CN1128534C - Correcting method for touch scanner with image sensor - Google Patents

Abstract

The present invention belongs to the technical field of a correction method of a scanner. The correction method comprises: firstly, different scanning conditions between 0% and 100% of lighting intensity for n times are set. Then, a pure white labelled strip is scanned according to the set condition. Invalid noise data in the obtained data is removed, and then, a correction table is generated. By an interpolation algorithm, the invalid data of the correction table is filled. By using the obtained correction table, an actual scanning algorithm is carried out, and the digital filtering operation is carried out. The correction method can effectively avoid the contradiction that the labelled strip needs to be pure white. A practical standard for uniform image brightness distribution is achieved during scanning.

Description

Correcting method for touch scanner with image sensor

Technical field

The invention belongs to scanner calibration method and technology field.

Background technology

Scanner is as the computer peripheral equipment of optical, mechanical and electronic integration, developed the course of more than ten years, perfect development along with CCD (charge coupled device) device, the gold period of also having entered of falt bed scanner, but traditional CCD device needs huge light path support, is difficult to satisfy ultra-thin, the extra small developing direction of scanner now.Therefore adopt the corresponding generation of scanner of contact-type image sensor (being called for short CIS) device.

RGB (RGB) color space that scanner adopts carries out color description, needs to accomplish the balance unanimity between every kind of color, and scanning result at this moment just can be true and reliable.

The all hardware product all can have worker's difference scope, and scanner is no exception, and for guaranteeing the product quality when dispatching from the factory, producer all can define some examination criterias that dispatch from the factory.Owing to need the project of test a lot, be example with the scanner standard of dispatching from the factory, it is as follows to list some indexs relevant with the present invention:

1, final output image evenness errors is less than 10%.The brightness number in-scope width of exporting when the uniformity is meant the even original copy of scanning same color.When for example scanning the grey original copy of neutrality, output numerical value should be 128, but actual result may be between 122～132, and the criterion of acceptability of scanner is that the difference of maximum and minimum value can not be greater than 10%.

2, final output image color consistency error is less than 10%.Three channel offset of brightness number RGB of exporting when color consistency is meant the even original copy of scanning same color.When for example scanning certain brightness ash original copy, three passage numerical value of the RGB mean value of output should be consistent, but actual result can be not quite similar, and the criterion of acceptability of scanner is that the difference between three passage numerical value of the RGB mean value is less than 10%

Because initial data do not reach the criterion of acceptability that dispatches from the factory, therefore having an important function in scanner drives is exactly image rectification, and this is that all scanners all need the function that possesses.

Adopt the scanner of CIS device not need extra light path support just can directly scan whole breadth, can think that therefore its sensitivity speck is that the position concerns one to one for actual original copy information.But because the problem of manufacturing process, the photobehavior between the sensitivity speck is not quite similar because of process deviation.Simultaneously the CIS device is owing to make width less than A4, and mostly the CIS probe inside that can buy at present is to be made up of multi-disc CIS splicing.With the present very common 600bCIS of Toshiba is example, and it is made up of 8 splicings.

The brightness of Fig. 1 shows when being the pure white and neutral grey original copy of scanning respectively data line is to pixel curve 1,2.Under different grey original copy situations, the proportionate relationship between the pixel is not constant as can be seen.With the 613rd pixel in scheming is example, and numerical value of two pixels is a left side＞right side in scanning during pure white original copy about it, and is a left side＜right side when the neutral grey original copy of scanning;

What Fig. 2 showed is exactly in whole GTG (black～white) scope, the curved line relation between 8 segment data averages.Error ratio between 8 segment data averages is bigger, and when causing the even sample manuscript of actual scanning color like this, different position luminance differences are apart from bigger, as shown in Figure 3.

Scanner oneself has self correcting capability in order to make, a pure white demarcation bar is placed in the capital above probe, demarcate bar by scanning this, just can obtain the proportionate relationship between each sensitivity speck under the illumination saturated conditions, by this ratio, can guarantee that the uniformity degree of image when the pure white original copy of scanning reaches criterion of acceptability.But the proportionate relationship of CIS device between each sensitivity speck under the different GTG original copy situations is not constant, satisfy the demand of uniformity degree of image, will have a demarcation bar that contains the above gray feature in ten rank at least by scanner, the relativity between the sensitivity speck when being provided at each Luminance Distribution.But this is on scanner design and the production technology, all can't realize on the manufacturing cost.Also there is this defective of nicking clocklike in the original image of touch scanner with image sensor except brightness irregularities.With seeing after the local amplification of original image, there is periodic light and shade alternate nicking above, as shown in Figure 4.

Summary of the invention

The objective of the invention is for overcoming the weak point of prior art, a kind of new contact-type image sensor CIS scanner calibration method is proposed, according to the analog-to-digital principle of scanner AD, when reducing illumination intensity of light source, make brightness not reach saturated and original copy still remains when pure white, can equivalently must think and use full illumination to scan the original copy of a grey.Adopt the characteristic that the led light source can the free adjustment light intensity in conjunction with CIS, can effectively avoid demarcating bar and can only be pure from contradiction, image brightness distribution practical standard comparatively uniformly when reaching scanning.Further adopt the digital filter of band resistance to remove with the signal of nicking on the frequency domain, thereby also can reach the purpose of removing nicking with one-period.

The contact-type image sensor CIS scanner calibration method that the present invention proposes is characterized in that, may further comprise the steps:

1) at first sets the different scanning condition of N light intensity between 100% ~ 0%, scan pure white demarcation bar by the condition that sets then;

2) the invalid noise data that contains in the data that the first step is obtained is removed; The concrete grammar that noise data is removed is: establishing every row has M pixel, then obtains M*N value; Set thresholding T earlier, carry out N operation then, each mean value M that calculates an one's own profession M pixel brightness earlier, then with brightness value in the one's own profession M pixel at (M-T, M+T) all think noise outside, the mean value of pixel brightness before and after the brightness value of noise position pixel is made as, concrete grammar is: the pixel value of supposing X position is a noise, then its numerical value will be replaced by the mean value that before and after it is the numerical value of X-1 and two positions of X+1;

3) produce checking list; The concrete grammar that produces checking list is: the size of checking list is M*256, is the look-up table of corresponding 256 gray scales of said M pixel; This checking list is a two-dimensional array, and the length of first dimension is M, and the length of second dimension is 256; According to said M*N value, calculate M*N data in M*256 the data of checking list, 256 data of each row of corresponding checking list, it is that effectively 256-N data are invalid that N data are wherein arranged; The generation formula of this checking list is Table[X] [Y]=M, wherein X is the horizontal level of pixel, Y is the original value of pixel intensity; Table[X then] [Y] be exactly its corrected value M;

4) fill the invalid data of this checking list by interpolation algorithm; For each row, it is that effectively 256-N data are invalid that N data are all arranged;

5) use checking list Table to proofread and correct the actual scanning result; Use in the checking list: the horizontal level of supposing certain its place of pixel of obtaining is A, and the original brightness value is B, then target output value=the Table[A of this pixel at this moment] [B].

On the basis of described five steps, also can comprise:

(6) adopt digital filter to remove the initial data nicking; Formula is

YYY[X]＝0.9692*XXX[X]-1.8426*XXX[X-1]+0.9692*XXX[X-2]

+1.8426*YYY[X-1]-0.9385*YYY[X-2]；

XXX[X wherein] the expression original image is at the brightness value at horizontal level X place, XXX[X-1] and XXX[X-2] represent the brightness value of original image at horizontal level X-1 and X-2 place; YYY[X] the filtered image of expression is at the brightness value at horizontal level X place, YYY[X-1] and YYY[X-2] represent the brightness value of filtered image at horizontal level X-1 and X-2 place.

Adopt bearing calibration of the present invention, can obtain comparatively uniform image of Luminance Distribution during scanner work.Further can reach the purpose of removing nicking.

Description of drawings

Fig. 1 is 8 sections device red datas of the pure white and neutral grey original copy gained of CIS scanning.

8 sections device red data brightness distribution curves that Fig. 2 obtains for CIS GTG figure.

Fig. 3 is scanning blank sheet of paper picture before proofreading and correct.

Fig. 4 is the scanner original image behind local the amplification, has nicking phenomenon clocklike above.

Fig. 5 proofreaies and correct back scanning from the paper picture for adopting embodiment of the invention method.

Fig. 6 is original and filtered partial image data luminance pixel curve chart.

The contact-type image sensor CIS scanner calibration method that embodiment the present invention proposes reaches embodiment in conjunction with the accompanying drawings and is described in detail as follows:

1) at first sets the different scanning condition (80%, 50%, 30%) of 3 light intensity between 100% ~ 0%, scan pure white demarcation bar by the condition that sets then;

2) the invalid noise data that contains in the data that the first step is obtained is removed; The concrete grammar that noise data is removed is: establish M=3, one group of data that certain scanning obtains are 200,230,190; Set thresholding 15 earlier, the mean value M=206 of these 3 pixel brightness, wherein brightness value is (191,221) all think noise outside, the mean value of pixel brightness before and after the brightness value of noise position pixel is made as, outside scope, think noise as the 2nd data 230, its value replaces with the mean value 195 of preceding (200) backs (190); The data of handling are 200,195,190.

3) in the 3rd step, produce checking list; If three groups of scan-datas are (180,200,220), (140,150,160), (40,50,60).Then the M=200 that scans for the first time obtains Table[0] [180]=200, Table[1] [200]=200, Table[2] [220]=200.

The M=150 of scanning obtains Table[0 for the second time] [140]=150, Table[1] [150]=150, Table[2] [160]=150.

Sao Miao M=50 obtains Table[0 for the third time] [40]=50, Table[1] [50]=50, Table[2] [60]=50.

4) in the 4th step, as for the 0th row, it is effectively (180,140,40 these triplex rows) that 3 data are arranged, and other data are all invalid.The data of head and tail all need to set, as Table[0] [255]=255, Table[0] [0]=0.Can obtain all values of these row according to these 5 value linear interpolations like this.As Table[0] [254]=255-(255-200) * (255-254)/(255-180)=255 (going to put in order) Table[0] [253]=255-(255-200) * (255-253)/(255-180)=254 (going to put in order)

5) use the checking list Table that tries to achieve to carry out the actual scanning computing.

Present embodiment also comprises on the basis of above-mentioned five steps: adopt digital filter to remove nicking;

Go up several steps in order to simplify computing, establish M=3, promptly each row of data has 3.If the cycle of nicking is 20 probably, for filtering, this step is established M=60, and promptly every row has 60 data, so just can embody the effect of filtering. establishes input data XXX and designs suitable wave filter for [183.1067 185.1159 187.9625 189.2290 188.6609 189.1078 187.8861184.6942,182.8442 180.1903 177.9600 176.7761 174.8172 172.7494172.9248,173.1171 174.7219 177.1919 179.7698 182.3143 183.9316186.5761,187.1957 188.4964 187.9955 188.8345 186.8214 184.0072181.7398,180.9203 177.7896 174.9755 174.8055 172.9793 172.1822174.6541,176.2798 177.6882 180.2403 182.2994 185.0965 186.4795188.0061,188.9596 187.9519 188.9838 186.0571 183.8413 182.1046179.2456,176.8214 174.8444 174.0858 172.2380 172.9679 174.6705175.1086 178.2982 179.9201 183.4270] for this cycle, and formula is

YYY[X]=0.9692*XXX[X]-1.8426*XXX[X-1]+0.9692*XXX[X-2]+1.8426*YYY[X-1]-0.9385*YYY[X-2]; According to zero state response condition enactment dateout YYY[0]=YYY[1]=Average (X)=181.0110, promptly get the average of importing data.From YYY[2] begin to calculate:

YYY[2]＝0.9692*XXX[2]-1.8426*XXX[1]+0.9692*XXX[0]+1.8426*YYY[1]-0.9385*YYY[0]

＝0.9692*187.9625-1.8426*185.1159+0.9692*183.1067

+1.8426*181.0110-0.9385*181.0110

＝182.2161；

YYY[3]＝0.9692*XXX[3]-1.8426*XXX[2]+0.9692*XXX[1]+1.8426*YYY[2]-0.9385*YYY[1]

＝0.9692*189.2290-1.8426*187.9625+0.9692*185.1159+1.8426*182.2161-0.9385*181.0110

＝182.3664；

Last YYY=

[181.0110?181.0110 182.2161 182.3664 181.3872 182.1491 181.9141

180.3563 180.6108 180.2846 180.3800 181.2517 180.8897 179.8623

180.3904 180.1868 180.6870 181.4233 181.8133 181.9509 181.2163

181.7827 180.8113 181.1716 180.4658 181.8239 181.0061 180.0196

180.0060 181.5628 180.7222 179.9590 181.3277 180.3820 179.7821

181.6591 181.9290 181.4301 181.7118 181.3571 181.8300 181.2163

181.2841 181.4365 180.3999 182.1278 180.5698 180.3227 180.8925

180.4556 180.3509 180.3480 180.9985 179.8905 180.6157 181.5060

180.4606 181.6332 180.8948 181.9351]；

After coming filtering as can be seen, the variation of data is mild, has not had the sort of periodic variation.

Wherein, XXX is an initial data, and YYY is a dateout row after the filtering, in the concrete operation, makes Y[0]=X[0], Y[1]=X[1]=183, from Y[2] begin to use given formula calculating, can obtain final filtering data.After coming filtering as can be seen, the variation of data is mild, has not had the sort of periodic variation.Through behind the new scaling method, the very approaching coincidence of 8 sections device brightness distribution curves of CIS, the entire image Luminance Distribution is also comparatively even.For same pure white paper slip, proofread and correct preceding scanning effect as shown in Figure 3, scan image left and right sides brightness irregularities; As shown in Figure 5, even about scan image brightness after the correction.Fig. 6 shows is that original and filtered partial image data brightness are to pixel curve C 1, C2.Come as can be seen from figure, original brightness curve C1 has fluctuation clocklike, and amplitude is bigger; Filtered brightness curve C2 has not had fluctuation clocklike, and irregular fluctuating range is less.So just meaning has not had nicking on the image after this filtering.

Claims (2)

1, a kind of correcting method for touch scanner with image sensor is characterized in that, may further comprise the steps:

1) at first sets the different scanning condition of N light intensity between 100% ~ 0%, scan pure white demarcation bar by the condition that sets then;

2) the invalid noise data that contains in the data that the first step is obtained is removed; The concrete grammar that noise data is removed is: establishing every row has M pixel, then obtains M*N value; Set thresholding T earlier, carry out N operation then, each mean value M that calculates an one's own profession M pixel brightness earlier, then with brightness value in the one's own profession M pixel at (M-T, M+T) all think noise outside, the mean value of pixel brightness before and after the brightness value of noise position pixel is made as, concrete grammar is: the pixel value of supposing X position is a noise, then its numerical value will be replaced by the mean value that before and after it is the numerical value of X-1 and two positions of X+1;

3) produce checking list; The concrete grammar that produces checking list is: the size of checking list is M*256, is the look-up table of corresponding 256 gray scales of said M pixel; This checking list is a two-dimensional array, and the length of first dimension is M, and the length of second dimension is 256; According to said M*N value, calculate M*N data in M*256 the data of checking list, 256 data of each row of corresponding checking list, it is that effectively 256-N data are invalid that N data are wherein arranged; The generation formula of this checking list is Table[X] [Y]=M, wherein X is the horizontal level of pixel, Y is the original value of pixel intensity; Table[X then] [Y] be exactly its corrected value M;

4) fill the invalid data of this checking list by interpolation algorithm; For each row, it is that effectively 256-N data are invalid that N data are all arranged;

5) use checking list Table to proofread and correct the actual scanning result; Use in the checking list: the horizontal level of supposing certain its place of pixel of obtaining is A, and the original brightness value is B, then target output value=the Table[A of this pixel at this moment] [B].

2, correcting method for touch scanner with image sensor as claimed in claim 1 is characterized in that, also comprises on the basis of described five steps:

6) adopt digital filter to remove the initial data nicking; Formula is

YYY[X]＝0.9692*XXX[X]-1.8426*XXX[X-1]+0.9692*XXX[X-2]

+ 1.8426*YYY[X-1]-0.9385*YYY[X-2]; XXX[X wherein] the expression original image is at the brightness value at horizontal level X place, XXX[X-1] and XXX[X-2] represent the brightness value of original image at horizontal level X-1 and X-2 place; YYY[X] the filtered image of expression is at the brightness value at horizontal level X place, YYY[X-1] and YYY[X-2] represent the brightness value of filtered image at horizontal level X-1 and X-2 place.

Images