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

Abstract

The invention provides a kind of classification quantitative coding method based on biorthogonal lapped transform, belong to remote sensing image data transmission technique field.The present invention first completes coding parameter design procedure by test pattern sequence; Then Image Coding Algorithms step is realized to input picture.Image is divided into different types according to encoding characteristics by the method for classification based training by the present invention, suitable quantization method is selected to every type, thus efficiently solve dissimilar remote sensing images and determining to the Image Coding performance difference problem that causes of coding performance parameter requirement difference under quality coded prerequisite, thus reach dissimilar image and determine the object that objective quality encodes.The present invention is used for, in remote sensing satellite application, substantially increasing the lower transfer efficiency of remote sensing satellite image data.

Description

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

Technical field

The invention belongs to remote sensing image data transmission technique field, be specifically related to a kind of classification quantitative coding method based on biorthogonal lapped transform.

Background technology

Due to the needs of observation, satellite sensor needs the image captured by observation system to be transmitted back to ground.Along with the demand of user, the resolution of image will be more and more higher, and this will cause the sharp increase of image data amount, and the data link channel capacity of space communication is at present limited.In order to make ground can receive high-quality image, solving the contradiction between the transmission of remote sensing images big data quantity and limited channel by Remote Sensing Image Coding method, being necessary.

Embedded encoded method based on discrete cosine transform and wavelet transformation is the focus in current Remote Sensing Image Coding method.But in conventional methods where, the lifting of coding efficiency is often along with the increase of complexity, and this brings more requirement with regard to giving the hardware designs of coded system.And in satellite system, hardware device all has a lot of restrictions in operational capability, internal memory and power consumption.Therefore, conventional method is difficult to the application demand meeting the transmission of hardware system Real-time Collection.Based on biorthogonal lapped transform image coding technique for realize low complex degree, high-performance image coding provide a kind of effective method, its adopt by JPEGXR standard.

In remote sensing earth observation, widely different between atural object, in traditional coding method, carries out coding transmission by all cartographic features according to identical mode often.Therefore, the image of texture-rich often has more distortion than the simple image of texture.And for user, often wish that various types of image all can meet quality requirement, namely have similar coding distortion, this just proposes new demand to method for encoding images.

The people such as Li, by analysing in depth contacting between image liveness and Image Coding performance, establish a kind of forecast model for describing JPEG2000 coding quality.(see document: LingLiandZhen-SongWang, CompressionQualityPredictionModelforJPEG2000, IEEETransactionsonImageProcessing, 2010) can find further, dissimilar image has different coding characteristics, and this feature can be weighed by image liveness.

Summary of the invention

For background technology Problems existing, the present invention proposes a kind of classification quantitative coding method based on biorthogonal lapped transform, can determine to require the different Image Coding performance difference problem caused to coding performance parameter under quality coded prerequisite by efficient solution remote sensing images never of the same type, thus reaching the object that dissimilar image determines quality coded.

For solving the problems of the technologies described above, the present invention adopts following technical scheme:

Based on a classification quantitative coding method for biorthogonal lapped transform, 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:

$B_{\mathrm{ij}}=\mathrm{round}\left(\frac{A_{\mathrm{ij}}}{Q_{\mathrm{step}}}\right)$

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:

Ci _j＝Bi _j×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:

$\mathrm{PSNR}=201g\sqrt{\frac{{255}^2}{\mathrm{MSE}}},$ $\mathrm{MSE}=\frac{1}{M\×N}\underset{i=0}{\overset{M-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{N-1}{\mathrm{\Σ}}}{[I(i,j)-\tilde{I}(i,j)]}^2$

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:

$\mathrm{IAMD}1=\frac{1}{(M-1)\&times;N}\underset{i=0}{\overset{M-2}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N-1}{\mathrm{\&Sigma;}}}|x(i,j)-x(i+1,j)|$

$+\frac{1}{M\&times;(N-1)}\underset{i=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N-2}{\mathrm{\&Sigma;}}}|x(i,j)-x(i,j+1)|$

$\mathrm{IAMD}2=\frac{1}{(M-1)\&times;(N-1)}\underset{i=1}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N-2}{\mathrm{\&Sigma;}}}|x(i,j)-x(i-1,j+1)|$

$+\frac{1}{(M-1)\&times;(N-1)}\underset{i=0}{\overset{M-2}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N-2}{\mathrm{\&Sigma;}}}|x(i,j)-x(i+1,j+1)|$

$\mathrm{IAME}1=\frac{1}{(M-2)\&times;N}\underset{i=1}{\overset{M-2}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N-1}{\mathrm{\&Sigma;}}}|x(i-1,j)-x(i+1,j)|$

$+\frac{1}{M\&times;(N-2)}\underset{i=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N-2}{\mathrm{\&Sigma;}}}|x(i,j-1)-x(i,j+1)|$

$\mathrm{IAME}2=\frac{1}{(M-2)\&times;(N-2)}\underset{i=1}{\overset{M-2}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N-2}{\mathrm{\&Sigma;}}}|x(i-1,j-1)-x(i+1,j+1)|$

$+\frac{1}{(M-2)\&times;(N-2)}\underset{i=1}{\overset{M-2}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N-2}{\mathrm{\&Sigma;}}}|x(i+1,j-1)-x(i-1,j+1)|$

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

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.

Compared with prior art, the present invention can determine to require the different Image Coding performance difference problem caused to coding performance parameter under quality coded prerequisite by efficient solution remote sensing images never of the same type, thus reaches the object that dissimilar image determines objective quality coding.For in remote sensing satellite application, the lower transfer efficiency of remote sensing satellite image data greatly can be improved.

Accompanying drawing explanation

Fig. 1 is flow chart of the present invention;

Fig. 2 (a) is the typical test pattern after classification in category-A;

Fig. 2 (b) is the typical test pattern after classification in category-B;

Fig. 2 (c) is the typical test pattern after classification in C class;

Fig. 3 (a) is input picture Road;

Fig. 3 (b) is input picture Farmland;

Fig. 3 (c) is input picture Meadow;

Fig. 3 (d) is input picture City;

Fig. 4 (a) is decompressed image Road;

Fig. 4 (b) is decompressed image Farmland;

Fig. 4 (c) is decompressed image Meadow;

Fig. 4 (d) is decompressed image City;

Embodiment

Embodiment 1:

(1) set up test pattern storehouse, picture number is 15 width;

(2) the Objective image quality PSNR of setting expection is 35dB;

(3) calculate the image of test pattern and enliven angle value, and by after biorthogonal lapped transform after quantizing inverse transformation, quantization parameter value time near 35dB, its result is as shown in table 1 respectively;

Table 1

(4) resemble 1,2,3,4,5,11,12,13,14,15 according to large young pathbreaker's resolution chart of quantization parameter and be defined as category-A, test image 6,15 is defined as category-B, and test image 7,8,9,10 is divided into C class;

(5) image of different test pattern is enlivened angle value is used for SVMs training as characteristic value;

(6) by the quantization step of category-A test image, to get average be 47, and it is that to get average be 66 for the quantization step of 59, C class testing image that the quantization step of category-B test image gets average;

(7) input picture comprises dissimilar image Road, Farmland, Meadow, City tetra-width, and its liveness calculated value is as shown in table 2 respectively;

Table 2

Image	IAMD1	IAMD2	IAME1	IAME2
					Road	21.7111	27.7085	33.7382	38.3088
Farmland	7.9463	13.0992	12.4334	19.5948
					Meadow	8.8276	12.1177	13.4544	16.9056
City	36.7683	48.1544	52.3366	62.4131

(8) enliven angle value according to image and the training result of SVMs in integrating step 5, by SVMs, input picture is classified.Classification results is: Road and City is divided into category-A; Meadow is divided into category-B; Farmland is divided into C class;

(9) successively carry out biorthogonal lapped transform and quantification to input picture, quantization step selects 47,66,59 and 47 respectively.

(10) entropy code is completed to the data after quantification.Entropy coding method have employed adaptive arithmetic code method.

(11) in order to assess the validity of coding method, to the decoding data after entropy code, decoded reconstruction image is obtained.Weigh input picture and the difference of rebuilding image by PSNR, its result is respectively 35.45,35.06,35.02 and 34.53, can find out that worst error is no more than 1dB, reaches the object that objective quality coding is determined in expection.

Claims (2)

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:

$B_{ij}=round\left(\frac{A_{ij}}{Q_{step}}\right)$

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:

$PSNR=20\lg \sqrt{\frac{{255}^2}{MSE}},MSE=\frac{1}{M\&times;N}\underset{i=0}{\overset{M-1}{\&Sigma;}}{\underset{j=0}{\overset{N-1}{\&Sigma;}}\&lsqb;I(i,j)-\tilde{I}(i,j)\&rsqb;}^2$

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:

$\begin{array}{c}IAMD1=\frac{1}{(M-1)\&times;N}\underset{i=0}{\overset{M-2}{\&Sigma;}}\underset{j=0}{\overset{N-1}{\&Sigma;}}|x(i,j)-x(i+1,j)|\\ +\frac{1}{M\&times;(N-1)}\underset{i=0}{\overset{M-2}{\&Sigma;}}\underset{j=0}{\overset{N-2}{\&Sigma;}}|x(i,j)-x(i,j+1)|\end{array}$

$\begin{array}{c}IAMD2=\frac{1}{(M-1)\&times;(N-1)}\underset{i=1}{\overset{M-1}{\&Sigma;}}\underset{j=0}{\overset{N-2}{\&Sigma;}}|x(i,j)-x(i-1,j+1)|\\ +\frac{1}{(M-1)\&times;(N-1)}\underset{i=0}{\overset{M-2}{\&Sigma;}}\underset{j=0}{\overset{N-2}{\&Sigma;}}|x(i,j)-x(i+1,j+1)|\end{array}$

$\begin{array}{c}IAME1=\frac{1}{(M-2)\&times;N}\underset{i=1}{\overset{M-2}{\&Sigma;}}\underset{j=0}{\overset{N-1}{\&Sigma;}}|x(i-1,j)-x(i+1,j)|\\ +\frac{1}{M\&times;(N-2)}\underset{i=0}{\overset{M-1}{\&Sigma;}}\underset{j=1}{\overset{N-2}{\&Sigma;}}|x(i,j-1)-x(i,j+1)|\end{array}$

$\begin{array}{c}IAME2=\frac{1}{(M-2)\&times;(N-2)}\underset{i=1}{\overset{M-2}{\&Sigma;}}\underset{j=1}{\overset{N-2}{\&Sigma;}}|x(i-1,j-1)-x(i+1,j+1)|\\ +\frac{1}{(M-2)\&times;(N-2)}\underset{i=1}{\overset{M-2}{\&Sigma;}}\underset{j=1}{\overset{N-2}{\&Sigma;}}|x(i+1,j-1)-x(i-1,j+1)|\end{array}$

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.