CN104683818B - Method for compressing image based on biorthogonal invariant set m ultiwavelet - Google Patents
Method for compressing image based on biorthogonal invariant set m ultiwavelet Download PDFInfo
- Publication number
- CN104683818B CN104683818B CN201510112891.4A CN201510112891A CN104683818B CN 104683818 B CN104683818 B CN 104683818B CN 201510112891 A CN201510112891 A CN 201510112891A CN 104683818 B CN104683818 B CN 104683818B
- Authority
- CN
- China
- Prior art keywords
- image
- row
- inner product
- extension
- sequence number
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The present invention discloses a kind of method for compressing image based on biorthogonal invariant set m ultiwavelet, and its step is：1. input image data；2.DC level shifts；3. composition electric-wave filter matrix is sentenced；4. the biorthogonal invariant set multi-wavelet transformation number of plies initializes；5. the height and width of expanded images；6. biorthogonal invariant set m ultiwavelet line translation；7. biorthogonal invariant set m ultiwavelet rank transformation；8. quantization parameter；9. arithmetic coding；10. rate-distortion optimized truncation；11. code stream organization.The present invention has symmetrical, tight branch, orthogonal biorthogonal not invariant set multi-wavelet filter using one group, biorthogonal not invariant set multi-wavelet transformation calculate when using inner product method, so that algorithm complex substantially reduces, after conversion energy and entropy intensity it is high have it is bigger openness, be advantageous to follow-up compressed encoding, and just handle with piecemeal and accelerate parallel.
Description
Technical field
The present invention relates to technical field of image processing, further relates in JPEG2000 scheme in Image Compression field
As a kind of method for compressing image based on biorthogonal invariant set m ultiwavelet realized in compressibility.The present invention can be used for various numbers
Word stationary image compression coding.
Background technology
With the development and application of multimedia and internet, traditional image compression algorithm can not meet practical application
Requirement, for this, International Standards Organization has formulated the new standard JPEG2000 of still image compression in November, 2000.The new mark
Standard uses the rate-distortion optimized truncation built-in code block encryption algorithm (EBCOT) based on wavelet transformation technique, achieves preferable figure
As compression effectiveness.But JPEG2000 systems need to pinpoint with floating-point two systems to realize 5/3 and 9/7 wavelet transformation.
What Xian Electronics Science and Technology University possessed " realizes the fixed point small echo of JPEG2000 compression of images in its patented technology application
Transform method " (number of patent application：201210148213, publication number：CN102685501B a kind of realization is disclosed in)
The fixed point small wave converting method of JPEG2000 compression of images.This method introduces BIBO gain control methods, uses 9/7 small echo BIBO
Gain determines the storage locating depth of wavelet transformation median, using 5/3 small echo BIBO gains determines 9/7 lifting wavelet transform amount
Change the selection mode and quantification manner of parameter, save JPEG2000 system memory resources and run time.But the patent skill
Weak point is existing for art, and 5/3 small echo and 9/7 small echo used due to the patented technology is all single wavelet, and every layer of small echo becomes
The feature that only can obtain on 4 different directions is changed, the intensity of energy and entropy is not also high after wavelet transformation, and wavelet transformation
When need to carry out boundary extension to image to cause Boundary Distortion, be unfavorable for piecemeal processing and parallel accelerate.
(Zhongshan University's journal is (certainly in paper " the Image Coding Algorithms research based on multi-wavelet transformation " by Liu Wei and Chen Dongli
Right science version) page 50 to page 53 of 5 phases in 2011) in propose a kind of Image Coding Algorithms method based on multi-wavelet transformation.
This method with reference to EBCOT (most preferably blocking embedded code block coding) algorithm, carries out various conventional multi-wavelet bases to gray level image
Compressed encoding.Although this method achieves certain compression effectiveness, still, the deficiency that this method still has is this method
The 2 simple multi-wavelet bases for multiplying 2 are only gived, it is difficult to the symmetry for embodying m ultiwavelet while having, short supportive, orthogonality
With high-order vanishing moment property, and Coding Compression Algorithm computational complexity not improve it is suitable with JPEG2000.
The content of the invention
It is an object of the invention to overcome the shortcomings of above-mentioned prior art, there is provided in a kind of JPEG2000 image compression systems
Based on biorthogonal invariant set multi-wavelet transformation implementation method.The present invention using one group there is symmetrical, tight branch, orthogonal and high-order to disappear
Lose the 4 of square characteristic and multiply 4 biorthogonal not invariant set multi-wavelet filter matrix, entering ranks again to the conversion of image advanced every trade becomes
Change, using the method for inner product during transformation calculations.The present invention does not need boundary extension, avoids Boundary Distortion, is advantageous at piecemeal
Reason and parallel acceleration, and the present invention algorithm complex while effect same with using method under traditional convolution is achieved
Be a quarter of conventional method, after the present invention converts to image in addition energy and entropy intensity it is high have it is bigger sparse
Property, be advantageous to follow-up compressed encoding.
To achieve the above object, method of the invention comprises the following steps：
(1) input image data：
Two-dimensional image data to be compressed is inputted in JPEG2000 image compression systems；
(2) DC level shifts：
DC level shifts are carried out to the two-dimensional image data to be compressed of input, obtain 0 symmetrical DC level shifts
View data afterwards；
(3) electric-wave filter matrix is formed：
By each biorthogonal invariant set multi-wavelet filter
WithAs a line of matrix, composition 4 multiplies 4 electric-wave filter matrix；
(4) the biorthogonal invariant set multi-wavelet transformation number of plies initializes；
(5) height and width of expanded images：
(5a) judge image after DC level shifts it is high whether the multiple for being 4, if so, then performing step (5c), otherwise, hold
Row step (5b)；
(5b) enters row bound symmetric extension to the height of image after DC level shifts, and the height for making image after extension is 4 multiple；
(5c) judge image after DC level shifts it is wide whether 4 multiple, if so, then performing step (6), otherwise, hold
Row step (5d)；
(5d) enters row bound symmetric extension to the width of image, make image after extension it is wide be 4 multiple, scheme after being expanded
Picture；
(6) biorthogonal invariant set m ultiwavelet line translation：
The row sequence number m of the row sequence number k of image after extension and line translation image is initialized as 1 by (6a) respectively；
1 to 4th four elements of electric-wave filter matrix the first row and image row k after extension are carried out inner product by (6b), will
First element of m rows of inner product result as line translation image, k represent the row sequence number of image after extension, and m represents line translation figure
The row sequence number of picture；5 to 8th four elements of electric-wave filter matrix the first row and image row k after extension are subjected to inner product, by
Second element of m rows of product result as line translation image；The rest may be inferred, by electric-wave filter matrix the first row and image after extension
4 × i+1 to 4 × i+4 of row k four elements carry out inner product, using inner product result as line translation image m rows i-th+
1 element, i represent pictorial element sequence number, and i span isW represents the width of image after extension；
The row of electric-wave filter matrix second and the 1 to 4th four elements of image row k after extension are carried out inner product by (6c), will
M row of the inner product result as line translation imageIndividual element, k represent the row sequence number of image, and m represents line translation image
Row sequence number, W represent the width of image after extension；By the four of the 5 to 8th of image row k after the row of electric-wave filter matrix second and extension the
Individual element carries out inner product, the m rows the using inner product result as line translation imageIndividual element；The rest may be inferred, by wave filter
Four elements of the row of matrix second and 4 × i+1 to 4 × i+4 of image row k after extension carry out inner product, and inner product result is made
For the m rows of line translation imageIndividual element, i represent pictorial element sequence number, and i span is
Electric-wave filter matrix the third line and the 1 to 4th four elements of image row k after extension are carried out inner product by (6d), will
M row of the inner product result as line translation imageIndividual element, k represent the row sequence number of image, and m represents line translation image
Row sequence number, W represent extension after image width；By electric-wave filter matrix the third line and the 5 to 8th of image row k after extension the
Four elements carry out inner products, the m rows the using inner product result as line translation imageIndividual element；The rest may be inferred, will filter
Four elements of ripple device matrix the third line and 4 × i+1 to 4 × i+4 of image row k after extension carry out inner product, by inner product knot
M row of the fruit as line translation imageIndividual element, i represent pictorial element sequence number, and i span is
1 to 4th four elements of electric-wave filter matrix fourth line and image row k after extension are carried out inner product by (6e), will
M row of the inner product result as line translation imageIndividual element, the expression row sequence number of k images, m represent line translation image
Row sequence number, W represent extension after image width；By electric-wave filter matrix fourth line and the 5 to 8th of image row k after extension the
Four elements carry out inner products, the m rows the using inner product result as line translation imageIndividual element；The rest may be inferred, will filter
Ripple device matrix fourth line and four elements of 4 × i+1 to 4 × i+4 of image row k after extension carry out inner product, by inner product knot
M row of the fruit as line translation imageIndividual element, i represent pictorial element sequence number, and i span is
The row sequence number m of the row sequence number k of image after extension and line translation image is added 1 by (6f) respectively；
Whether the row sequence number k of image is equal to the height of image after extension after extension after (6g) judgement plus 1, if so, then
To line translation image, step (7) is performed, otherwise, performs step (6b)；
(7) biorthogonal invariant set m ultiwavelet rank transformation：
The row sequence number r of the row sequence number n of image after extension and rank transformation image is initialized as 1 by (7a) respectively；
The 1 to 4th four elements that (7b) arranges electric-wave filter matrix the first row and line translation image n-th carry out inner product, will
First element of r row of inner product result as rank transformation image, n represent the row sequence number of image；By electric-wave filter matrix the first row
Inner product is carried out with the 5 to 8th four elements that line translation image n-th arranges, the is arranged using inner product result as the r of rank transformation image
Two elements；The rest may be inferred, four of 4 × i+1 to 4 × i+4 that electric-wave filter matrix the first row and line translation image n-th are arranged
Element carries out inner product, the r row i+1 elements using inner product result as rank transformation image, and i represents pictorial element sequence number, i's
Span isL represents the height of image after extension；
The 1 to 4th four elements that (7c) arranges the row of electric-wave filter matrix second and line translation image n-th carry out inner product, will
Inner product result arranges as the r of rank transformation imageIndividual element, n represent the row sequence number of image, and L represents image after extending
Highly；The 5 to 8th four elements that the row of electric-wave filter matrix second and line translation image n-th are arranged carry out inner product, by inner product result
R as rank transformation image arrangesIndividual element；The rest may be inferred, by the row of electric-wave filter matrix second and line translation image n-th
4 × i+1 to 4 × i+4 of row four elements carry out inner product, and the is arranged using inner product result as the r of rank transformation imageIndividual element, i represent pictorial element sequence number, and i span is
The 1 to 4th four elements that (7d) arranges electric-wave filter matrix the third line and line translation image n-th carry out inner product, will
Inner product result arranges as the r of rank transformation imageIndividual element, n represent the row sequence number of image, and L represents image after extension
Height；The 5 to 8th four elements that electric-wave filter matrix the third line and line translation image n-th are arranged carry out inner product, by inner product knot
Fruit arranges as the r of rank transformation imageIndividual element；The rest may be inferred, by electric-wave filter matrix the third line and line translation image
4 × i+1 to 4 × i+4 of n-th row four elements carry out inner product, and the is arranged using inner product result as the r of rank transformation imageIndividual element, i represent pictorial element sequence number, and i span is
The 1 to 4th four elements that (7e) arranges electric-wave filter matrix fourth line and line translation image n-th carry out inner product, will
Inner product result arranges as the r of rank transformation imageIndividual element, n represent the row sequence number of image, and L represents image after extension
Height；The 5 to 8th four elements that electric-wave filter matrix fourth line and line translation image n-th are arranged carry out inner product, by inner product knot
Fruit arranges as the r of rank transformation imageIndividual element；The rest may be inferred, by electric-wave filter matrix fourth line and line translation image
4 × i+1 to 4 × i+4 of n-th row four elements carry out inner product, and the is arranged using inner product result as the r of rank transformation imageIndividual element, i represent pictorial element sequence number, and i span is
The row sequence number r of the row sequence number n of image after extension and rank transformation image is added 1 by (7f) respectively；
Whether the row sequence number n of image is equal to the width of image after extension after extension after (7g) judgement plus 1, if so, then
Biorthogonal invariant set multi-wavelet transformation image is obtained, performs step (8), otherwise, performs step (7b)；
(8) more new images：
Orthogonally-persistent collection multi-wavelet transformation image is existedScope height andIn the range of wide part, make
For image after DC level shifts；
(9) the conversion number of plies of biorthogonal invariant set m ultiwavelet is subtracted 1；
(10) judge whether the biorthogonal invariant set multi-wavelet transformation number of plies is zero after subtracting 1, if so, image after being expanded
Multilayer biorthogonal invariant set multi-wavelet transformation coefficient, then step (11) is performed, otherwise, perform step (5)；
(11) quantization parameter：
By the multilayer biorthogonal invariant set multi-wavelet transformation coefficient quantization of image after extension；
(12) arithmetic coding：
Using the bit plane arithmetic coding module based on context of JPEG2000 image compression system Plays to small echo
Coefficient after conversion is handled, and obtains the code stream of arithmetic coding；
(13) rate-distortion optimization：
Rate-distortion optimized truncation module using JPEG2000 image compression system Plays is entered to the code stream of arithmetic coding
Row rate-distortion optimized truncation, record intercept point information；
(14) tissue code stream：
The code stream organization module of JPEG2000 image compression system Plays uses intercept point information, to the code of arithmetic coding
Stream carries out code stream organization, obtains JPEG2000 compressed bit stream.
The present invention compared with prior art, has the following advantages that：
First, due to multiplying 4 biorthogonal invariant set multi-wavelet filter matrix present invention employs 4, each biorthogonal is not
Change collection m ultiwavelet filtering transformation can obtain the feature on 16 different directions, overcome every layer of small echo of single wavelet in the prior art
Conversion only can obtain the shortcomings that feature on 4 different directions so that the present invention has the advantages of more direction represents image.
Second, employed the characteristics of due to biorthogonal invariant set multi-wavelet filter matrix of the present invention and during transformation calculations in
Long-pending method, overcome the shortcomings that intensity of single wavelet transformation energy and entropy in the prior art is not high so that present invention tool
There is the advantages of more preferable compression performance.
3rd, employed the characteristics of due to biorthogonal invariant set multi-wavelet filter matrix of the present invention and during transformation calculations in
Long-pending method, need to cause Boundary Distortion to image progress boundary extension when overcoming single wavelet conversion in the prior art, it is unfavorable
In piecemeal processing and parallel the shortcomings that accelerating so that the present invention has quick the advantages of realizing compression.
4th, the characteristics of 4 biorthogonal invariant set multi-wavelet filter matrix is multiplied due to the present invention 4 and adopt during transformation calculations
With the method for inner product, overcome and 2 multiply that 2 simple multi-wavelet bases Coding Compression Algorithm computational complexities are high to be lacked in the prior art
Point so that there is the present invention algorithm complex to reduce, and save operation time, quick the advantages of realizing compression.
Brief description of the drawings
Fig. 1 is the flow chart of the present invention.
Embodiment
Step is described in detail for 1 pair of of the invention realizing below in conjunction with the accompanying drawings.
Step 1. input image data.
Two-dimensional image data to be compressed is inputted in JPEG2000 image compression systems, pixel value uses 1~16 ratio
It is special.
Step 2.DC level shifts.
DC level shifts are carried out to the two-dimensional image data to be compressed of input, after obtaining 0 symmetrical DC level shifts
View data.
Step 3. forms electric-wave filter matrix.
By each biorthogonal invariant set multi-wavelet filter
WithAs a line of matrix, the electric-wave filter matrix that composition 4 multiplies 4 is shown below：
Wherein, H4×4Represent that 4 multiply 4 biorthogonal invariant set multi-wavelet filter matrix.
Biorthogonal invariant set multi-wavelet filter H4×4It is in text according to Charles A.Micchelli and Yuesheng Xu
Chapter " Reconstruction and Decomposition Algorithms for Biorthogonal
That is established in Multiwavelets " has the biorthogonal invariant set m ultiwavelet from imitative map feature theoretical, with self affine
Delta Region be support Interval, using constant function thereon as scaling function, having constructed one group has symmetrical, tight branch, just
The biorthogonal of friendship not invariant set multi-wavelet filter.The electric-wave filter matrix has most short support length, does not have support overlapping portion
Point, after wavelet decomposition is made can Accurate Reconstruction, non-boundary distortion effect, application when avoid boundary extension, just with piecemeal
Reason and parallel acceleration.
The step 4. biorthogonal invariant set multi-wavelet transformation number of plies initializes.
Biorthogonal invariant set multi-wavelet transformation is usually no more than 3 layers, generally takes 1 layer or 2 layers, therefore typically by biorthogonal not
Become the collection multi-wavelet transformation number of plies and be initialized as 1 or 2.
The height and width of step 5. expanded images.
(5a) judge image after DC level shifts it is high whether the multiple for being 4, if so, then performing step (5c), otherwise, hold
Row step (5b).
(5b) enters row bound symmetric extension to the height of image after DC level shifts, and the height for making image after extension is 4 multiple.
The wide of image is 4 multiple after (5c) judges DC level shifts, if so, then performing step 6, otherwise, performs
Step (5d).
(5d) enters row bound symmetric extension to the width of image, make image after extension it is wide be 4 multiple, scheme after being expanded
Picture.
By the height and width symmetric extension of image after DC level shifts be 4 multiple refer to, make the image after extension height and
Wide to be divided exactly by 4, the line number or columns of extension are in the range of 1 to 3.
Step 6. biorthogonal invariant set m ultiwavelet line translation.
The row sequence number m of the row sequence number k of image after extension and line translation image is initialized as 1 by (6a) respectively.
1 to 4th four elements of electric-wave filter matrix the first row and image row k after extension are carried out inner product by (6b), will
First element of m rows of inner product result as line translation image, k represent the row sequence number of image after extension, and m represents line translation figure
The row sequence number of picture；5 to 8th four elements of electric-wave filter matrix the first row and image row k after extension are subjected to inner product, by
Second element of m rows of product result as line translation image；The rest may be inferred, by electric-wave filter matrix the first row and image after extension
4 × i+1 to 4 × i+4 of row k four elements carry out inner product, using inner product result as line translation image m rows i-th+
1 element, i represent pictorial element sequence number, and i span isW represents the width of image after extension.
The row of electric-wave filter matrix second and the 1 to 4th four elements of image row k after extension are carried out inner product by (6c), will
M row of the inner product result as line translation imageIndividual element, k represent the row sequence number of image, and m represents line translation image
Row sequence number, W represent the width of image after extension；By the four of the 5 to 8th of image row k after the row of electric-wave filter matrix second and extension the
Individual element carries out inner product, the m rows the using inner product result as line translation imageIndividual element；The rest may be inferred, by wave filter
Four elements of the row of matrix second and 4 × i+1 to 4 × i+4 of image row k after extension carry out inner product, and inner product result is made
For the m rows of line translation imageIndividual element, i represent pictorial element sequence number, and i span is
Electric-wave filter matrix the third line and the 1 to 4th four elements of image row k after extension are carried out inner product by (6d), will
M row of the inner product result as line translation imageIndividual element, k represent the row sequence number of image, and m represents line translation figure
The row sequence number of picture, W represent the width of image after extension；By the 5 to 8th of image row k after electric-wave filter matrix the third line and extension the
Four elements carry out inner product, the m rows the using inner product result as line translation imageIndividual element；The rest may be inferred, will
Four elements of electric-wave filter matrix the third line and 4 × i+1 to 4 × i+4 of image row k after extension carry out inner product, by inner product
As a result the m rows as line translation imageIndividual element, i represent pictorial element sequence number, and i span is
1 to 4th four elements of electric-wave filter matrix fourth line and image row k after extension are carried out inner product by (6e), will
M row of the inner product result as line translation imageIndividual element, the expression row sequence number of k images, m represent line translation figure
The row sequence number of picture, W represent the width of image after extension；By the 5 to 8th of image row k after electric-wave filter matrix fourth line and extension the
Four elements carry out inner product, the m rows the using inner product result as line translation imageIndividual element；The rest may be inferred, will
Electric-wave filter matrix fourth line and four elements of 4 × i+1 to 4 × i+4 of image row k after extension carry out inner product, by inner product
As a result the m rows as line translation imageIndividual element, i represent pictorial element sequence number, and i span is
The row sequence number m of the row sequence number k of image after extension and line translation image is added 1 by (6f) respectively.
Whether the row sequence number k of image is equal to the height of image after extension after extension after (6g) judgement plus 1, if so, then
To line translation image, step 7 is performed, otherwise, performs step (6b).
Step 7. biorthogonal invariant set m ultiwavelet rank transformation.
The row sequence number r of the row sequence number n of image after extension and rank transformation image is initialized as 1 by (7a) respectively.
The 1 to 4th four elements that (7b) arranges electric-wave filter matrix the first row and line translation image n-th carry out inner product, will
First element of r row of inner product result as rank transformation image, n represent the row sequence number of image；By electric-wave filter matrix the first row
Inner product is carried out with the 5 to 8th four elements that line translation image n-th arranges, the is arranged using inner product result as the r of rank transformation image
Two elements；The rest may be inferred, four of 4 × i+1 to 4 × i+4 that electric-wave filter matrix the first row and line translation image n-th are arranged
Element carries out inner product, the r row i+1 elements using inner product result as rank transformation image, and i represents pictorial element sequence number, i's
Span isL represents the height of image after extension.
The 1 to 4th four elements that (7c) arranges the row of electric-wave filter matrix second and line translation image n-th carry out inner product, will
Inner product result arranges as the r of rank transformation imageIndividual element, n represent the row sequence number of image, and L represents image after extending
Highly；The 5 to 8th four elements that the row of electric-wave filter matrix second and line translation image n-th are arranged carry out inner product, by inner product result
R as rank transformation image arrangesIndividual element；The rest may be inferred, by the row of electric-wave filter matrix second and line translation image n-th
4 × i+1 to 4 × i+4 of row four elements carry out inner product, and the is arranged using inner product result as the r of rank transformation imageIndividual element, i represent pictorial element sequence number, and i span is
The 1 to 4th four elements that (7d) arranges electric-wave filter matrix the third line and line translation image n-th carry out inner product, will
Inner product result arranges as the r of rank transformation imageIndividual element, n represent the row sequence number of image, and L represents image after extension
Height；The 5 to 8th four elements that electric-wave filter matrix the third line and line translation image n-th are arranged carry out inner product, by inner product knot
Fruit arranges as the r of rank transformation imageIndividual element；The rest may be inferred, by electric-wave filter matrix the third line and line translation figure
As 4 × i+1 to 4 × i+4 of the n-th row four elements carry out inner product, the is arranged using inner product result as the r of rank transformation imageIndividual element, i represent pictorial element sequence number, and i span is
The 1 to 4th four elements that (7e) arranges electric-wave filter matrix fourth line and line translation image n-th carry out inner product, will
Inner product result arranges as the r of rank transformation imageIndividual element, n represent the row sequence number of image, and L represents image after extension
Height；The 5 to 8th four elements that electric-wave filter matrix fourth line and line translation image n-th are arranged carry out inner product, by inner product knot
Fruit arranges as the r of rank transformation imageIndividual element；The rest may be inferred, by electric-wave filter matrix fourth line and line translation image
4 × i+1 to 4 × i+4 of n-th row four elements carry out inner product, and the is arranged using inner product result as the r of rank transformation imageIndividual element, i represent pictorial element sequence number, and i span is
The row sequence number r of the row sequence number n of image after extension and rank transformation image is added 1 by (7f) respectively.
Whether the row sequence number n of image is equal to the width of image after extension after extension after (7g) judgement plus 1, if so, then
Biorthogonal invariant set multi-wavelet transformation image is obtained, performs step 8, otherwise, performs step (7b).
Step 8. more new images.
Orthogonally-persistent collection multi-wavelet transformation image is existedScope height andIn the range of wide part, make
For image after DC level shifts.
The conversion number of plies of biorthogonal invariant set m ultiwavelet is subtracted 1 by step 9..
Step 10. judges whether the biorthogonal invariant set multi-wavelet transformation number of plies is zero after subtracting 1, if so, after being expanded
The multilayer biorthogonal invariant set multi-wavelet transformation coefficient of image, then step 11 is performed, otherwise, perform step 5.
The characteristics of this biorthogonal invariant set multi-wavelet transformation process make use of biorthogonal invariant set multi-wavelet filter is adopted
With the method for inner product, so as to obtain transform effect with using method same under traditional convolution, algorithm complex is tradition
The advantages of a quarter of method, and energy and entropy intensity are high after converting has bigger openness, are advantageous to follow-up
Compressed encoding.
Step 11. quantization parameter.
JPEG2000 is utilized by the multilayer biorthogonal invariant set multi-wavelet transformation coefficient quantization of the image after extension, during quantization
The international standard ISO/IEC 15444-1 of image compression system:2000 quantizing process.
Due to the feature on 16 different directions of the original image of every layer of biorthogonal invariant set multi-wavelet transformation, and every layer every
9/7 wavelet transformation of layer can only obtain the feature on 4 different directions of original image, in order to utilize JPEG2000 compression of images system
Unite the quantizing process of Plays, biorthogonal invariant set multi-wavelet transformation and 9/7 wavelet transformation can be done one it is corresponding.When to corresponding to
Mainly biorthogonal invariant set multi-wavelet transformation is mapped can with the low frequency part of 9/7 wavelet transformation, other directions can
To carry out as needed correspondingly.
Step 12. arithmetic coding.
Using the bit plane arithmetic coding module based on context of JPEG2000 image compression system Plays to small echo
Coefficient after conversion is handled, and obtains the code stream of arithmetic coding.
Step 13. rate-distortion optimization.
Rate-distortion optimized truncation module using JPEG2000 image compression system Plays is entered to the code stream of arithmetic coding
Row rate-distortion optimized truncation, record intercept point information.
Step 14. organizes code stream.
The code stream organization module of JPEG2000 image compression system Plays uses intercept point information, to the code of arithmetic coding
Stream carries out code stream organization, obtains JPEG2000 compressed bit stream.
Due to obtained JPEG2000 compressed bit stream, wavelet transformation part becomes using biorthogonal invariant set m ultiwavelet
Change, so when this JPEG2000 compressed bit stream is decompressed, inverse wavelet transform part also needs to use biorthogonal accordingly
Invariant set m ultiwavelet inverse transformation.
Claims (3)
1. a kind of method for compressing image based on biorthogonal invariant set m ultiwavelet, comprises the following steps：
(1) input image data：
Two-dimensional image data to be compressed is inputted in JPEG2000 image compression systems；
(2) DC level shifts：
DC level shifts are carried out to the two-dimensional image data to be compressed of input, after obtaining 0 symmetrical DC level shifts
View data；
(3) electric-wave filter matrix is formed：
By each biorthogonal invariant set multi-wavelet filter
WithAs a line of matrix, composition 4 multiplies 4 electric-wave filter matrix；
(4) the biorthogonal invariant set multi-wavelet transformation number of plies initializes；
(5) height and width of expanded images：
(5a) judge image after DC level shifts it is high whether the multiple for being 4, if so, then performing step (5c), otherwise, perform step
Suddenly (5b)；
The multiple that the described height and width symmetric extension by image after DC level shifts is 4 refers to, make after extension the height of image and
Wide to be divided exactly by 4, the line number or columns of extension are in the range of 1 to 3；
(5b) enters row bound symmetric extension to the height of image after DC level shifts, and the height for making image after extension is 4 multiple；
(5c) judge image after DC level shifts it is wide whether 4 multiple, if so, then performing step (6), otherwise, perform step
Suddenly (5d)；
(5d) enters row bound symmetric extension to the width of image, make image after extension it is wide be 4 multiple, image after being expanded；
(6) biorthogonal invariant set m ultiwavelet line translation：
The row sequence number m of the row sequence number k of image after extension and line translation image is initialized as 1 by (6a) respectively；
1 to 4th four elements of electric-wave filter matrix the first row and image row k after extension are carried out inner product by (6b), by inner product
As a result first element of m rows as line translation image, k represent the row sequence number of image after extension, and m represents line translation image
Row sequence number；5 to 8th four elements of electric-wave filter matrix the first row and image row k after extension are subjected to inner product, by inner product knot
Second element of m rows of fruit as line translation image；The rest may be inferred, by electric-wave filter matrix the first row and image kth after extension
Capable 4 × i+1 to 4 × i+4 four elements carry out inner product, the m rows i+1 using inner product result as line translation image
Element, i represent pictorial element sequence number, and i span isW represents the width of image after extension；
The row of electric-wave filter matrix second and the 1 to 4th four elements of image row k after extension are carried out inner product by (6c), by inner product
As a result the m rows as line translation imageIndividual element, k represent the row sequence number of image, and m represents the row sequence of line translation image
Number, W represents the width of image after extension；By the row of electric-wave filter matrix second and the 5 to 8th four members of image row k after extension
Element carries out inner product, the m rows the using inner product result as line translation imageIndividual element；The rest may be inferred, by electric-wave filter matrix
Four elements of the second row and 4 × i+1 to 4 × i+4 of image row k after extension carry out inner product, using inner product result as row
The m rows of changing imageIndividual element, i represent pictorial element sequence number, and i span is
Electric-wave filter matrix the third line and the 1 to 4th four elements of image row k after extension are carried out inner product by (6d), by inner product
As a result the m rows as line translation imageIndividual element, k represent the row sequence number of image, and m represents the row of line translation image
Sequence number, W represent the width of image after extension；By four of the 5 to 8th of image row k after electric-wave filter matrix the third line and extension the
Element carries out inner product, the m rows the using inner product result as line translation imageIndividual element；The rest may be inferred, by wave filter
Four elements of matrix the third line and 4 × i+1 to 4 × i+4 of image row k after extension carry out inner product, and inner product result is made
For the m rows of line translation imageIndividual element, i represent pictorial element sequence number, and i span is
1 to 4th four elements of electric-wave filter matrix fourth line and image row k after extension are carried out inner product by (6e), by inner product
As a result the m rows as line translation imageIndividual element, the expression row sequence number of k images, m represent the row of line translation image
Sequence number, W represent the width of image after extension；By four of the 5 to 8th of image row k after electric-wave filter matrix fourth line and extension the
Element carries out inner product, the m rows the using inner product result as line translation imageIndividual element；The rest may be inferred, by wave filter
Matrix fourth line and four elements of 4 × i+1 to 4 × i+4 of image row k after extension carry out inner product, and inner product result is made
For the m rows of line translation imageIndividual element, i represent pictorial element sequence number, and i span is
The row sequence number m of the row sequence number k of image after extension and line translation image is added 1 by (6f) respectively；
Whether the row sequence number k of image is equal to the height of image after extension after extension after (6g) judgement plus 1, if so, then being gone
Changing image, step (7) is performed, otherwise, perform step (6b)；
(7) biorthogonal invariant set m ultiwavelet rank transformation：
The row sequence number r of the row sequence number n of image after extension and rank transformation image is initialized as 1 by (7a) respectively；
The 1 to 4th four elements that (7b) arranges electric-wave filter matrix the first row and line translation image n-th carry out inner product, by inner product
As a result first element of r row as rank transformation image, n represent the row sequence number of image；By electric-wave filter matrix the first row and row
The 5 to 8th four elements that changing image n-th arranges carry out inner product, the r row second using inner product result as rank transformation image
Element；The rest may be inferred, four elements for 4 × i+1 to 4 × i+4 that electric-wave filter matrix the first row and line translation image n-th are arranged
Inner product, the r row i+1 elements using inner product result as rank transformation image are carried out, i represents pictorial element sequence number, i value
Scope isL represents the height of image after extension；
The 1 to 4th four elements that (7c) arranges the row of electric-wave filter matrix second and line translation image n-th carry out inner product, by inner product
As a result the r as rank transformation image arrangesIndividual element, n represent the row sequence number of image, and L represents the height of image after extension
Degree；The 5 to 8th four elements that the row of electric-wave filter matrix second and line translation image n-th are arranged carry out inner product, and inner product result is made
Is arranged for the r of rank transformation imageIndividual element；The rest may be inferred, and the row of electric-wave filter matrix second and line translation image n-th are arranged
4 × i+1 to 4 × i+4 four elements carry out inner product, arrange the using inner product result as the r of rank transformation image
Individual element, i represent pictorial element sequence number, and i span is
The 1 to 4th four elements that (7d) arranges electric-wave filter matrix the third line and line translation image n-th carry out inner product, by inner product
As a result the r as rank transformation image arrangesIndividual element, n represent the row sequence number of image, and L represents the height of image after extension
Degree；The 5 to 8th four elements that electric-wave filter matrix the third line and line translation image n-th are arranged carry out inner product, and inner product result is made
Is arranged for the r of rank transformation imageIndividual element；The rest may be inferred, by electric-wave filter matrix the third line and line translation image n-th
4 × i+1 to 4 × i+4 of row four elements carry out inner product, and the is arranged using inner product result as the r of rank transformation imageIndividual element, i represent pictorial element sequence number, and i span is
The 1 to 4th four elements that (7e) arranges electric-wave filter matrix fourth line and line translation image n-th carry out inner product, by inner product
As a result the r as rank transformation image arrangesIndividual element, n represent the row sequence number of image, and L represents the height of image after extension
Degree；The 5 to 8th four elements that electric-wave filter matrix fourth line and line translation image n-th are arranged carry out inner product, and inner product result is made
Is arranged for the r of rank transformation imageIndividual element；The rest may be inferred, by electric-wave filter matrix fourth line and line translation image n-th
4 × i+1 to 4 × i+4 of row four elements carry out inner product, and the is arranged using inner product result as the r of rank transformation imageIndividual element, i represent pictorial element sequence number, and i span is
The row sequence number r of the row sequence number n of image after extension and rank transformation image is added 1 by (7f) respectively；
Whether the row sequence number n of image is equal to the width of image after extension after extension after (7g) judgement plus 1, if so, then obtaining
Biorthogonal invariant set multi-wavelet transformation image, step (8) is performed, otherwise, perform step (7b)；
(8) more new images：
Orthogonally-persistent collection multi-wavelet transformation image is existedScope height andIn the range of wide part, as DC
Image after level shift；
(9) the conversion number of plies of biorthogonal invariant set m ultiwavelet is subtracted 1；
(10) judge subtract 1 after the biorthogonal invariant set multi-wavelet transformation number of plies whether be zero, if so, after being expanded image multilayer
Biorthogonal invariant set multi-wavelet transformation coefficient, then step (11) is performed, otherwise, perform step (5)；
(11) quantization parameter：
By the multilayer biorthogonal invariant set multi-wavelet transformation coefficient quantization of image after extension；
(12) arithmetic coding：
Using the bit plane arithmetic coding module based on context of JPEG2000 image compression system Plays to wavelet transformation
Coefficient afterwards is handled, and obtains the code stream of arithmetic coding；
(13) rate-distortion optimization：
Rate is carried out to the code stream of arithmetic coding using the rate-distortion optimized truncation module of JPEG2000 image compression system Plays
Aberration optimizing intercepts, and records intercept point information；
(14) tissue code stream：
The code stream organization module of JPEG2000 image compression system Plays uses intercept point information, and the code stream of arithmetic coding is entered
Row code stream organization, obtain JPEG2000 compressed bit stream.
2. the method for compressing image according to claim 1 based on biorthogonal invariant set m ultiwavelet, it is characterised in that：Step
(1) two-dimensional image data to be compressed described in uses 1~16 bit.
3. the method for compressing image according to claim 1 based on biorthogonal invariant set m ultiwavelet, it is characterised in that：Step
(11) the invariant set multi-wavelet transformation coefficient quantization of multilayer biorthogonal described in refers to, utilizes the state of JPEG2000 image compression systems
Border standard ISO/IEC 15444-1:2000 quantizing process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510112891.4A CN104683818B (en) | 2015-03-15 | 2015-03-15 | Method for compressing image based on biorthogonal invariant set m ultiwavelet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510112891.4A CN104683818B (en) | 2015-03-15 | 2015-03-15 | Method for compressing image based on biorthogonal invariant set m ultiwavelet |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104683818A CN104683818A (en) | 2015-06-03 |
| CN104683818B true CN104683818B (en) | 2017-11-21 |
Family
ID=53318298
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510112891.4A Active CN104683818B (en) | 2015-03-15 | 2015-03-15 | Method for compressing image based on biorthogonal invariant set m ultiwavelet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104683818B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR3022064A1 (en) * | 2014-06-04 | 2015-12-11 | I Ces Innovative Compression Engineering Solutions | ADAPTIVE PRECISION AND QUANTIFICATION OF WAVELET-TRANSFORMED MATRIX |
| CN106056158B (en) * | 2016-06-03 | 2019-06-21 | 西安电子科技大学 | Template image global search method based on exposure mask matrix and Fast Fourier Transform |
| CN107071482B (en) * | 2017-03-31 | 2019-12-10 | 东南大学 | Image compression and reconstruction method based on multilayer discrete wavelet transform coefficients |
| CN111885390B (en) * | 2020-07-30 | 2021-04-13 | 河南大学 | High spectral image compression method based on fractal multi-wavelet |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102685501A (en) * | 2012-05-14 | 2012-09-19 | 西安电子科技大学 | Fixed-point wavelet transform method for joint photographic experts group 2000 (JPEG2000) image compression |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB0321954D0 (en) * | 2003-09-18 | 2003-10-22 | Wave Ltd M | Data compression |
-
2015
- 2015-03-15 CN CN201510112891.4A patent/CN104683818B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102685501A (en) * | 2012-05-14 | 2012-09-19 | 西安电子科技大学 | Fixed-point wavelet transform method for joint photographic experts group 2000 (JPEG2000) image compression |
Non-Patent Citations (1)
| Title |
|---|
| JPEG2000 算法中基于有界输入有界输出增益控制的小波变换定点实现技术;张静 等;《电子与信息学报》;20121231;第34卷(第12期);2802-2808 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104683818A (en) | 2015-06-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104683818B (en) | Method for compressing image based on biorthogonal invariant set m ultiwavelet | |
| US9179162B2 (en) | Image transform zero coefficient selection and zero-skip transmission for arbitrary shape transform coding | |
| US8024389B2 (en) | Inverse discrete cosine transform supporting multiple decoding processes | |
| US7634148B2 (en) | Image signal transforming and inverse-transforming method and computer program product with pre-encoding filtering features | |
| CN104331913A (en) | Polarized SAR polarization method based on sparse K-SVD (Singular Value Decomposition) | |
| Yadav et al. | A review on image compression techniques | |
| Kumar et al. | A review: DWT-DCT technique and arithmetic-Huffman coding based image compression | |
| Mulla et al. | Image compression scheme based on zig-zag 3D-DCT and LDPC coding | |
| EP2869575B1 (en) | Adaptive inter-channel transform for wavelet color image compression | |
| Deshlahra et al. | A comparative study of DCT, DWT & hybrid (DCT-DWT) transform | |
| Sadkhan | A proposed image compression technique based on DWT and predictive techniques | |
| CN112906874A (en) | Convolutional neural network characteristic graph data compression method and device | |
| US8031782B2 (en) | Systems and methods for compression, transmission and decompression of video codecs | |
| Divakara et al. | High speed modular systolic array-based DTCWT with parallel processing architecture for 2D image transformation on FPGA | |
| Rani et al. | Comparative analysis of image compression using dct and dwt transforms | |
| Thepade et al. | Varying proportions of constituent transforms to generate hybrid wavelet transform for image compression | |
| Kountchev et al. | Decorrelation of multispectral images, based on hierarchical adaptive PCA | |
| CN109255770B (en) | Image transform domain down-sampling method | |
| Zhang et al. | Color image compression based on directional all phase biorthogonal transform | |
| Mehta et al. | Image Compression using DCT and DWT-Technique | |
| Rajeshwari et al. | DWT based Multimedia Compression | |
| Mallikarjuna et al. | Compression of noisy images based on sparsification using discrete rajan transform | |
| Wiewel | FPGA implementation of an energy-efficient real-time image compression algorithm for the EIVE satellite mission | |
| Yang et al. | Infinity-norm rotation transforms | |
| Manoria et al. | An efficient DCT compression technique using Strassen's matrix multiplication algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |