CN103793897A - Digital image halftone method based on small-wavelet-domain multi-scale information fusion - Google Patents

Digital image halftone method based on small-wavelet-domain multi-scale information fusion Download PDF

Info

Publication number
CN103793897A
CN103793897A CN201410016561.0A CN201410016561A CN103793897A CN 103793897 A CN103793897 A CN 103793897A CN 201410016561 A CN201410016561 A CN 201410016561A CN 103793897 A CN103793897 A CN 103793897A
Authority
CN
China
Prior art keywords
wavelet
image
measure function
error measure
represent
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.)
Granted
Application number
CN201410016561.0A
Other languages
Chinese (zh)
Other versions
CN103793897B (en
Inventor
何自芬
詹肇麟
张印辉
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.)
Kunming University of Science and Technology
Original Assignee
Kunming 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 Kunming University of Science and Technology filed Critical Kunming University of Science and Technology
Priority to CN201410016561.0A priority Critical patent/CN103793897B/en
Publication of CN103793897A publication Critical patent/CN103793897A/en
Application granted granted Critical
Publication of CN103793897B publication Critical patent/CN103793897B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention relates to a digital image halftone method based on small-wavelet-domain multi-scale information fusion and belongs to the technical field of digital image processing before printing. The digital image halftone method based on small-wavelet-domain multi-scale information fusion comprises the steps that a continuous-tone image is converted into a grayscale image, and whether the grayscale image is a standard 2n*2n grayscale image is judged; small-wavelet-domain four-scale information of the standard grayscale image is obtained through two-dimension discrete wavelet transformation, and an edge error measure function is obtained by means of wavelet detail coefficient relation between autocorrelation fusion layers of inter-scale wavelet coefficients; the standard grayscale image is divided into k areas with the K-means cluster method, and an area error measure function is established with the reciprocals of variances of the areas serving as weights; a mixed error measure function is established according to the additivity of the edge error measure function and the area error measure function, and then errors of the standard grayscale image and a rough halftone image are minimized with the direct binary searching method. The digital image halftone method based on small-wavelet-domain multi-scale information fusion solves the problem that according to an existing method, the smoothness and the definition of images can not be guaranteed at the same time.

Description

A kind of digital picture halftoning method based on wavelet multi-scale information fusion
Technical field
The present invention relates to a kind of digital picture halftoning method based on wavelet multi-scale information fusion, belong to digital picture and print pretreatment technology field.
Background technology
The engraving mode of laser gravure plate-making is a kind of scan-type engraving method that imitates print out equipment.The way of output of Laser engraved gravure is the on off state by controlling laser, and high energy laser beam is radiated on the specified point of surface of the work, makes this point produce rapidly evaporation, forms a depression points, thereby show two kinds of tones of black and white at surface of the work.So for multi-grey image, must convert it to the black and white binary image that is applicable to Laser output by digital halftone technology before output.Remove outside the factor such as beam quality and rapidoprint characteristic of optical maser wavelength, pulsed frequency, laser, the halftoning algorithm of multi-grey image plays a part especially crucial to laser image output effect.
Digital picture shadow tone is continuous toned image is developed on the two-value equipment such as such as laser platemaker, digital printer, laser printer and in human visual system, produce the gordian technique of continuous toned image illusion.It has obtained application more and more widely in producing, live people.The current small desk ink-jet from family, office use, laser printer, laser platemaker are to large-scale publication and printing system, and digital halftone technology can be described as ubiquitous.Digital halftone technology also can be applicable to data and hides and digital watermark technology, and 3-dimensional digital halftone technique is also used widely in layering manufacturing technology.Existing halftoning method carries out image modeling from single yardstick, although obtain successful Application to a certain extent under static environment, under dynamic environment, half tone image quality is unsatisfactory.Reason is because the information of fine dimension under dynamic environment is very unreliable, makes space, frequency self-adaption error in single yardstick estimate information and have stronger dynamic random.And due to human visual system's attention mechanism, people can be observed image under different resolution, and then picture quality is produced and evaluated, but be subject to the impact such as such as illumination variation, shade, background clutter extraneous factor, the image information in fine dimension has stronger dynamic random on time and space.Multiresolution O&A mechanism under this dynamic environment just, require to consider the multi-scale information that error is estimated in digital picture halftone process process, relate to how merging and comprise thick scale error and estimate and estimate information in interior multiple dimensioned error and realize this technical barrier of digital halftone.
Summary of the invention
The invention provides a kind of digital picture halftoning method based on wavelet multi-scale information fusion, comprise thick scale error and estimate and estimate information in interior multiple dimensioned error and realize this technical matters of digital halftone for solving under dynamic environment multiresolution O&A mechanism, merging.
Technical scheme of the present invention is: a kind of digital picture halftoning method based on wavelet multi-scale information fusion, and the concrete steps of described method are as follows:
A, convert continuous toned image to gray level image and judge whether gray level image is 2 of standard n× 2 ngray level image;
B, employing two-dimensional discrete wavelet conversion obtain four yardstick information of the wavelet field of standard grayscale image, and the wavelet details Relationship of Coefficients between recycling yardstick between the autocorrelation fused layer of wavelet coefficient is set up marginal error measure function;
C, employing K-means clustering procedure are divided into standard grayscale image respectively kindividual region, recycles each Local Deviation inverse as weight, sets up domain error measure function;
D, the additive property of marginal error measure function and domain error measure function is set up to combined error measure function, then adopt direct two-value searching method to minimize the error of standard grayscale image and coarse half tone image, obtain optimum digital halftone image.
In described step B, the concrete steps that marginal error measure function is set up are as follows:
The Harr small echo that B1, employing have orthogonality carries out two-dimensional discrete wavelet conversion, realizes four yardstick wavelet transformations of gray level image and obtains four yardstick wavelet images;
B2, according to four yardstick wavelet images obtain level, vertical and diagonal angle subband wavelet coefficient distributes;
B3, independently suppose according to wavelet sub-band, level, vertical and three, diagonal angle direction wavelet coefficient are normalized, and the wavelet coefficient obtaining after normalized distributes;
B4, utilize the autocorrelation of wavelet coefficient between yardstick to merge the each yardstick wavelet coefficient after normalization, the wavelet coefficient after being merged;
B5, using merge after wavelet coefficient as weight, set up marginal error measure function
Figure DEST_PATH_IMAGE002
; Wherein, lrepresent wavelet transformation progression; lH, hL, hHrepresent respectively multi-scale wavelet territory vertical direction, horizontal direction and to angular direction wavelet coefficient fuse information; i,jpresentation video pixel; b i,j represent pixel eight neighborhoods; z i,j represent the collimation error; w i,j represent model printer error.
In described step C, the concrete steps that domain error measure function is set up are as follows:
C1, employing K-means clustering procedure are divided into standard grayscale image kindividual area image, wherein k=2,3,4;
C2, the average of calculating each region and variance; Using the inverse of variance as weight, utilize weighted least squares to set up domain error measure function
Figure DEST_PATH_IMAGE004
; Wherein, k=1,2,3,4 represent cut zone; λ k represent weight coefficient; i,jpresentation video pixel; b i,j represent pixel eight neighborhoods; z i, j represent the collimation error; w i, j represent model printer error.
In described step D, the concrete steps that direct two-value searching algorithm minimizes combined error measure function are as follows:
D1, the additive property of marginal error measure function and domain error measure function is set up to combined error measure function
Figure DEST_PATH_IMAGE006
;
D2, utilize in direct two-value searching algorithm conversion and the exchange of jumping for eight neighborhood regions of each pixel in standard grayscale image and coarse half tone image, make combined error measure function minimum;
D3, when entire image combined error measure function hour, algorithm stops, and obtains optimum half tone image.
Described coarse half tone image adopts error-diffusion method processing to obtain.
Principle of work of the present invention is:
Combined error measure function formulation process:
One, set up marginal error measure function:
If image is after level Four wavelet decomposition jwavelet coefficient on yardstick is c j,k , scale correlations coefficient is w j,k = c j,k × c j+1 , k .Between wavelet coefficient interband and the brotgher of node of same father node, in band, wavelet coefficient all has stronger correlativity.Utilize amplitude wavelet coefficient in conjunction with correlativity in interband band, define relative coefficient between new wavelet coefficient. p j,k represent junder yardstick kthe all child node wavelet coefficients of individual wavelet coefficient maximal value:
Figure DEST_PATH_IMAGE008
, λ j,k , λ ' j,k represent that small echo two enters to cut apart, p j,k comprised ( j,k) all child node information when position, wavelet coefficient is decayed along with yardstick, and different yardsticks is multiplied by different scale factor 2 - js , ( svalue is determined by experiment), reach the order at outstanding its edge.Define new maximum child node related coefficient:
Figure DEST_PATH_IMAGE010
.Related coefficient r j,k be equal to the product of all wavelet coefficients.Revise r j,k for the product of adjacent yardstick, define the related coefficient of new related coefficient-adjacent maximum child node
Figure DEST_PATH_IMAGE012
.After having determined related coefficient, for making related coefficient and wavelet coefficient there is comparability, related coefficient is normalized, obtains new wavelet coefficient d j,k .In wavelet field, wavelet coefficient is distributed on different yardstick and subband, independently supposes according to wavelet sub-band, and wavelet coefficient adds up independent on yardstick and subband, utilize the method that between yardstick, wavelet coefficient merges to obtain fine dimension Wavelet Fusion coefficient, considered the approximate information of image simultaneously.Wavelet coefficient after fusion can be as weight coefficient, sets up boundary error measure function.
Figure DEST_PATH_IMAGE014
Two, set up domain error measure function:
At uniform cartesian grid
Figure DEST_PATH_IMAGE016
in ube defined as a secondary continuous toned image, in normal range scale, suppose
Figure DEST_PATH_IMAGE018
.?
Figure 118599DEST_PATH_IMAGE016
in to any image
Figure DEST_PATH_IMAGE020
application linear factor
Figure DEST_PATH_IMAGE022
represent human visual system's (HVS) the perception factor, grey level range is [0,1] (no matter whether continuous toned image or half tone image),
Figure DEST_PATH_IMAGE024
be assumed to be it is weak low pass, therefore
Figure DEST_PATH_IMAGE026
.
HVS model by Mannos and Sakrison can well predictive-coded picture quality:
Figure DEST_PATH_IMAGE028
Frequency
Figure DEST_PATH_IMAGE030
, be horizontal frequency f x and vertical frequency f y effective value, unit is cycles/deg.
Suppose that pixel is square,
Figure DEST_PATH_IMAGE032
( tbe the length of a cell).Adjacent stain may the overlapping and adjacent white point in meeting cover part.The radius of point is not less than
Figure DEST_PATH_IMAGE034
, only in this way, just can make the complete blackening of print image zone.This means between stain and adjacent white point and always have some overlapping regions, cause having produced the gray shade scale of pixel on white point, make printing images occur distortion.A kind of simple model printer is called " circular shaped lattice point double exposure " model and in the time evaluating the gray shade scale of the each pixel of output image, has considered this distortion.Pixel in cell ( i,j) central point be ( x i , y j ), wherein x i= iT x+ t x / 2, y j= jT y+ t y / 2, the bianry image of output be [ b i,j ], when b i,j within=1 o'clock, represent pixel center ( x i , y j ) be stain, b i,j within=0 o'clock, represent pixel center ( x i , y j ) be white point.
Figure DEST_PATH_IMAGE036
Window function in formula w i,j comprise b i,j with its eight neighborhood systems, f 1represent horizontal and vertical direction black site, f 2represent black site, angular direction and non-conterminous black site, f 3representing one is horizontal direction, and one is vertical direction adjacent black site in pairs.In above formula α, βwith γfor the ratio of dash area in figure and grid area, establish actual print point radius and ideally the ratio of radius be ρrepresent, for desirable site, parameter
Figure DEST_PATH_IMAGE038
=1, but can there is " some gain " phenomenon in output image gray shade scale, therefore consider this distortion, adopts
Figure 334554DEST_PATH_IMAGE038
=1.25,
Figure DEST_PATH_IMAGE040
.
In the time optimizing, the variance of the weight of unknown parameter and each predictive variable value is inversely proportional to:
Figure DEST_PATH_IMAGE042
.
The visual quality of shadow tone depends on the selection of initial point and the optimisation strategy of taking.A kind of simple iterative strategy is as follows, for arbitrary picture point ( i, j), a given initial estimation
Figure DEST_PATH_IMAGE044
, find binarized pixel
Figure DEST_PATH_IMAGE046
, make weighted quadratic value minimum.
Figure DEST_PATH_IMAGE048
In formula, be point ( i,j) eight neighborhoods, if vision wave filter and neighborhood system are selected enough greatly, minimize
Figure DEST_PATH_IMAGE052
just equal to minimize global error
Figure DEST_PATH_IMAGE054
.Attempt to produce best shadow tone duplicating image, minimize the square error between two-value half tone image and original gray scale image.A given secondary gray level image [ x i,j ], i=1 ...,
Figure DEST_PATH_IMAGE056
, j=1 ...,
Figure DEST_PATH_IMAGE058
. n w for piece image xthe number of the pixel in direction, n h for piece image ythe number of the pixel in direction, x i,j represent that pixel is positioned at grid irow and jrow, suppose a secondary grayscale image [ x i,j ], the gray shade scale of each pixel is from 0(white) to 1(black), suppose that a site can produce a pixel, therefore, grayscale image [ x i,j ] and bianry image [ b i,j ] there is an identical yardstick.First obtain a secondary initial half tone image
Figure DEST_PATH_IMAGE060
, for arbitrary picture point
Figure DEST_PATH_IMAGE062
can find binarized pixel
Figure DEST_PATH_IMAGE064
minimize the difference of two squares.
Figure DEST_PATH_IMAGE066
Figure DEST_PATH_IMAGE068
Figure DEST_PATH_IMAGE070
w i,j be by b i,j shown neighborhood system, * represents convolution.
Figure DEST_PATH_IMAGE074
with represent the different impulse response of vision wave filter of half tone image and continuous toned image.Boundary condition is that supposition does not have ink to be in outside image border.
The invention has the beneficial effects as follows:
1, solve existing method and cannot take into account image smoothing and sharpness problems, can realize half tone image and there is good marginal sharpness and the flatness in region, obtained high-quality half tone image.
2, by application weighted least squares, minimize combined error function, reach local optimum.
3, utilize the digital halftone method of multi-scale information fusion of the present invention, can obtain the laser plate-making image of high-quality.
Accompanying drawing explanation
Fig. 1 is that multi-scale information of the present invention merges halftone technique process flow diagram;
Fig. 2 is the hierarchical relationship figure of image of the present invention through level Four wavelet decomposition;
Fig. 3 is the quad-tree structure figure of wavelet coefficient between yardstick of the present invention;
Fig. 4 is the level Four wavelet transformation figure of image of the present invention;
Fig. 5 is the present invention's cluster segmentation image while working as K=2;
Fig. 6 is the present invention's cluster segmentation image while working as K=3;
Fig. 7 is the present invention's cluster segmentation image while working as K=4;
Fig. 8 is quality evaluation of halftone image parameter square error of the present invention (MSEv) comparison diagram;
Fig. 9 is quality evaluation of halftone image parameter Y-PSNR of the present invention (PSNR) comparison diagram.
Embodiment
Embodiment 1: as shown in Fig. 1-9, a kind of digital picture halftoning method based on wavelet multi-scale information fusion, the concrete steps of described method are as follows:
A, convert continuous toned image to gray level image and judge whether gray level image is 2 of standard n× 2 ngray level image;
B, employing two-dimensional discrete wavelet conversion obtain four yardstick information of the wavelet field of standard grayscale image, and the wavelet details Relationship of Coefficients between recycling yardstick between the autocorrelation fused layer of wavelet coefficient is set up marginal error measure function;
C, employing K-means clustering procedure are divided into standard grayscale image respectively kindividual region, recycles each Local Deviation inverse as weight, sets up domain error measure function;
D, the additive property of marginal error measure function and domain error measure function is set up to combined error measure function, then adopt direct two-value searching method to minimize the error of standard grayscale image and coarse half tone image, obtain optimum digital halftone image.
In described step B, the concrete steps that marginal error measure function is set up are as follows:
The Harr small echo that B1, employing have orthogonality carries out two-dimensional discrete wavelet conversion, realizes four yardstick wavelet transformations of gray level image and obtains four yardstick wavelet images;
B2, according to four yardstick wavelet images obtain level, vertical and diagonal angle subband wavelet coefficient distributes;
B3, independently suppose according to wavelet sub-band, level, vertical and three, diagonal angle direction wavelet coefficient are normalized, and the wavelet coefficient obtaining after normalized distributes;
B4, utilize the autocorrelation of wavelet coefficient between yardstick to merge the each yardstick wavelet coefficient after normalization, the wavelet coefficient after being merged;
B5, using merge after wavelet coefficient as weight, set up marginal error measure function
Figure 477566DEST_PATH_IMAGE002
; Wherein, lrepresent wavelet transformation progression; lH, hL, hHrepresent respectively multi-scale wavelet territory vertical direction, horizontal direction and to angular direction wavelet coefficient fuse information; i,jpresentation video pixel; b i,j represent pixel eight neighborhoods; z i,j represent the collimation error; w i,j represent model printer error.
In described step C, the concrete steps that domain error measure function is set up are as follows:
C1, employing K-means clustering procedure are divided into standard grayscale image kindividual area image, wherein k=2,3,4;
C2, the average of calculating each region and variance; Using the inverse of variance as weight, utilize weighted least squares to set up domain error measure function
Figure 478889DEST_PATH_IMAGE004
; Wherein, k=1,2,3,4 represent cut zone; λ k represent weight coefficient; i,jpresentation video pixel; b i,j represent pixel eight neighborhoods; z i, j represent the collimation error; w i, j represent model printer error.
In described step D, the concrete steps that direct two-value searching algorithm minimizes combined error measure function are as follows:
D1, the additive property of marginal error measure function and domain error measure function is set up to combined error measure function ;
D2, utilize in direct two-value searching algorithm conversion and the exchange of jumping for eight neighborhood regions of each pixel in standard grayscale image and coarse half tone image, make combined error measure function minimum;
D3, when entire image combined error measure function hour, algorithm stops, and obtains optimum half tone image.
Described coarse half tone image adopts error-diffusion method processing to obtain.
Embodiment 2: as shown in Fig. 1-9, a kind of digital picture halftoning method based on wavelet multi-scale information fusion, the concrete steps of described method are as follows:
A, convert continuous toned image to gray level image and judge whether gray level image is 2 of standard n× 2 ngray level image;
B, employing two-dimensional discrete wavelet conversion obtain four yardstick information of the wavelet field of standard grayscale image, and the wavelet details Relationship of Coefficients between recycling yardstick between the autocorrelation fused layer of wavelet coefficient is set up marginal error measure function:
B1, as shown in Figure 3, adopts the Harr small echo with orthogonality to carry out two-dimensional discrete wavelet conversion, realizes four yardstick wavelet transformations of gray level image and obtains four yardstick wavelet images.
B2, according to four yardstick wavelet images obtain level, vertical and diagonal angle subband wavelet coefficient distributes;
B3, independently suppose according to wavelet sub-band, level, vertical and three, diagonal angle direction wavelet coefficient are normalized, and the wavelet coefficient obtaining after normalized distributes;
B4, utilize the autocorrelation of wavelet coefficient between yardstick to merge the each yardstick wavelet coefficient after normalization, the wavelet coefficient after being merged;
B5, using merge after wavelet coefficient as weight, set up marginal error measure function
Figure 239035DEST_PATH_IMAGE002
; Wherein, lrepresent wavelet transformation progression; lH, hL, hHrepresent respectively multi-scale wavelet territory vertical direction, horizontal direction and to angular direction wavelet coefficient fuse information; i,jpresentation video pixel; b i,j represent pixel eight neighborhoods; z i,j represent the collimation error; w i,j represent model printer error;
C, employing K-means clustering procedure are divided into standard grayscale image respectively kindividual region, recycles each Local Deviation inverse as weight, sets up domain error measure function:
C1, employing K-means clustering procedure are divided into standard grayscale image kindividual area image, wherein k=2,3,4; (the continuous tune standard grayscale image shown in Fig. 4 is divided into kindividual area image, when k=2 o'clock, obtain two cluster areas images shown in Fig. 5, when k=3 o'clock, obtain three cluster areas images shown in Fig. 6, when k, obtain four cluster areas images shown in Fig. 7 at=4 o'clock).
C2, the average of calculating each region and variance; Using the inverse of variance as weight, utilize weighted least squares to set up domain error measure function
Figure 693018DEST_PATH_IMAGE004
; Wherein, k=1,2,3,4 represent cut zone; λ k represent that weight coefficient is as shown in table 1; i,jpresentation video pixel; b i,j represent pixel eight neighborhoods; z i, j represent the collimation error; w i, j represent model printer error;
Figure DEST_PATH_IMAGE080
D, the additive property of marginal error measure function and domain error measure function is set up to combined error measure function, then adopts direct two-value searching method to minimize the error of standard grayscale image and coarse half tone image, obtain optimum digital halftone image:
D1, the additive property of marginal error measure function and domain error measure function is set up to combined error measure function
Figure 829602DEST_PATH_IMAGE006
;
D2, utilize in direct two-value searching algorithm conversion and the exchange of jumping for eight neighborhood regions of each pixel in standard grayscale image and coarse half tone image, make combined error measure function minimum;
D3, when entire image combined error measure function hour, algorithm stops, and obtains optimum half tone image.
Whether obtained shadow tone gray level image is carried out to quality judging, be optimum to evaluate half tone image.Utilize quality assessment parameter square mean error amount (Means Squares Error value, and Y-PSNR (Peak Signal-to-Noise Ratio MSEv), PSNR), gained half tone image is carried out to quality assessment, square mean error amount (MSEv) represents half tone image visual matrix, for measuring the visual deformation between original-gray image and two-value half tone image.Y-PSNR (PSNR) is a ratio that represents signal maximum possible power and affect the destructive noise power of its expression precision.Evaluation result as shown in Figure 8 and Figure 9, as can be seen from the figure, has reduced by 0.0129 ~ 0.2102, PSNR and has increased by 0.53 ~ 3.17 along with cluster areas is increased to 4, MSEv value from 2.The visual deformation that half tone image is described reduces, and visual effect is better.The degree of closeness of shadow tone duplicating image and original image is higher.The noise that in halftoning process, image is introduced is less.
By reference to the accompanying drawings the specific embodiment of the present invention is explained in detail above, but the present invention is not limited to above-mentioned embodiment, in the ken possessing those of ordinary skills, can also under the prerequisite that does not depart from aim of the present invention, make various variations.

Claims (5)

1. the digital picture halftoning method based on wavelet multi-scale information fusion, is characterized in that: the concrete steps of described method are as follows:
A, convert continuous toned image to gray level image and judge whether gray level image is 2 of standard n× 2 ngray level image;
B, employing two-dimensional discrete wavelet conversion obtain four yardstick information of the wavelet field of standard grayscale image, and the wavelet details Relationship of Coefficients between recycling yardstick between the autocorrelation fused layer of wavelet coefficient is set up marginal error measure function;
C, employing K-means clustering procedure are divided into standard grayscale image respectively kindividual region, recycles each Local Deviation inverse as weight, sets up domain error measure function;
D, the additive property of marginal error measure function and domain error measure function is set up to combined error measure function, then adopt direct two-value searching method to minimize the error of standard grayscale image and coarse half tone image, obtain optimum digital halftone image.
2. the digital picture halftoning method based on wavelet multi-scale information fusion according to claim 1, is characterized in that: in described step B, the concrete steps that marginal error measure function is set up are as follows:
The Harr small echo that B1, employing have orthogonality carries out two-dimensional discrete wavelet conversion, realizes four yardstick wavelet transformations of gray level image and obtains four yardstick wavelet images;
B2, according to four yardstick wavelet images obtain level, vertical and diagonal angle subband wavelet coefficient distributes;
B3, independently suppose according to wavelet sub-band, level, vertical and three, diagonal angle direction wavelet coefficient are normalized, and the wavelet coefficient obtaining after normalized distributes;
B4, utilize the autocorrelation of wavelet coefficient between yardstick to merge the each yardstick wavelet coefficient after normalization, the wavelet coefficient after being merged;
B5, using merge after wavelet coefficient as weight, set up marginal error measure function
Figure 2014100165610100001DEST_PATH_IMAGE001
; Wherein, lrepresent wavelet transformation progression; lH, hL, hHrepresent respectively multi-scale wavelet territory vertical direction, horizontal direction and to angular direction wavelet coefficient fuse information; i,jpresentation video pixel; b i,j represent pixel eight neighborhoods; z i,j represent the collimation error; w i,j represent model printer error.
3. the digital picture halftoning method based on wavelet multi-scale information fusion according to claim 1, is characterized in that: in described step C, the concrete steps that domain error measure function is set up are as follows:
C1, employing K-means clustering procedure are divided into standard grayscale image kindividual area image, wherein k=2,3,4;
C2, the average of calculating each region and variance; Using the inverse of variance as weight, utilize weighted least squares to set up domain error measure function
Figure 673565DEST_PATH_IMAGE002
; Wherein, k=1,2,3,4 represent cut zone; λ k represent weight coefficient; i,jpresentation video pixel; b i,j represent pixel eight neighborhoods; z i, j represent the collimation error; w i, j represent model printer error.
4. the digital picture halftoning method based on wavelet multi-scale information fusion according to claim 1, is characterized in that: in described step D, the concrete steps that direct two-value searching algorithm minimizes combined error measure function are as follows:
D1, the additive property of marginal error measure function and domain error measure function is set up to combined error measure function
Figure DEST_PATH_IMAGE003
;
D2, utilize in direct two-value searching algorithm conversion and the exchange of jumping for eight neighborhood regions of each pixel in standard grayscale image and coarse half tone image, make combined error measure function minimum;
D3, when entire image combined error measure function hour, algorithm stops, and obtains optimum half tone image.
5. according to the digital picture halftoning method based on wavelet multi-scale information fusion described in claim 1 or 4, it is characterized in that: described coarse half tone image adopts error-diffusion method processing to obtain.
CN201410016561.0A 2014-01-15 2014-01-15 A kind of digital picture halftoning method based on wavelet multi-scale information fusion Expired - Fee Related CN103793897B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410016561.0A CN103793897B (en) 2014-01-15 2014-01-15 A kind of digital picture halftoning method based on wavelet multi-scale information fusion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410016561.0A CN103793897B (en) 2014-01-15 2014-01-15 A kind of digital picture halftoning method based on wavelet multi-scale information fusion

Publications (2)

Publication Number Publication Date
CN103793897A true CN103793897A (en) 2014-05-14
CN103793897B CN103793897B (en) 2016-08-24

Family

ID=50669522

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410016561.0A Expired - Fee Related CN103793897B (en) 2014-01-15 2014-01-15 A kind of digital picture halftoning method based on wavelet multi-scale information fusion

Country Status (1)

Country Link
CN (1) CN103793897B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107133936A (en) * 2017-05-09 2017-09-05 上海理工大学 Digital halftoning method
CN108921886A (en) * 2018-06-11 2018-11-30 昆明理工大学 A kind of texture information fusion Multi-scale model forest digital picture halftoning method
CN110880004A (en) * 2019-11-21 2020-03-13 北京三缘聚科技有限公司 Digital image mode class feature extraction network and method
CN112330757A (en) * 2019-08-05 2021-02-05 复旦大学 Complementary color wavelet measurement for evaluating color image automatic focusing definition

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1863267A (en) * 2005-05-10 2006-11-15 致伸科技股份有限公司 Digital halftoning technique based on 1-d multi-scale dot assignment
US7280252B1 (en) * 2001-12-19 2007-10-09 Ricoh Co., Ltd. Error diffusion of multiresolutional representations
CN101600039A (en) * 2008-06-05 2009-12-09 佳世达科技股份有限公司 The method of half tone image conversion method, Method of printing and generation halftone shield

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7280252B1 (en) * 2001-12-19 2007-10-09 Ricoh Co., Ltd. Error diffusion of multiresolutional representations
CN1863267A (en) * 2005-05-10 2006-11-15 致伸科技股份有限公司 Digital halftoning technique based on 1-d multi-scale dot assignment
CN101600039A (en) * 2008-06-05 2009-12-09 佳世达科技股份有限公司 The method of half tone image conversion method, Method of printing and generation halftone shield

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107133936A (en) * 2017-05-09 2017-09-05 上海理工大学 Digital halftoning method
CN107133936B (en) * 2017-05-09 2019-11-01 上海理工大学 Digital halftoning method
CN108921886A (en) * 2018-06-11 2018-11-30 昆明理工大学 A kind of texture information fusion Multi-scale model forest digital picture halftoning method
CN108921886B (en) * 2018-06-11 2021-09-14 昆明理工大学 Texture information fusion multi-scale structured forest digital image halftone method
CN112330757A (en) * 2019-08-05 2021-02-05 复旦大学 Complementary color wavelet measurement for evaluating color image automatic focusing definition
CN112330757B (en) * 2019-08-05 2022-11-29 复旦大学 Complementary color wavelet measurement for evaluating color image automatic focusing definition
CN110880004A (en) * 2019-11-21 2020-03-13 北京三缘聚科技有限公司 Digital image mode class feature extraction network and method

Also Published As

Publication number Publication date
CN103793897B (en) 2016-08-24

Similar Documents

Publication Publication Date Title
CN102063713B (en) Neighborhood normalized gradient and neighborhood standard deviation-based multi-focus image fusion method
CN101968883A (en) Method for fusing multi-focus images based on wavelet transform and neighborhood characteristics
CN103793897A (en) Digital image halftone method based on small-wavelet-domain multi-scale information fusion
CN1352855A (en) Narrow band, anisotropic stochastic halftone patterns and methods for creating and using the same
JP2003046777A (en) Mask preparation method, image processor, software program and mask data
WO2006041812A9 (en) Method of producing improved lenticular images
WO2006114031A1 (en) Frequency modulated screening method based on the dual feedback error diffusion
Goyal et al. Clustered-dot halftoning with direct binary search
EP1596572B1 (en) Reduction of artifacts in halftone screens
US4758886A (en) Optimal color half-tone patterns for raster-scan images
US20160255240A1 (en) Halftoning
Son Inverse halftoning through structure-aware deep convolutional neural networks
JPH0785272A (en) Frequency-modulated halftone image and formation method
JPH11234513A (en) Pseudo gradation image processing unit
Wen et al. A novel classification method of halftone image via statistics matrices
CN106515241B (en) A kind of fluorescence coloured picture screening method and device
Abedini et al. 3D halftoning based on iterative method controlling dot placement
CN113989436A (en) Three-dimensional mesh tone reconstruction method based on HVS and random printer model
CN102158630A (en) Image-tone-based adaptive green-noise screening method
US10587774B2 (en) 3D printed object halftone image generation containing updated voxel data
CN103955908B (en) Digital image halftone method based on multi-scale perception error measure approximation global optimization
Son Inverse halftoning through structure-aware deep convolutional neural networks
JP3917634B2 (en) How to create a mesh screen
KR101028697B1 (en) Apparatus and method for luminace independent texure synthesis
RU2308167C2 (en) Method for adaptive screening of halftone original and device for realization of the method

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20160824

Termination date: 20220115

CF01 Termination of patent right due to non-payment of annual fee