancement by using the three components Rn, Gn, Bn.

14. The image intensifier as set forth in one of claims 12 and 13, wherein the color image is obtained by a portable terminal.