CN103618845B - A kind of based on minimum colour developing error laser printer model green noise halftone algorithm - Google Patents

Abstract

The invention discloses a kind of based on minimum colour developing error laser printer model green noise halftone algorithm, belong to the halftone screen dot generation method that image hard copy copies field.The non-linear distortion due to printer with dot gains and some loss exists, and needs to use model printer to simulate non-linear distortion.And current model printer uses density or reflectivity to convert measurement result usually, do not meet the linear change of human eye vision gray scale, so usually need repeatedly to print the parameter measured and could determine model printer due to density and reflectivity.Printout results conversion is Munsell value value by the present invention, calculates the optimized parameter of model printer, obtains the laser printer model of minimum colour developing error, and be applied in green noise halftone algorithm.Only need once to print measurement, the model printer that the optimized parameter generated is corresponding just inhibits the non-linear distortion of printer preferably.

Description

A kind of based on minimum colour developing error laser printer model green noise halftone algorithm

Technical field

The invention belongs to digital image processing field, relate to a kind of halftoning method, be specifically related to a kind of based on minimum colour developing error laser printer model green noise halftone algorithm.

Background technology

Because ink powder physical diffusion and electrostatic charge adsorption capacity are not enough, printer there will be dot gains and some Loss, and these two kinds of phenomenons are otherwise known as printer non-linear distortion.Printer non-linear distortion will cause the gray scale of output image and the situations such as level, relatively relationship between light and dark change.In order to suppress the non-linear distortion of printer, domestic and international researcher passes through feature and the printer printing mechanism of printer emulation print point, sets up model printer.

Current model printer, is normally converted to density value or reflectivity by measured value, determines model printer by the relation of the gray value comparing the density value after conversion, reflectivity and original image.But do not meet human eye vision grey scale change situation due to density value and reflectivity, so directly can not be obtained the optimized parameter of model printer by the method for minimum variance, and need repeatedly to print measurement, judge optimized parameter by human eye.

Summary of the invention

In order to solve the problems of the technologies described above, the present invention proposes a kind of based on minimum colour developing error laser printer model green noise halftone algorithm, be Munsell value value by printout results conversion, calculate the optimized parameter of model printer, obtain the laser printer model of minimum colour developing error, and be applied in green noise halftone algorithm, only need once to print measurement, the model printer that the optimized parameter generated is corresponding just inhibits the non-linear distortion of printer preferably.

The technical solution adopted in the present invention is: a kind of based on minimum colour developing error laser printer model green noise halftone algorithm, it is characterized in that, comprises the following steps:

Step 1: by calculating the minimum colour developing error of test pattern, obtain corresponding optimized parameter t ₁, T ₂combination; Its specific implementation comprises following sub-step:

Step 1.1: draw and comprise by the gray level image of black 20 color lumps to the gray-value variation such as white as test pattern I _s;

Step 1.2: select green noise halftone algorithm to test pattern I _scarry out halftone process, obtain halftone process result images I _h;

Step 1.3: by halftone process result images I _huse target laser printer to print out, obtain Output rusults image I _p;

Step 1.4: use colorimeter or densimeter measurement Output rusults image I _p, measurement data is converted to gray value, obtains 20 color lumps and measure gray value array arrPG; Wherein said is converted to gray value by measurement data, and its specific implementation comprises following sub-step:

Step 1.4.1: measurement data is converted to the luminance factor Y in CIE1931XYZ colorimeter system; Wherein:

If use colorimeter to measure printing Output rusults, then measure the luminance factor Y obtaining printing out result;

If use densitometer to measure printing Output rusults, then the density D derived according to optical reflection density defined formula and the functional relation of luminance factor Y, calculate luminance factor Y:

$D=\lg \[Y_0/\left({\mathrm{Y\ρ}}_0\right)\]\⇒Y=\frac{Y_0}{{10}^D{\mathrm{\ρ}}_0}$ (formula one)

Wherein, Y ₀for the luminance factor of Munsell system the 10th grade of lightness, Y ₀=102.75, ρ ₀for the reflectivity of Munsell system the 10th grade of lightness, ρ ₀=1, D is the optical reflection density in certain region, and Y is the luminance factor in this region;

Step 1.4.2: luminance factor Y is converted to Munsell value V, according to the functional relation of luminance factor Y and Munsell value V, calculates Munsell value V:

V=2.217Y ^0.352-1.324 (formulas two)

Step 1.4.3: Munsell value V is converted to gray value G,

$G=\frac{V-V_p}{V_s-V_p}$ (formula three)

Wherein, V is to be converted is Munsell value, V _pfor the white corresponding Munsell value of paper, V _sit is the Munsell value that 100% site printing effect is corresponding;

Step 1.5: for halftone process result images I _h, in [0,1] scope with 0.1 for step-length, constantly change formula wantonly in parameter single pixel segmentation is become 10 × 10 points, at the light energy distribution value P (x of each point when calculating single pixel exposure _i, y _i):

$P(x_i,y_i)={\exp }^{-\∂\[{(x_i-x_c)}^2+{(y_i-y_c)}^2\]/l^2}$ (formula wantonly)

Wherein, (x _c, y _c) be the exposure center of pixel (m, n), (x _i, y _i) be this pixel exposure each point pericentral, l is the centre-to-centre spacing of laser printer two neighbouring device print points, l=10;

Step 1.6: the light energy of each point of statistics exposing pixels and eight neighborhood thereof, the average light energy value A of the region of calculation exposure pixel own, adjacent area, diagonal zones _center, A _nerb, A _diag, its universal calculation equation is:

$A=\frac{{\∫}_{y_i-l/2}^{y_i+l/2}{\∫}_{x_i-l/2}^{x_i+l/2}P(x,y)dxdy}{l\×l}$ (formula 5)

Wherein, (x _i-l/2, x _i+ l/2), (y _i-l/2, y _i+ l/2) be respectively the transverse and longitudinal coordinate span in unit are region;

Step 1.7: pointwise judges halftone process result images I _hin the eight neighborhood of each pixel whether be print point, calculate the superimposed light energy L on each pixel region:

L=uA _center+ vA _nerb+ wA _diag(formula land)

Wherein, whether u is exposed decision by current pixel, A _centerfor current pixel is at the average light energy value in pixel region own, v is the number of pixels will be exposed in the adjacent domain of current pixel, A _nerbfor the average light energy value of print pixel point in current pixel region of proximal direction, w is the number of pixels will be exposed in the diagonal zones of current pixel, A _diagfor the average light energy value of print point in current pixel region at diagonal angle;

Step 1.8: in [0,3] scope with 0.1 for step-length, change parameter T in formula seven successively ₁, T ₂, calculate the modulating output gray value p of each pixel in substitution formula seven:

$p=\left\{\begin{array}{cc}0& L<T_1\\ \frac{L-T_1}{T_2-T_1}& T_1\&le;L<T_2\\ 1& T_2\&le;L\end{array}\right.$ (formula seven)

Wherein, if in conversion threshold process, T ₁>T ₂, then this time conversion is skipped, and directly enters and converts next time;

Step 1.9: the mean value calculating modulating output gray value p corresponding to all pixels in same gray scale color lump, obtains the analog gray scale value array arrIG of 20 gray scale color lumps;

Step 1.10: the root-mean-square error RMSE between computation and measurement gray value array arrPG and analog gray scale value array arrIG:

$RMSE=\sqrt{\frac{\underset{i=1}{\overset{n}{\&Sigma;}}{({\mathrm{arrPG}}_i-{\mathrm{arrIG}}_i)}^2}{n}}$ (formula eight)

Wherein, n is the number of gray scale color lump, n=20;

Step 1.11: judge, does is RMSE minimum value?

If not, then the step 1.5 described in revolution execution;

If: then order performs following step 1.12;

Step 1.12: select minimum colour developing error RMSE _mincorresponding parameter t ₁, T ₂as optimized parameter;

Step 2: by optimized parameter t ₁, T ₂be loaded in green noise halftone algorithm, halftone process is carried out to image; Its specific implementation comprises following sub-step:

Step 2.1: substitute into optimized parameter and calculate the light energy value A of single print point in pixel region own, adjacent domain and diagonal zones respectively according to formula four and formula 5 _center, A _nerband A _diag, and preserve;

Step 2.2: set pending image as a width size be the gray level image of M × N, x _m,nrepresent the gray value at pixel (m, the n) place of pending image, wherein x _m,n∈ [0,1], 0 is black, and 1 is white; The error amount of the around processed pixel of eight neighborhood and error diffusion filter convolution are obtained error amount, and the gray value of input pixel of current point and the difference of error amount are as inputting threshold quantizer gray value u _m,n:

$u_{m,n}=x_{m,n}-\underset{k,j\&Element;S}{\&Sigma;}{F}_{k,j}{e}_{m-k,n-j}$ (formula nine)

Wherein, S is the coefficient template scope of error diffusion filter F (), and k, j are the ranks value of the coefficient template of F ();

Step 2.3: adopt the analog pixel value of sluggish filters H () to the adjacent processed pixel in pixel (m, n) place to carry out weights computing, then be multiplied by hysteresis index h as feedback, adds input threshold quantizer gray value u _m,nin, then the output valve b of pixel (m, n) _m,n:

$b_{m,n}=Q\left(u_{m,n}\right)=\left\{\begin{array}{cc}1& (u_{m,n}-h\underset{p,q\&Element;R}{\mathrm{\&Sigma;}}{H}_{p,q}{b}_{m-p,n-q}^{\&prime;})>0.5\\ 0& else\end{array}\right.$ (formula is picked up)

Wherein, R is the coefficient template scope of sluggish filters H (), and output valve 1 is print point, and 0 is not print point; P, q are the ranks value of the coefficient template of H ();

Step 2.4: by light energy value A _center, A _nerband A _diagsubstitution formula land, the light energy superposition value on calculating pixel point (m, n) region, the value of u, v, w in its Chinese style land is determined by the output valve of point processed in pixel (m, n) eight neighborhood.

$u=\left\{\begin{array}{cc}1& b_{m,n}=0\\ 0& b_{m,n}=1\end{array}\right.$ (formula 11)

(formula 12)

(formula 13)

Step 2.5: by optimized parameter T ₁, T ₂the modulating output gray value of current point is calculated in substitution formula seven;

Step 2.6: the error amount e calculating current point _m,nfor analog pixel value and the difference e of input threshold quantizer gray value _m,n:

E _m,n=b ' _m,n-u _m,n(formula 14)

Step 2.7: described in step 2.1 to step 2.6, from angle point beginning, pending image upper left reason, by order left-to-right, from top to bottom, until terminate to pending image bottom right angle point.

Innovative point of the present invention is:

(1) measurement data is converted to Munsell value value, Munsell value is converted to analog gray scale;

(2) by loop modification parameter t ₁, T ₂, ask minimal error, using the parameter of correspondence as optimized parameter;

(3) optimized parameter is simulated the gray value obtained and replace the original gray value error of calculation;

Printout results conversion is Munsell value value by the present invention, calculate the optimized parameter of model printer, obtain the laser printer model of minimum colour developing error, and be applied in green noise halftone algorithm, only need once to print measurement, the model printer that the optimized parameter generated is corresponding just inhibits the non-linear distortion of printer preferably.

Accompanying drawing explanation

Fig. 1: be the flow chart based on minimum colour developing error optimization calculation of parameter of the present invention.

Fig. 2: be the flow chart based on minimum colour developing error laser printer model green noise halftone algorithm of the present invention

Fig. 3: be the luminance factor Y of the embodiment of the present invention and the transformational relation schematic diagram of Munsell value V.

Fig. 4: 20 the pending gray scale color lumps being the embodiment of the present invention.

Fig. 5-1: the coefficient template being the error diffusion filter F () adopted in the embodiment of the present invention.

Fig. 5-2: the coefficient template being the sluggish filters H () adopted in the embodiment of the present invention.

Fig. 6: the lena test pattern being the embodiment of the present invention.

Fig. 7: be the embodiment of the present invention by common green noise halftone algorithm process and print scanned lena resolution chart.

Fig. 8: be the embodiment of the present invention by based on minimum colour developing error laser printer model green noise halftone algorithm process and print scanned lena resolution chart.

Embodiment

Below with Pantum2000 laser printer be target printer, to hat board height white manifold paper carry out printing out, measuring equipment is for Eyeone colorimetric instrument, be pending image for lena resolution chart, with common green noise halftone algorithm for contrast, use the high score resolution scanner of 1200dpi to scan halftoning printout image, with this, the present invention is further described.

Ask for an interview Fig. 1, Fig. 2, the technical solution adopted in the present invention is: a kind of based on minimum colour developing error laser printer model green noise halftone algorithm, it is characterized in that, comprises the following steps:

Step 1: by calculating the minimum colour developing error of test pattern, obtain corresponding optimized parameter t ₁, T ₂combination; Its specific implementation comprises following sub-step:

Step 1.1: ask for an interview Fig. 4, draws 20 the gray scale color lumps of 600 × 600 pixels by 0 to 0.95 every 0.05 grey scale change as test pattern I _s, wherein 0 represents black, and 1 represents white;

Step 1.2: select green noise halftone algorithm to test pattern I _scarry out halftone process, obtain halftone process result images I _h;

Step 1.3: by halftone process result images I _huse Pantum2000 laser printer to print out, obtain Output rusults image I _p;

Step 1.4: use Eyeone colorimetric instrument to measure Output rusults image I _p, and by the following method measurement data is converted to gray value, obtain 20 color lumps and measure gray value array arrPG: wherein said is converted to gray value by measurement data, and its specific implementation comprises following sub-step:

Step 1.4.1: measurement data is converted to the luminance factor Y in CIE1931XYZ colorimeter system; Its specific implementation process is: the density D derived according to optical reflection density defined formula and the functional relation of luminance factor Y, calculate luminance factor Y:

$D=\lg \&lsqb;Y_0/\left({\mathrm{Y\&rho;}}_0\right)\&rsqb;\&DoubleRightArrow;Y=\frac{Y_0}{{10}^D{\mathrm{\&rho;}}_0}$ (formula one)

Wherein, Y ₀for the luminance factor of Munsell system the 10th grade of lightness, Y ₀=102.75, ρ ₀for the reflectivity of Munsell system the 10th grade of lightness, ρ ₀=1, D is the optical reflection density in certain region, and Y is the luminance factor in this region;

Step 1.4.2: luminance factor Y is converted to Munsell value V, asks for an interview Fig. 3, according to the functional relation of luminance factor Y and Munsell value V, calculates Munsell value V:

V=2.217Y ^0.352-1.324 (formulas two)

Step 1.4.3: Munsell value V is converted to gray value G,

$G=\frac{V-V_p}{V_s-V_p}$ (formula three)

Wherein, V is to be converted is Munsell value, V _pfor the white corresponding Munsell value of paper, V _sit is the Munsell value that 100% site printing effect is corresponding;

Step 1.5: for halftone process result images I _h, in [0,1] scope with 0.1 for step-length, constantly change formula wantonly in parameter single pixel segmentation is become 10 × 10 points, at the light energy distribution value P (x of each point when calculating single pixel exposure _i, y _i):

$P(x_i,y_i)={\exp }^{-\&part;\&lsqb;{(x_i-x_c)}^2+{(y_i-y_c)}^2\&rsqb;/l^2}$ (formula wantonly)

Wherein, (x _c, y _c) be the exposure center of pixel (m, n), (x _i, y _i) be this pixel exposure each point pericentral, l is the centre-to-centre spacing of laser printer two neighbouring device print points, l=10;

Step 1.6: the light energy of each point of statistics exposing pixels and eight neighborhood thereof, the average light energy value A of the region of calculation exposure pixel own, adjacent area, diagonal zones _center, A _nerb, A _diag, its universal calculation equation is:

$A=\frac{{\&Integral;}_{y_i-l/2}^{y_i+l/2}{\&Integral;}_{x_i-l/2}^{x_i+l/2}P(x,y)dxdy}{l\&times;l}$ (formula 5)

Wherein, (x _i-l/2, x _i+ l/2), (y _i-l/2, y _i+ l/2) be respectively the transverse and longitudinal coordinate span in unit are region;

Step 1.7: pointwise judges halftone process result images I _hin the eight neighborhood of each pixel whether be print point, calculate the superimposed light energy L on each pixel region:

L=uA _center+ vA _nerb+ wA _diag(formula land)

Wherein, whether u is exposed decision by current pixel, A _centerfor current pixel is at the average light energy value in pixel region own, v is the number of pixels will be exposed in the adjacent domain of current pixel, A _nerbfor the average light energy value of print pixel point in current pixel region of proximal direction, w is the number of pixels will be exposed in the diagonal zones of current pixel, A _diagfor the average light energy value of print point in current pixel region at diagonal angle;

Step 1.8: in [0,3] scope with 0.1 for step-length, change parameter T in formula seven successively ₁, T ₂, in substitution formula seven, calculate the modulating output gray value p of each pixel

$p=\left\{\begin{array}{cc}0& L<T_1\\ \frac{L-T_1}{T_2-T_1}& T_1\&le;L<T_2\\ 1& T_2\&le;L\end{array}\right.$ (formula seven)

Wherein, if in conversion threshold process, T ₁>T ₂, then this time conversion is skipped, and directly enters and converts next time;

Step 1.9: the mean value calculating modulating output gray value p corresponding to all pixels in same gray scale color lump, obtains the analog gray scale value array arrIG of 20 gray scale color lumps;

Step 1.10: the root-mean-square error RMSE between computation and measurement gray value array arrPG and analog gray scale value array arrIG:

$RMSE=\sqrt{\frac{\underset{i=1}{\overset{n}{\&Sigma;}}{({\mathrm{arrPG}}_i-{\mathrm{arrIG}}_i)}^2}{n}}$ (formula eight)

Wherein, n is the number of gray scale color lump, n=20;

Step 1.11: judge, does is RMSE minimum value?

If not, then the step 1.5 described in revolution execution;

If: then order performs following step 1.12;

Step 1.12: select minimum colour developing error RMSE _mincorresponding parameter t ₁, T ₂as optimized parameter; The present embodiment selects minimum colour developing error RMSE _minwhen=3.28, corresponding parameter t ₁=0.4, T ₂=2.3 as optimized parameter;

Step 2: by optimized parameter t ₁, T ₂be loaded in green noise halftone algorithm, halftone process is carried out to image; Its specific implementation comprises following sub-step:

Step 2.1: substitute into optimized parameter and calculate the light energy value A of single print point in pixel region own, adjacent domain and diagonal zones respectively according to formula four and formula 5 _center, A _nerband A _diag, and preserve;

Step 2.2: set pending image as a width size be the gray level image of M × N, x _m,nrepresent the gray value at pixel (m, the n) place of pending image, wherein x _m,n∈ [0,1], 0 is black, and 1 is white; The error amount of the around processed pixel of eight neighborhood and error diffusion filter convolution are obtained error amount, and the gray value of input pixel of current point and the difference of error amount are as inputting threshold quantizer gray value u _m,n:

$u_{m,n}=x_{m,n}-\underset{k,j\&Element;S}{\&Sigma;}{F}_{k,j}{e}_{m-k,n-j}$ (formula nine)

Wherein, S is the coefficient template scope of error diffusion filter F (), and k, j are the ranks value of the coefficient template of F (), ask for an interview Fig. 5-1;

Step 2.3: adopt the analog pixel value of sluggish filters H () to the adjacent processed pixel in pixel (m, n) place to carry out weights computing, then be multiplied by hysteresis index h as feedback, adds input threshold quantizer gray value u _m,nin, then the output valve b of pixel (m, n) _m,n:

$b_{m,n}=Q\left(u_{m,n}\right)=\left\{\begin{array}{cc}1& (u_{m,n}-h\underset{p,q\&Element;R}{\mathrm{\&Sigma;}}{H}_{p,q}{b}_{m-p,n-q}^{\&prime;})>0.5\\ 0& else\end{array}\right.$ (formula is picked up)

Wherein, R is the coefficient template scope of sluggish filters H (), and output valve 1 is print point, and 0 is not print point; P, q are the ranks value of the coefficient template of H (), ask for an interview Fig. 5-2;

Step 2.4: by light energy value A _center, A _nerband A _diagsubstitution formula land, the light energy superposition value on calculating pixel point (m, n) region, the value of u, v, w in its Chinese style land is determined by the output valve of point processed in pixel (m, n) eight neighborhood.

$u=\left\{\begin{array}{cc}1& b_{m,n}=0\\ 0& b_{m,n}=1\end{array}\right.$ (formula 11)

(formula 12)

(formula 13)

Step 2.5: by optimized parameter T ₁=0.4, T ₂the modulating output gray value of current point is calculated in=2.3 substitution formulas seven;

Step 2.6: the error amount e calculating current point _m,nfor analog pixel value and the difference e of input threshold quantizer gray value _m,n:

E _m,n=b ' _m,n-u _m,n(formula 14)

Step 2.7: described in step 2.1 to step 2.6, from angle point beginning, pending image upper left reason, by order left-to-right, from top to bottom, until terminate to pending image bottom right angle point.

In the present embodiment, measurement data is converted to the luminance factor Y in CIE1931XYZ colorimeter system; Also colorimeter can be used to measure printing Output rusults, and measure the luminance factor Y obtaining printing out result.

Asking for an interview Fig. 6, is the lena test pattern of the embodiment of the present invention; Ask for an interview Fig. 7, be the embodiment of the present invention by common green noise halftone algorithm process and print scanned lena resolution chart; Its overall light and shade is obviously partially dark, and the contrast of the flower on its hair and cap completely and level, cannot resolve minutiae profile; Ask for an interview Fig. 8, be the embodiment of the present invention by based on minimum colour developing error laser printer model green noise halftone algorithm process and print scanned lena resolution chart, its overall light and shade reduction comparatively meets former figure, still can reproduce preferably in shadow part details, restrained effectively printer non-linear distortion.

These are only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention, therefore, all any amendments done within the spirit and principles in the present invention, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (1)

1., based on a minimum colour developing error laser printer model green noise halftone algorithm, it is characterized in that, comprise the following steps:

Step 1: by calculating the minimum colour developing error of test pattern, obtain corresponding optimized parameter , T ₁, T ₂combination; Its specific implementation comprises following sub-step:

Step 1.1: draw and comprise by the gray level image of black 20 color lumps to the gray-value variation such as white as test pattern I _s;

Step 1.2: select green noise halftone algorithm to test pattern I _scarry out halftone process, obtain halftone process result images I _h;

Step 1.3: by halftone process result images I _huse target laser printer to print out, obtain Output rusults image I _p;

Step 1.4: use colorimeter or densimeter measurement Output rusults image I _p, measurement data is converted to gray value, obtains 20 color lumps and measure gray value array arrPG; Wherein said is converted to gray value by measurement data, and its specific implementation comprises following sub-step:

Step 1.4.1: measurement data is converted to the luminance factor Y in CIE1931XYZ colorimeter system; Wherein:

If use colorimeter to measure printing Output rusults, then measure the luminance factor Y obtaining printing out result;

If use densitometer to measure printing Output rusults, then the density D derived according to optical reflection density defined formula and the functional relation of luminance factor Y, calculate luminance factor Y:

$D=\lg \&lsqb;Y_0/\left({\mathrm{Y\&rho;}}_0\right)\&rsqb;\&DoubleRightArrow;Y=\frac{Y_0}{{10}^D{\mathrm{\&rho;}}_0}$ (formula one)

Wherein, Y ₀for the luminance factor of Munsell system the 10th grade of lightness, Y ₀=102.75, ρ ₀for the reflectivity of Munsell system the 10th grade of lightness, ρ ₀=1, D is the optical reflection density in certain region, and Y is the luminance factor in this region;

Step 1.4.2: luminance factor Y is converted to Munsell value V, according to the functional relation of luminance factor Y and Munsell value V, calculates Munsell value V:

V=2.217Y ^0.352-1.324 (formulas two)

Step 1.4.3: Munsell value V is converted to gray value G,

$G=\frac{V-V_p}{V_s-V_p}$ (formula three)

Wherein, V is to be converted is Munsell value, V _pfor the white corresponding Munsell value of paper, V _sit is the Munsell value that 100% site printing effect is corresponding;

Step 1.5: for halftone process result images I _h, in [0,1] scope with 0.1 for step-length, constantly change formula wantonly in parameter , single pixel segmentation is become 10 × 10 points, at the light energy distribution value P (x of each point when calculating single pixel exposure _i, y _i):

$P(x_i,y_i)={\exp }^{-\&part;\&lsqb;{(x_i-x_c)}^2+{(y_i-y_c)}^2\&rsqb;/l^2}$ (formula wantonly)

Wherein, (x _c, y _c) be the exposure center of pixel (m, n), (x _i, y _i) be this pixel exposure each point pericentral, l is the centre-to-centre spacing of laser printer two neighbouring device print points, l=10;

Step 1.6: the light energy of each point of statistics exposing pixels and eight neighborhood thereof, the average light energy value A of the region of calculation exposure pixel own, adjacent area, diagonal zones _center, A _nerb, A _diag, its universal calculation equation is:

$A=\frac{{\&Integral;}_{y_i-l/2}^{y_i+l/2}{\&Integral;}_{x_i-l/2}^{x_i+l/2}P(x,y)dxdy}{l\&times;l}$ (formula 5)

Wherein, (x _i-l/2, x _i+ l/2), (y _i-l/2, y _i+ l/2) be respectively the transverse and longitudinal coordinate span in unit are region;

Step 1.7: pointwise judges halftone process result images I _hin the eight neighborhood of each pixel whether be print point, calculate the superimposed light energy L on each pixel region:

L=uA _center+ vA _nerb+ wA _diag(formula land)

Wherein, whether u is exposed decision by current pixel, A _centerfor current pixel is at the average light energy value in pixel region own, v is the number of pixels will be exposed in the adjacent domain of current pixel, A _nerbfor the average light energy value of print pixel point in current pixel region of proximal direction, w is the number of pixels will be exposed in the diagonal zones of current pixel, A _diagfor the average light energy value of print point in current pixel region at diagonal angle;

Step 1.8: in [0,3] scope with 0.1 for step-length, change parameter T in formula seven successively ₁, T ₂, and substitute into the modulating output gray value p calculating each pixel in formula seven

$p=\left\{\begin{array}{cc}0& L<T_1\\ \frac{L-T_1}{T_2-T_1}& T_1\&le;L<T_2\\ 1& T_2\&le;L\end{array}\right.$ (formula seven)

Wherein, if in conversion threshold process, T ₁>T ₂, then this time conversion is skipped, and directly enters and converts next time;

Step 1.9: the mean value calculating modulating output gray value p corresponding to all pixels in same gray scale color lump, obtains the analog gray scale value array arrIG of 20 gray scale color lumps;

Step 1.10: the root-mean-square error RMSE between computation and measurement gray value array arrPG and analog gray scale value array arrIG:

$RMSE=\sqrt{\frac{\underset{i=1}{\overset{n}{\&Sigma;}}{({\mathrm{arrPG}}_i-{\mathrm{arrIG}}_i)}^2}{n}}$ (formula eight)

Wherein, n is the number of gray scale color lump, n=20;

Step 1.11: judge, does is RMSE minimum value?

If not, then the step 1.5 described in revolution execution;

If: then order performs following step 1.12;

Step 1.12: select minimum colour developing error RMSE _mincorresponding parameter t ₁, T ₂as optimized parameter;

Step 2: by optimized parameter t ₁, T ₂be loaded in green noise halftone algorithm, halftone process is carried out to image; Its specific implementation comprises following sub-step:

Step 2.1: substitute into optimized parameter and calculate the light energy value A of single print point in pixel region own, adjacent domain and diagonal zones respectively according to formula four and formula 5 _center, A _nerband A _diag, and preserve;

Step 2.2: set pending image as a width size be the gray level image of M × N, x _m,nrepresent the gray value at pixel (m, the n) place of pending image, wherein x _m,n∈ [0,1], 0 is black, and 1 is white; The error amount of the around processed pixel of eight neighborhood and error diffusion filter convolution are obtained error amount, and the gray value of input pixel of current point and the difference of error amount are as inputting threshold quantizer gray value u _m,n:

$u_{m,n}=x_{m,n}-\underset{k,j\&Element;S}{\&Sigma;}{F}_{k,j}{e}_{m-k,n-j}$ (formula nine)

Wherein, S is the coefficient template scope of error diffusion filter F (), and k, j are the ranks value of the coefficient template of F ();

Step 2.3: adopt the analog pixel value of sluggish filters H () to the adjacent processed pixel in pixel (m, n) place to carry out weights computing, then be multiplied by hysteresis index h as feedback, adds input threshold quantizer gray value u _m,nin, then the output valve b of pixel (m, n) _m,n:

$b_{m,n}=Q\left(u_{m,n}\right)=\left\{\begin{array}{cc}1& (u_{m,n}-h\underset{p,q\&Element;R}{\&Sigma;}{H}_{p,q}{b}_{m-p,n-q}^{\&prime;})>0.5\\ 0& else\end{array}\right.$ (formula is picked up)

Wherein, R is the coefficient template scope of sluggish filters H (), and output valve 1 is print point, and 0 is not print point; P, q are the ranks value of the coefficient template of H ();

Step 2.4: by light energy value A _center, A _nerband A _diagsubstitution formula land, the light energy superposition value on calculating pixel point (m, n) region, the value of u, v, w in its Chinese style land is determined by the output valve of point processed in pixel (m, n) eight neighborhood;

$u=\left\{\begin{array}{cc}1& b_{m,n}=0\\ 0& b_{m,n}=1\end{array}\right.$ (formula 11) (formula 12)

(formula 13)

Step 2.5: by optimized parameter T _1,t ₂the modulating output gray value of current point is calculated in substitution formula seven;

Step 2.6: the error amount e calculating current point _m,nfor analog pixel value and the difference e of input threshold quantizer gray value _m,n:

E _m,n=b ' _m,n-u _m,n(formula 14)

Step 2.7: described in step 2.1 to step 2.6, from angle point beginning, pending image upper left reason, by order left-to-right, from top to bottom, until pending image bottom right angle point terminates.