CN110636307B - Deterministic image coding method based on Gabor filter - Google Patents

Deterministic image coding method based on Gabor filter Download PDF

Info

Publication number
CN110636307B
CN110636307B CN201910792886.0A CN201910792886A CN110636307B CN 110636307 B CN110636307 B CN 110636307B CN 201910792886 A CN201910792886 A CN 201910792886A CN 110636307 B CN110636307 B CN 110636307B
Authority
CN
China
Prior art keywords
gabor filter
image
coding
coding matrix
matrix
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
Application number
CN201910792886.0A
Other languages
Chinese (zh)
Other versions
CN110636307A (en
Inventor
肖泽龙
何蕾
胡泰洋
薛文
张晋宇
周卓辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Science and Technology
Original Assignee
Nanjing University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing University of Science and Technology filed Critical Nanjing University of Science and Technology
Priority to CN201910792886.0A priority Critical patent/CN110636307B/en
Publication of CN110636307A publication Critical patent/CN110636307A/en
Application granted granted Critical
Publication of CN110636307B publication Critical patent/CN110636307B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/117—Filters, e.g. for pre-processing or post-processing
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/42—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a deterministic image coding method based on a Gabor filter, wherein the size of a coding matrix is related to the size and the measured number of a target image, and a two-dimensional Gabor filter is constructed; secondly, adjusting variable parameters in the Gabor filter to determine the Gabor filter, and then carrying out binarization on the obtained matrix to obtain a required coding matrix; and coding the target image by using the coding matrix, and finally reconstructing the coded image by using an orthogonal matching pursuit reconstruction algorithm. The invention can be applied to a millimeter wave compression sensing system aiming at the characteristics of millimeter wave images, realizes the compression imaging and reconstruction of the millimeter wave images, can effectively reduce the reconstruction time of the millimeter wave images and improve the reconstruction performance.

Description

Deterministic image coding method based on Gabor filter
Technical Field
The invention relates to the technical field of compressed sensing, in particular to a deterministic image coding method based on a Gabor filter.
Background
Because the traditional image coding needs full sampling recompression, the storage space required by a coding end is large, the complexity is high, the proposal and the development of a compressed sensing theory are induced, the Nyquist sampling theorem is broken through, and the original signal is accurately recovered from the linear projection lower than the Nyquist rate.
In recent years, a Compressed Sensing (CS) theory is gradually applied to a millimeter wave imaging system, and compared with the traditional millimeter wave sampling imaging, the CS sampling method which breaks through the nyquist sampling theorem has an application advantage that the sampling time of hardware equipment can be effectively shortened; the data storage capacity after compression sampling is less, and the space is saved; compared with the array scanning imaging system adopted at present, the imaging system adopting compressed sensing can realize single-channel coding rapid imaging and effectively reduce the system cost. The application of the CS theory enables the speed of the millimeter wave imaging system for acquiring the target image to be remarkably improved, and the method is more suitable for various complex occasions.
The construction of the coding matrix is the research focus of the compressed sensing theory, and determines the number of samples required by a target and determines the quality of an image reconstruction effect. The existing coding matrix is mainly constructed by a random coding matrix, such as a Bernoulli coding matrix, and the matrixes have a good reconstruction effect in experiments due to strong randomness, but are difficult to realize on hardware in practical application.
Disclosure of Invention
The invention aims to provide a deterministic image coding method based on a Gabor filter, which considers the quality, speed and practicability of image sampling and has better reconstruction effect.
The invention is realized by adopting the following scheme: a deterministic image coding method based on Gabor filter includes the following steps:
step 1, constructing a two-dimensional Gabor filter according to the size of a target image;
step 2, adjusting variable parameters in the Gabor filter, and determining the Gabor filter;
step 3, carrying out binarization on the matrix obtained in the step 2 to obtain a coding matrix, and multiplying the coding matrix by a target image to carry out coding;
and 4, decoding the coded image obtained in the step 3 by using an orthogonal matching pursuit reconstruction algorithm.
Compared with the prior art, the invention has the following remarkable advantages: the invention provides a deterministic image coding matrix based on a Gabor filter, which can be better applied to a millimeter wave imaging system, reduces the reconstruction time of an image, improves the reconstruction efficiency and is very suitable for sampling and processing millimeter wave images. The method can be used for constructing a compressed sensing coding matrix suitable for millimeter wave images, and rapid acquisition of target images is realized.
Drawings
FIG. 1 is a schematic diagram of a Gabor filter constructed according to the present invention under different parameters.
FIG. 2 is a schematic diagram of the coding matrix of the present invention under different Gabor parameters λ, θ, and σ.
Fig. 3 is a schematic diagram of experimental effects of the present invention and the toeplitz code matrix under a millimeter wave gun chart, and the original chart, the toeplitz code matrix effect chart, and the method effect chart are sequentially shown from left to right.
FIG. 4 is a line graph of PSNR and measurement data for a millimeter wave firearms map with a Toepltiz encoding matrix in accordance with the present invention.
Detailed Description
Based on the characteristics of more background information, strong sparsity, obvious characteristic regions and the like of millimeter wave images, in order to acquire characteristic information in coding information as much as possible, reduce background information and reconstruct images with excellent effects at the sampling rate as low as possible, the invention provides a deterministic image coding matrix based on a Gabor filter, and the coding matrix can be better applied to a millimeter wave imaging system, reduces the reconstruction time of the images, improves the reconstruction efficiency and is very suitable for sampling processing of millimeter wave images.
A deterministic image coding method based on Gabor filter includes the following steps:
step 1, constructing a two-dimensional Gabor filter according to the size of a target image, for example, the size of the image is M, and the size of the Gabor filter is M;
step 2, adjusting variable parameters in the Gabor filter to determine the Gabor filter; the variable parameters comprise sine function wavelength lambda, Gabor kernel function direction theta and Gaussian function standard deviation sigma;
step 3, carrying out binarization on the matrix obtained in the step 2 to obtain a coding matrix, and multiplying the coding matrix by a target image to carry out coding;
and 4, decoding the coded image obtained in the step 3 by using an orthogonal matching pursuit reconstruction algorithm.
Further, step 1 specifically includes the following steps:
designing a Gabor filter g (x, y) according to the size of the target image:
Figure GDA0003259169060000021
wherein: lambda, lambda,θ、σ、
Figure GDA00032591690600000312
Respectively representing the wavelength of the sine function, the direction of the Gabor kernel function, the standard deviation of the Gaussian function and the phase difference, and (x, y) representing the position of each point in the matrix.
Further, the step 2 specifically comprises the following steps:
selecting three variable parameters of sine function wavelength lambda, Gabor kernel function direction theta and standard deviation sigma of Gaussian function to obtain peak signal-to-noise ratio
Figure GDA0003259169060000031
For the determination criteria, M, N represents the row and column numbers of the image, x (i, j) represents the pixel value of each point of the reconstructed image,
Figure GDA0003259169060000032
the pixel value of each point of the target image is represented, and the corresponding Gabor filter g (x, y) with the maximum PSNR is selected.
Further, step 3 specifically includes the following steps:
step S31: carrying out binarization processing on g (x, y) obtained in the step 2 to obtain a coding matrix phi;
step S32: multiplying the encoding matrix phi with the target image to obtain an encoded image YiIs represented by Yi＝ΦX。
Further, step 4 specifically includes the following steps:
step S41: initialization, in which residual vectors
Figure GDA0003259169060000033
The iteration number i is 1;
step S42: finding residual vector B and column of coding matrix
Figure GDA0003259169060000034
Subscript λ corresponding to maximum value in producti，
Figure GDA0003259169060000035
Wherein N represents that the coding matrix has N columns;
step S43: updating index collections
Figure GDA0003259169060000036
Recording reconstructed atom sets in found coding matrices
Figure GDA0003259169060000037
Ψ represents the product of the coding matrix and the signal sparse basis;
step S44: derived from least squares
Figure GDA0003259169060000038
Wherein
Figure GDA0003259169060000039
A vector of the measured values is represented,
Figure GDA00032591690600000310
representing a reconstruction approximation of a sparse coefficient alpha of an original signal beta;
step S45: updating residual vectors
Figure GDA00032591690600000311
Step S46: judging whether i is larger than K, wherein K is the sparsity of the original signal beta, and if so, stopping iteration; if not, go to step S42.
The invention is further explained below with reference to the drawings and the embodiments.
Examples
As shown in fig. 1, the present embodiment provides a method for constructing a deterministic coding matrix based on a Gabor filter, which includes the following steps:
s1: a two-dimensional Gabor filter g (x, y) (256 × 256 pixels) is designed according to the size of the target image (256 × 256 pixels):
Figure GDA0003259169060000041
wherein: lambda, theta, sigma,
Figure GDA0003259169060000042
Respectively representing the wavelength of the sine function, the direction of the Gabor kernel function, the standard deviation of the Gaussian function and the phase difference, and (x, y) representing the position of each point in the matrix.
S2: adjusting variable parameters in the Gabor filter to determine the Gabor filter; the method specifically comprises the following steps:
at peak signal-to-noise ratio
Figure GDA0003259169060000043
For the determination of the criterion, three variable parameters are selected, namely a sine function wavelength lambda, a Gabor kernel function direction theta and a standard deviation sigma of a Gaussian function, in the invention, lambda is 10, theta is pi/4 and sigma is 10, wherein M, N respectively represents the number of image rows and columns, x (i, j) represents each point pixel value of the reconstructed image,
Figure GDA0003259169060000044
the pixel value of each point of the target image is represented, and the corresponding Gabor filter g (x, y) with the maximum PSNR is selected.
S3: carrying out binarization on the matrix obtained in the step S2 to obtain an encoding matrix, and multiplying the encoding matrix by a target image for encoding, wherein the encoding matrix specifically comprises the following steps:
s31: performing binarization processing on g (x, y) obtained in step S2 to obtain an encoding matrix Φ (256 × 256 pixels) of the text;
s32: multiplying the encoding matrix phi with the target image to obtain an encoded image YiIs represented by Yi＝ΦX。
S4: decoding the coded image obtained in the step S3 by using an Orthogonal Matching Pursuit (OMP) reconstruction algorithm; the method specifically comprises the following steps:
s41: initialization, in which residual vectors
Figure GDA0003259169060000045
The iteration number i is 1;
s42: finding residual vector B and column of coding matrix
Figure GDA0003259169060000051
Subscript λ corresponding to maximum value in producti，
Figure GDA0003259169060000052
Wherein N represents that the coding matrix has N columns;
s43: update index set Λi＝Λi-1∪{λiiRecording the set of reconstructed atoms in the found coding matrix
Figure GDA0003259169060000053
Ψ represents the product of the coding matrix and the signal sparse basis;
s44: derived from least squares
Figure GDA0003259169060000054
Wherein
Figure GDA0003259169060000055
A vector of the measured values is represented,
Figure GDA0003259169060000056
representing a reconstruction approximation of a sparse coefficient alpha of an original signal beta;
s45: updating residual vectors
Figure GDA0003259169060000057
S46: judging whether i is larger than K, wherein K is the sparsity of the original signal beta, and if so, stopping iteration; if not, go to step S42.
The parameters λ, θ, σ of the Gabor filter represent the wavelength of the sine function, the direction of the Gabor kernel function, and the standard deviation of the gaussian function, respectively. As shown in fig. 1 and 2, the λ, θ, and σ parameters for the Gabor filter are selected as follows: the three dimensions of 5, 10 and 15 are selected for lambda, the smaller the lambda is, the smaller the central pattern is, the direction theta is selected from pi/4, pi/2, 3 pi/2 and pi, the four directions determine the rotation angle of the pattern, the larger the sigma is, the more dispersed the energy is, the more concentrated the energy is, the smaller the sigma is, the smaller the central stripe number is, and 5, 10 and 15 are often selected. The total of 36 filters, the PSNR values are compared by a program to select lambda as 10(m), theta as pi/4 and sigma as 10 to form a deterministic coding matrix.
Through a Matlab tool, effect simulation experiments are carried out on millimeter wave gun images (256 pixels by 256 pixels) by using the image coding matrix, and it can be seen that under the condition that the same OMP reconstruction algorithm is used, when the sampling rates are all 50%, the PSNR under the Toepltiz coding matrix is 27.8059dB, and the PSNR under the method is 28.3601dB, so that the method is superior to the Toepltiz coding matrix method. The effect is shown in the following fig. 3 and table 1.
TABLE 1 PSNR comparison results
Figure GDA0003259169060000058
As can be seen from fig. 3, the boundary of the firearm is clearer and the image noise inside the firearm is smaller under the method of the present invention. As shown in fig. 4, the triangle broken line represents the reconstruction result under the toeplitz coding matrix, and the asterisk broken line represents the reconstruction result under the present invention. It can be found that the PSNR value of the method is better than that of a Toepltiz coding matrix under the sampling rate of 20% -50%, and the method is more suitable for image reconstruction under the low sampling rate. It can also be seen by comparing the time for processing the same millimeter wave image by the two methods that 836.3059ms is needed for processing one millimeter wave image by the Toeplitz coding matrix, only 395.5224ms is needed for the method of the present invention (the specific processing time is determined according to the performance of the computer, and a MacBook Pro 2.3GHz eight-Core Intel Core i9 processor and a 16GB memory are used in the machine), and the processing time of the present invention is far shorter than that of the Toeplitz coding matrix, so that the reconstruction time of the millimeter wave image can be effectively reduced, and the reconstruction performance can be improved.

Claims (4)

1. A deterministic image coding method based on Gabor filter is characterized by comprising the following steps:
step 1, constructing a two-dimensional Gabor filter according to the size of a target image;
step 2, adjusting variable parameters in the Gabor filter, and determining the Gabor filter; the specific method comprises the following steps:
selecting three variable parameters of sine function wavelength lambda, Gabor kernel function direction theta and Gaussian function standard deviation sigma to obtain peak signal-to-noise ratio
Figure FDA0003259169050000011
For the determination criteria, M, N represents the row and column numbers of the image, x (i, j) represents the pixel value of each point of the reconstructed image,
Figure FDA0003259169050000012
representing the pixel value of each point of the target image, and selecting a corresponding Gabor filter g (x, y) when the PSNR is maximum;
step 3, carrying out binarization on the matrix obtained in the step 2 to obtain a coding matrix, and multiplying the coding matrix by a target image to carry out coding;
and 4, decoding the coded image obtained in the step 3 by using an orthogonal matching pursuit reconstruction algorithm, and specifically comprising the following steps:
step S41: initialization, in which residual vectors
Figure FDA0003259169050000013
Representing a measurement vector, wherein the iteration number i is 1;
step S42: finding residual vector B and column of coding matrix
Figure FDA0003259169050000014
Subscript λ corresponding to maximum value in producti，
Figure FDA0003259169050000015
Wherein N represents that the coding matrix has N columns;
step S43: update index set Λi＝Λi-1∪{λiRecording the set of reconstructed atoms in the found coding matrix
Figure FDA0003259169050000016
Ψ represents the product of the coding matrix and the signal sparse basis;
step S44: derived from least squares
Figure FDA0003259169050000017
Wherein
Figure FDA0003259169050000018
A vector of the measured values is represented,
Figure FDA0003259169050000019
representing a reconstruction approximation of a sparse coefficient alpha of an original signal beta;
step S45: updating residual vectors
Figure FDA00032591690500000110
i＝i+1；
Step S46: judging whether i is larger than K, wherein K is the sparsity of the original signal beta, and if so, stopping iteration; if not, go to step S42.
2. The method of claim 1, wherein the target image size is M, and the Gabor filter size is M.
3. A Gabor filter based deterministic image coding method as claimed in claim 1 or 2, characterized in that the Gabor filter g (x, y) is designed according to the size of the target image:
Figure FDA0003259169050000021
x '═ xcos θ + ysin θ, y' ═ xsin θ + ycos θ; wherein: lambda, theta, sigma,
Figure FDA0003259169050000022
Respectively representing the wavelength of the sine function, the direction of the Gabor kernel function, the standard deviation of the Gaussian function and the phase difference, and (x, y) representing the position of each point in the matrix.
4. The Gabor-filter-based deterministic image coding method according to claim 1, wherein step 3 specifically comprises the steps of:
step S31: carrying out binarization processing on g (x, y) obtained in the step 2 to obtain a coding matrix phi;
step S32: multiplying the encoding matrix phi with the target image to obtain an encoded image YiIs represented by Yi＝ΦX。
CN201910792886.0A 2019-08-26 2019-08-26 Deterministic image coding method based on Gabor filter Active CN110636307B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910792886.0A CN110636307B (en) 2019-08-26 2019-08-26 Deterministic image coding method based on Gabor filter

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910792886.0A CN110636307B (en) 2019-08-26 2019-08-26 Deterministic image coding method based on Gabor filter

Publications (2)

Publication Number Publication Date
CN110636307A CN110636307A (en) 2019-12-31
CN110636307B true CN110636307B (en) 2022-03-18

Family

ID=68969114

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910792886.0A Active CN110636307B (en) 2019-08-26 2019-08-26 Deterministic image coding method based on Gabor filter

Country Status (1)

Country Link
CN (1) CN110636307B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103955922A (en) * 2014-04-17 2014-07-30 西安工程大学 Method for detecting flaws of printed fabric based on Gabor filter
CN104299223A (en) * 2014-08-21 2015-01-21 中国人民解放军63620部队 Different-source image matching method based on Gabor coding
WO2017161892A1 (en) * 2016-03-23 2017-09-28 深圳大学 Classification method for hyperspectral remote sensing image, and system for same

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6753965B2 (en) * 2001-01-09 2004-06-22 The University Of Hong Kong Defect detection system for quality assurance using automated visual inspection

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103955922A (en) * 2014-04-17 2014-07-30 西安工程大学 Method for detecting flaws of printed fabric based on Gabor filter
CN104299223A (en) * 2014-08-21 2015-01-21 中国人民解放军63620部队 Different-source image matching method based on Gabor coding
WO2017161892A1 (en) * 2016-03-23 2017-09-28 深圳大学 Classification method for hyperspectral remote sensing image, and system for same

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Gabor二进制编码异源图像匹配方法;涂国勇 等;《国防科技大学学报》;20151031;全文 *

Also Published As

Publication number Publication date
CN110636307A (en) 2019-12-31

Similar Documents

Publication Publication Date Title
Torfason et al. Towards image understanding from deep compression without decoding
KR101420550B1 (en) Methods, devices and computer-readable storage media for fast subspace projection of descriptor patches for image recognition
Hawe et al. Dense disparity maps from sparse disparity measurements
CN106952317B (en) Hyperspectral image reconstruction method based on structure sparsity
Zhang et al. Kbnet: Kernel basis network for image restoration
Thapa et al. A performance comparison among different super-resolution techniques
Zhong et al. Channel-level variable quantization network for deep image compression
CN108242065A (en) A kind of light field image compression sensing method based on LDPC matrixes
Ahn Compressive sensing and recovery for binary images
CN110636307B (en) Deterministic image coding method based on Gabor filter
Chai et al. CSENMT: A deep image compressed sensing encryption network via multi-color space and texture feature
CN106101732B (en) The vector quantization scheme of Fast Compression bloom spectrum signal
CN110363823B (en) Gabor filter-based random image coding method
CN109559357B (en) Wavelet packet threshold-based image block compressed sensing reconstruction method
CN116797676A (en) Method suitable for coded aperture compression spectrum polarization imaging reconstruction
CN117011703A (en) Power transmission line inspection image duplication removing method
Yeganli et al. Super-resolution using multiple structured dictionaries based on the gradient operator and bicubic interpolation
CN108460777B (en) Extraction, blocking, compression and reconstruction method for hyperspectrum of plant
Zhang et al. Sedenion polar harmonic Fourier moments and their application in multi-view color image watermarking
CN110266318B (en) Measurement matrix optimization method based on gradient projection algorithm in compressed sensing signal reconstruction
Shoberg et al. The multiresolution analysis with radial symmetry elements
Kasem et al. DRCS-SR: Deep robust compressed sensing for single image super-resolution
CN111932473A (en) Phase information noise reduction algorithm of multi-resolution sparse coding and storage medium
Liu et al. Sparse Representation Based Image Super‐resolution Using Large Patches
CN111416980B (en) High-resolution camera imaging method based on compressed coded aperture

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant