CN103327337B

CN103327337B - A kind of classification quantitative coding method based on biorthogonal lapped transform

Info

Publication number: CN103327337B
Application number: CN201310270943.1A
Authority: CN
Inventors: 田昕; 李松; 郑国兴; 周辉; 杨晋陵; 高俊玲
Original assignee: Wuhan University WHU
Current assignee: Wuhan University WHU
Priority date: 2013-06-28
Filing date: 2013-06-28
Publication date: 2015-12-23
Anticipated expiration: 2033-06-28
Also published as: CN103327337A

Abstract

The invention provides a classification and quantization encoding method based on double-orthogonal overlapping transform, which belongs to the technical field of remote sensing image data transmission. The invention firstly completes the coding parameter design step through the test image sequence; then realizes the image coding algorithm step for the input image. The present invention divides the images into different types according to the coding characteristics through the method of classification training, and selects a suitable quantization method for each type, thereby effectively solving the problem caused by different requirements for coding performance parameters of different types of remote sensing images under the premise of constant quality coding The problem of image coding performance difference is solved, so as to achieve the goal of coding objective quality of different types of images. The invention is used in the application of remote sensing satellites, and greatly improves the downloading efficiency of remote sensing satellite image data.

Description

A classification and quantization coding method based on biorthogonal overlapping transform

技术领域technical field

本发明属于遥感图像数据传输技术领域，具体涉及一种基于双正交重叠变换的分类量化编码方法。The invention belongs to the technical field of remote sensing image data transmission, and in particular relates to a classification, quantization and encoding method based on biorthogonal overlapping transform.

背景技术Background technique

由于观测的需要，卫星探测器需要将观测系统所拍摄的图像传回至地面。随着用户的需求，图像的分辨率将会越来越高，这将会导致图像数据量的急剧增长,而目前空间通信的数据链路信道容量是有限的。为了使得地面能够接收高质量的图像，通过遥感图像编码方法解决遥感图像大数据量传输与有限信道之间的矛盾，是非常有必要的。Due to the needs of observation, satellite detectors need to transmit the images taken by the observation system back to the ground. With the needs of users, the resolution of images will be higher and higher, which will lead to a sharp increase in the amount of image data, and the current space communication data link channel capacity is limited. In order to enable the ground to receive high-quality images, it is very necessary to solve the contradiction between the large amount of remote sensing image data transmission and the limited channel through the remote sensing image coding method.

基于离散余弦变换和小波变换的嵌入式编码方法是目前遥感图像编码方法中的热点。但是，在传统方法中，编码性能的提升往往会伴随着复杂度的增加，这就给编码系统的硬件设计带来更多的要求。而在卫星系统中，硬件设备在运算能力、内存及功耗上均有很多的限制。因此，传统方法很难满足硬件系统实时采集传输的应用需求。基于双正交重叠变换的图像编码技术为实现低复杂度、高性能图像编码提供了一种有效的方法，其已被JPEGXR标准所采用。Embedded coding method based on discrete cosine transform and wavelet transform is a hotspot in remote sensing image coding methods at present. However, in traditional methods, the improvement of encoding performance is often accompanied by the increase of complexity, which brings more requirements to the hardware design of the encoding system. In satellite systems, hardware devices have many limitations in terms of computing power, memory, and power consumption. Therefore, traditional methods are difficult to meet the application requirements of real-time acquisition and transmission of hardware systems. The image coding technology based on biorthogonal overlapping transform provides an effective method for realizing low-complexity, high-performance image coding, which has been adopted by the JPEGXR standard.

在遥感对地观测中，地物之间的差异很大，在传统的编码方法中，往往是将所有地物图像按照相同的方式进行编码传输。因此，纹理丰富的图像往往会比纹理简单的图像具有更多的失真。而对于用户而言，往往会希望各种类型的图像均可满足质量要求，即具有相似的编码失真，这就对图像编码方法提出了新的需求。In remote sensing earth observation, there are great differences among ground features. In the traditional coding method, all ground feature images are often encoded and transmitted in the same way. Therefore, images with rich textures tend to have more distortion than images with simple textures. For users, they often hope that various types of images can meet the quality requirements, that is, have similar coding distortion, which puts forward new requirements for image coding methods.

Li等人通过深入分析图像活跃度与图像编码性能之间的联系，建立了一种用于描述JPEG2000编码质量的预测模型。（参见文献：LingLiandZhen-SongWang,CompressionQualityPredictionModelforJPEG2000，IEEETransactionsonImageProcessing，2010）进一步可以发现，不同类型的图像具有不同的编码特征，而这种特征可以通过图像活跃度来衡量。Li et al. established a predictive model for describing JPEG2000 encoding quality by in-depth analysis of the link between image liveness and image encoding performance. (See literature: LingLiandZhen-SongWang, CompressionQualityPredictionModelforJPEG2000, IEEETransaction on ImageProcessing, 2010) It can further be found that different types of images have different encoding characteristics, and this characteristic can be measured by image activity.

发明内容Contents of the invention

针对背景技术存在的问题，本发明提出一种基于双正交重叠变换的分类量化编码方法，可有效解决不同类型遥感图像在定质量编码前提下对编码性能参数要求不同造成的图像编码性能差异问题，从而达到了不同类型图像定质量编码的目的。Aiming at the problems existing in the background technology, the present invention proposes a classification and quantization coding method based on bio-orthogonal overlapping transform, which can effectively solve the problem of image coding performance differences caused by different requirements for coding performance parameters of different types of remote sensing images under the premise of constant quality coding , so as to achieve the purpose of quality coding for different types of images.

为解决上述技术问题，本发明采用如下技术方案：In order to solve the problems of the technologies described above, the present invention adopts the following technical solutions:

一种基于双正交重叠变换的分类量化编码方法，包括下述步骤，A classification and quantization encoding method based on biorthogonal overlapping transform, comprising the following steps,

步骤1、通过测试图像序列完成编码参数的设计；Step 1, complete the design of encoding parameters by testing image sequences;

步骤2、利用步骤1得到的编码参数对输入图像进行编码。Step 2. Using the encoding parameters obtained in Step 1 to encode the input image.

所述步骤1具体包括以下步骤，The step 1 specifically includes the following steps,

步骤1.1、对测试图像进行双正交重叠变换；Step 1.1, carry out bi-orthogonal overlapping transformation to test image;

变换方法采用JPEGXR标准的变换方法，具体为：首先将测试图像分割成大小为16×16的宏块，以每个宏块中4×4的块为单元进行第一次双正交重叠变换；变换后，每个4×4块中左上角数据为DC系数，剩余的15个为HP系数；将16×16宏块中所有的DC系数组成4×4的块进行第二次双正交重叠变换，最终得到一个DC系数和15个LP系数；每个16×16的宏块最终变换后的数据包括1个DC系数、15个LP系数和240个HP系数；The transformation method adopts the transformation method of the JPEGXR standard, which is as follows: firstly, the test image is divided into macroblocks with a size of 16×16, and the first bi-orthogonal overlapping transformation is performed in units of 4×4 in each macroblock; After transformation, the data in the upper left corner of each 4×4 block is the DC coefficient, and the remaining 15 are HP coefficients; all the DC coefficients in the 16×16 macroblock are formed into a 4×4 block for the second bi-orthogonal overlapping Transform to finally obtain a DC coefficient and 15 LP coefficients; the final transformed data of each 16×16 macroblock includes 1 DC coefficient, 15 LP coefficients and 240 HP coefficients;

步骤1.2、对每个宏块变换后的DC系数、LP系数和HP系数进行量化，量化公式为：Step 1.2, quantize the transformed DC coefficient, LP coefficient and HP coefficient of each macroblock, the quantization formula is:

${B B}_{ij ij} = = round round ((\frac{{A A}_{ij ij}}{{Q Q}_{step step}}))$

其中，A_ij代表变换后的系数，B_ij代表量化后的系数，Q_step代表量化步长，初始量化步长设置为1，round()表示四舍五入运算；Among them, A _ij represents the transformed coefficient, B _ij represents the quantized coefficient, Q _step represents the quantization step size, the initial quantization step size is set to 1, and round() represents the rounding operation;

步骤1.3、对B_ij进行反量化，反量化公式为：Step 1.3, perform inverse quantization on B _ij , the inverse quantization formula is:

Ci_j＝Bi_j×Q_step Ci _j =Bi _j ×Q _step

其中，C_ij代表宏块的反量化系数，i、j分别代表C_ij的行、列坐标；Wherein, C _ij represents the inverse quantization coefficient of the macroblock, and i and j represent the row and column coordinates of C _ij respectively;

步骤1.4、根据步骤1.1对C_ij进行双正交重叠变换逆变换，获得重建图像；Step 1.4, according to step 1.1, perform biorthogonal overlapping transformation and inverse transformation on C _ij to obtain a reconstructed image;

步骤1.5、计算测试图像和重建图像的峰值信噪比（PeakSignaltoNoiseRatio，PSNR），计算公式为：Step 1.5, calculate the peak signal-to-noise ratio (PeakSignaltoNoiseRatio, PSNR) of the test image and the reconstructed image, the calculation formula is:

$PSNR PSNR = = 201201 g g \sqrt{\frac{255255^{22}}{MSE MSE}},,$ $MSE MSE = = \frac{11}{M m \times \times N N} {Σ Σ}_{i i = = 00}^{M m - - 11} {Σ Σ}_{j j = = 00}^{N N - - 11} {[[I I ((i i,, j j)) - - \overset{~ ~}{I I} ((i i,, j j))]]}^{22}$

其中，M、N分别代表图像的长度和宽度，I(i,j)和分别代表原始图像和重建图像的像素大小，i、j分别代表原始图像和重建图像的行、列坐标；Among them, M and N represent the length and width of the image respectively, and I(i,j) and Represent the pixel size of the original image and the reconstructed image respectively, i and j represent the row and column coordinates of the original image and the reconstructed image respectively;

步骤1.6、设预期图像客观质量为T_Q，若PSNR-T_Q＞0.5，则Q_step=Q_step+1，然后重复步骤1.2至步骤1.6；若PSNR-T_Q＜-0.5，则Q_step=Q_step-1，然后重复步骤1.2至步骤1.6；其他情况则结束迭代计算过程；Step 1.6, set the expected image objective quality as T _Q , if PSNR-T _Q >0.5, then Q _step =Q _step +1, then repeat steps 1.2 to 1.6; if PSNR-T _Q <-0.5, then Q _step = Q _step -1, then repeat step 1.2 to step 1.6; otherwise, end the iterative calculation process;

步骤1.7、通过迭代计算过程确定量化步长，根据量化步长将图像划分为不同类型，并设定每个类型的量化步长；Step 1.7, determine the quantization step size through an iterative calculation process, divide the image into different types according to the quantization step size, and set the quantization step size of each type;

步骤1.8、计算图像活跃度，包括IAMD1、IAMD2、IAME1和IAME2，并将之作为图像的分类特征，构建特征向量并结合步骤1.7中的已分类类型完成支持向量机的训练；Step 1.8, calculate the image activity, including IAMD1, IAMD2, IAME1 and IAME2, and use it as the classification feature of the image, construct the feature vector and complete the training of the support vector machine in combination with the classified type in step 1.7;

具体计算公式如下：The specific calculation formula is as follows:

$IAMD IAMD 11 = = \frac{11}{((M m - - 11)) \times \times N N} {Σ Σ}_{i i = = 00}^{M m - - 22} {Σ Σ}_{j j = = 00}^{N N - - 11} | | x x ((i i,, j j)) - - x x ((i i + + 11,, j j)) | |$

$+ + \frac{11}{M m \times \times ((N N - - 11))} {Σ Σ}_{i i = = 00}^{M m - - 11} {Σ Σ}_{j j = = 00}^{N N - - 22} | | x x ((i i,, j j)) - - x x ((i i,, j j + + 11)) | |$

$IAMD IAMD 22 = = \frac{11}{((M m - - 11)) \times \times ((N N - - 11))} {Σ Σ}_{i i = = 11}^{M m - - 11} {Σ Σ}_{j j = = 00}^{N N - - 22} | | x x ((i i,, j j)) - - x x ((i i - - 11,, j j + + 11)) | |$

$+ + \frac{11}{((M m - - 11)) \times \times ((N N - - 11))} {Σ Σ}_{i i = = 00}^{M m - - 22} {Σ Σ}_{j j = = 00}^{N N - - 22} | | x x ((i i,, j j)) - - x x ((i i + + 11,, j j + + 11)) | |$

$IAME IAME 11 = = \frac{11}{((M m - - 22)) \times \times N N} {Σ Σ}_{i i = = 11}^{M m - - 22} {Σ Σ}_{j j = = 00}^{N N - - 11} | | x x ((i i - - 11,, j j)) - - x x ((i i + + 11,, j j)) | |$

$+ + \frac{11}{M m \times \times ((N N - - 22))} {Σ Σ}_{i i = = 00}^{M m - - 11} {Σ Σ}_{j j = = 11}^{N N - - 22} | | x x ((i i,, j j - - 11)) - - x x ((i i,, j j + + 11)) | |$

$IAME IAME 22 = = \frac{11}{((M m - - 22)) \times \times ((N N - - 22))} {Σ Σ}_{i i = = 11}^{M m - - 22} {Σ Σ}_{j j = = 11}^{N N - - 22} | | x x ((i i - - 11,, j j - - 11)) - - x x ((i i + + 11,, j j + + 11)) | |$

$+ + \frac{11}{((M m - - 22)) \times \times ((N N - - 22))} {Σ Σ}_{i i = = 11}^{M m - - 22} {Σ Σ}_{j j = = 11}^{N N - - 22} | | x x ((i i + + 11,, j j - - 11)) - - x x ((i i - - 11,, j j + + 11)) | |$

x(i,j)代表原始图像的像素大小，i、j分别代表原始图像的行、列坐标。x(i, j) represents the pixel size of the original image, and i and j represent the row and column coordinates of the original image, respectively.

所述步骤2具体包括以下步骤，The step 2 specifically includes the following steps,

步骤2.1、利用与步骤1.1相同的方法对输入图像进行双正交重叠变换；Step 2.1, using the same method as step 1.1 to carry out biorthogonal overlapping transformation to the input image;

步骤2.2、利用与步骤1.8相同的方法计算图像的活跃度；Step 2.2, using the same method as step 1.8 to calculate the liveness of the image;

步骤2.3、结合图像活跃度构建用于分类的特征向量，结合步骤1.8的分类训练结果，通过支持向量机确定输入图像的类型，进一步确定量化步长；Step 2.3, constructing a feature vector for classification in combination with the image activity, combined with the classification training results in step 1.8, determining the type of the input image through a support vector machine, and further determining the quantization step size;

步骤2.4、根据量化步长对输入图像双正交重叠变换后的系数进行量化；Step 2.4, quantize the coefficients after the biorthogonal overlapping transform of the input image according to the quantization step;

步骤2.5、对量化后的系数进行熵编码。Step 2.5, performing entropy coding on the quantized coefficients.

与现有技术相比，本发明可有效解决不同类型遥感图像在定质量编码前提下对编码性能参数要求不同造成的图像编码性能差异问题，从而达到了不同类型图像定客观质量编码的目的。用于遥感卫星应用中，可以大大提高遥感卫星图像数据的下传效率。Compared with the prior art, the present invention can effectively solve the problem of image coding performance differences caused by different requirements for coding performance parameters of different types of remote sensing images under the premise of constant quality coding, thereby achieving the purpose of fixed objective quality coding of different types of images. Used in remote sensing satellite applications, it can greatly improve the download efficiency of remote sensing satellite image data.

附图说明Description of drawings

图1是本发明的流程图；Fig. 1 is a flow chart of the present invention;

图2(a)是分类后第A类中的典型测试图像；Figure 2(a) is a typical test image in class A after classification;

图2(b)是分类后第B类中的典型测试图像；Figure 2(b) is a typical test image in class B after classification;

图2(c)是分类后第C类中的典型测试图像；Figure 2(c) is a typical test image in category C after classification;

图3(a)是输入图像Road；Figure 3(a) is the input image Road;

图3(b)是输入图像Farmland；Figure 3(b) is the input image Farmland;

图3(c)是输入图像Meadow；Figure 3(c) is the input image Meadow;

图3(d)是输入图像City；Figure 3(d) is the input image City;

图4(a)是解压图像Road；Figure 4(a) is the decompressed image Road;

图4(b)是解压图像Farmland；Figure 4(b) is the decompressed image Farmland;

图4(c)是解压图像Meadow；Figure 4(c) is the decompressed image Meadow;

图4(d)是解压图像City；Figure 4(d) is the decompressed image City;

具体实施方式Detailed ways

实施例1：Example 1:

（1）建立测试图像库，图像数目为15幅；(1) Establish a test image library with 15 images;

（2）设定预期的客观图像质量PSNR为35dB；(2) Set the expected objective image quality PSNR to 35dB;

（3）计算测试图像的图像活跃度值，以及通过双正交重叠变换后经过量化反变换后，在35dB附近时的量化参数值，其结果分别如表1所示；(3) Calculate the image activity value of the test image, and the quantization parameter value near 35dB after biorthogonal overlapping transformation and quantization inverse transformation. The results are shown in Table 1;

表1Table 1

（4）按照量化参数的大小将测试图象1、2、3、4、5、11、12、13、14、15定义为A类，测试图象6、15定义为B类，测试图象7、8、9、10分为C类；(4) Define test images 1, 2, 3, 4, 5, 11, 12, 13, 14, and 15 as Class A according to the size of the quantization parameters, and test images 6 and 15 as Class B. Test images 7, 8, 9, and 10 are divided into category C;

（5）将不同测试图像的图像活跃度值作为特征值用于支持向量机的训练；(5) The image activity values of different test images are used as feature values for the training of support vector machines;

（6）将A类测试图象的量化步长取均值为47，B类测试图象的量化步长取均值为59，C类测试图象的量化步长取均值为66；(6) The average quantization step length of the A-type test image is 47, the average quantization step length of the B-type test image is 59, and the average quantization step length of the C-type test image is 66;

（7）输入图像包括不同类型图像Road、Farmland、Meadow、City四幅，其活跃度计算值分别如表2所示；(7) The input images include four different types of images Road, Farmland, Meadow, and City, and their activity calculation values are shown in Table 2;

表2Table 2

图像image IAMD1IAMD1 IAMD2IAMD2 IAME1IAME1 IAME2IAME2 RoadRoad 21.711121.7111 27.708527.7085 33.738233.7382 38.308838.3088 FarmlandFarmland 7.94637.9463 13.099213.0992 12.433412.4334 19.594819.5948 MeadowMeadow 8.82768.8276 12.117712.1177 13.454413.4544 16.905616.9056 CityCity 36.768336.7683 48.154448.1544 52.336652.3366 62.413162.4131

（8）根据图像活跃度值并结合步骤5中支持向量机的训练结果,通过支持向量机对输入图像进行分类。分类结果为：Road和City分为A类；Meadow分为B类；Farmland分为C类；(8) According to the image activity value and combined with the training results of the support vector machine in step 5, the input image is classified by the support vector machine. The classification results are as follows: Road and City are classified into Class A; Meadow is classified into Class B; Farmland is classified into Class C;

（9）对输入图像逐次进行双正交重叠变换及量化，量化步长分别选择47、66、59和47。(9) Perform biorthogonal overlapping transformation and quantization on the input image successively, and the quantization steps are selected as 47, 66, 59 and 47 respectively.

（10）对量化后的数据完成熵编码。熵编码方法采用了自适应算术编码方法。(10) Complete entropy coding on the quantized data. The entropy coding method adopts an adaptive arithmetic coding method.

（11）为了对编码方法的有效性进行评估，对熵编码后的数据进行解码，获得解码后的重建图像。通过PSNR来衡量输入图像与重建图像的差异，其结果分别为35.45、35.06、35.02和34.53，可以看出最大误差不超过1dB，达到了预期定客观质量编码的目的。(11) In order to evaluate the effectiveness of the encoding method, the entropy encoded data is decoded to obtain the decoded reconstructed image. The difference between the input image and the reconstructed image is measured by PSNR, and the results are 35.45, 35.06, 35.02, and 34.53, respectively. It can be seen that the maximum error does not exceed 1dB, and the objective of objective quality coding is achieved.

Claims

1., based on a classification quantitative coding method for biorthogonal lapped transform, it is characterized in that: comprise the steps,

Step 1, completed the design of coding parameter by test pattern sequence;

Step 2, the coding parameter utilizing step 1 to obtain are encoded to input picture;

Described step 1 specifically comprises the following steps,

Step 1.1, biorthogonal lapped transform is carried out to test pattern;

Transform method adopts the transform method of JPEGXR standard, is specially: first test pattern is divided into size be 16 × 16 macro block, with in each macro block 4 × 4 block be that unit carries out first time biorthogonal lapped transform; After conversion, in each 4 × 4 pieces, upper left corner data are DC coefficient, and remaining 15 is HP coefficient; The block of DC coefficients all in 16 × 16 macro blocks composition 4 × 4 is carried out second time biorthogonal lapped transform, finally obtains a DC coefficient and 15 LP coefficients; Data after the macro block of each 16 × 16 finally converts comprise 1 DC coefficient, 15 LP coefficients and 240 HP coefficients;

Step 1.2, DC coefficient, LP coefficient and HP coefficient to after each macroblock transform quantize, and quantitative formula is:

B_{i j} = r o u n d (\frac{A_{i j}}{Q_{s t e p}})

Wherein, A _ijcoefficient after representation transformation, B _ijcoefficient after representative quantizes, Q _steprepresent quantization step, initial quantization step is set to 1, round () and represents the computing that rounds up;

Step 1.3, to B _ijcarry out inverse quantization, inverse quantization formula is:

C _ij＝B _ij×Q _step

Wherein, C _ijrepresent the dequantized coefficients of macro block, i, j represent C respectively _ijrow, column coordinate;

Step 1.4, according to step 1.1 couple C _ijcarry out biorthogonal lapped transform inverse transformation, obtain and rebuild image;

The Y-PSNR (PeakSignaltoNoiseRatio, PSNR) of step 1.5, calculating test pattern and reconstruction image, computing formula is:

P S N R = 20 \lg \sqrt{\frac{255^{2}}{M S E}}, M S E = \frac{1}{M \times N} Σ_{i = 0}^{M - 1} {Σ_{j = 0}^{N - 1} [I (i, j) - \tilde{I} (i, j)]}^{2}

Wherein, M, N be the length of representative image and width respectively, I (i, j) and represent the pixel size of original image and reconstruction image respectively, i, j represent original image respectively and rebuild the row, column coordinate of image;

Step 1.6, set expection image objective quality as T _qif, PSNR-T _q> 0.5, then Q _step=Q _step+ 1, then repeat step 1.2 to step 1.6; If PSNR-T _q<-0.5, then Q _step=Q _step-1, then repeat step 1.2 to step 1.6; Other situations are finishing iteration computational process then;

Step 1.7, by iterative process determination quantization step, image is divided into dissimilar according to quantization step, and sets the quantization step of each type;

Step 1.8, computed image liveness, comprise IAMD1, IAMD2, IAME1 and IAME2, and using it characteristic of division as image, and construction feature vector classification type in integrating step 1.7 complete the training of SVMs;

Specific formula for calculation is as follows:

\begin{matrix} I A M D 1 = \frac{1}{(M - 1) \times N} Σ_{i = 0}^{M - 2} Σ_{j = 0}^{N - 1} | x (i, j) - x (i + 1, j) | \\ + \frac{1}{M \times (N - 1)} Σ_{i = 0}^{M - 2} Σ_{j = 0}^{N - 2} | x (i, j) - x (i, j + 1) | \end{matrix}

\begin{matrix} I A M D 2 = \frac{1}{(M - 1) \times (N - 1)} Σ_{i = 1}^{M - 1} Σ_{j = 0}^{N - 2} | x (i, j) - x (i - 1, j + 1) | \\ + \frac{1}{(M - 1) \times (N - 1)} Σ_{i = 0}^{M - 2} Σ_{j = 0}^{N - 2} | x (i, j) - x (i + 1, j + 1) | \end{matrix}

\begin{matrix} I A M E 1 = \frac{1}{(M - 2) \times N} Σ_{i = 1}^{M - 2} Σ_{j = 0}^{N - 1} | x (i - 1, j) - x (i + 1, j) | \\ + \frac{1}{M \times (N - 2)} Σ_{i = 0}^{M - 1} Σ_{j = 1}^{N - 2} | x (i, j - 1) - x (i, j + 1) | \end{matrix}

\begin{matrix} I A M E 2 = \frac{1}{(M - 2) \times (N - 2)} Σ_{i = 1}^{M - 2} Σ_{j = 1}^{N - 2} | x (i - 1, j - 1) - x (i + 1, j + 1) | \\ + \frac{1}{(M - 2) \times (N - 2)} Σ_{i = 1}^{M - 2} Σ_{j = 1}^{N - 2} | x (i + 1, j - 1) - x (i - 1, j + 1) | \end{matrix}

Wherein, x (i, j) represents the pixel size of original image, and i, j represent the row, column coordinate of original image respectively.

2. a kind of classification quantitative coding method based on biorthogonal lapped transform as claimed in claim 1, is characterized in that: described step 2 specifically comprises the following steps,

Step 2.1, the utilization method identical with step 1.1 carry out biorthogonal lapped transform to input picture;

Step 2.2, utilize the liveness of the method computed image identical with step 1.8;

Step 2.3, combining image liveness build the characteristic vector for classifying, and the classification based training result of integrating step 1.8, by the type of SVMs determination input picture, determines quantization step further;

Step 2.4, according to quantization step, the coefficient after input picture biorthogonal lapped transform to be quantized;

Step 2.5, to quantize after coefficient carry out entropy code.