CN101088163A

CN101088163A - Image sensor with globally separate colour zones

Info

Publication number: CN101088163A
Application number: CNA2005800443244A
Authority: CN
Inventors: L·布里索
Original assignee: e2v Semiconductors SAS
Current assignee: Teledyne e2v Semiconductors SAS
Priority date: 2004-12-24
Filing date: 2005-12-01
Publication date: 2007-12-12
Also published as: JP2008526062A; WO2006069886A1; EP1829109A1; CA2590667A1; FR2880194B1; FR2880194A1

Abstract

The invention concerns electronic colour image sensors and in particular sensors of very small dimensions enabling miniature photographic cameras or miniature cameras (capable of being incorporated in a portable telephone) to be produced. The sensor is provided with an optical system enabling the image of a scene to be observed to be projected on an array of photosensitive zones provided in a monolithic semiconductor chip integral with the optical system; said optical system includes several optical subassemblies (L1, L2, L3, L4) and the array of photosensitive zones is divided into at least two matrices (M1, M2, M3, M4) capable of being read independently, each optical subassembly being adapted to project onto a respective matrix the whole scene to be observed, two uniform filters of different colours (F1, F2, F3, F4) being arranged respectively above the different matrices to allow through to each matrix a single-colour light.

Description

Imageing sensor with colour band of overall situation separation

Technical field

The present invention relates to the electronic color image transducer, relate in particular to the very little transducer that is used to make miniature photography device or camera (for example, it can be attached in the middle of the mobile phone).

Background technology

We wish and can make whole camera with the cheapest technology of cost as far as possible, can satisfy resolution, colorimetric quality and the compactedness requirement that improves constantly simultaneously again.

Can make color image sensor by following manner: various operations are carried out in the front to silicon wafer, for example, deposit that mask, impurity inject, have the temporary layer of various compositions or setting layer are set, to the etching of these layers, heat treatment etc.; Adopt these operations to define sensitivity speck matrix and being used to and handle circuit with the signal of telecommunication of these spot correlations; Afterwards, deposition of color filters layer on the front of described silicon wafer carries out independent etching to them, to define matrix pattern; Described matrix comprise be in the row and column, at group each pixel of transducer, that constitute by the colour filter of three or preferred four different colours.Each basic filter is positioned on the corresponding photosensitive district, is used to receive the light of solid color.The tight adjacent filter that is positioned on next-door neighbour's the photosensitive area has different colors.Therefore, each pixel comprises four colour filters (common two is red for green, and is blue) that are positioned on four adjacent photosensitive areas that define pixel in theory.

Because the high cause of resolution, adjacent area leans on very closely mutually, thereby there is a risk of suitable generation optical crosstalk, described optical crosstalk or owing to be that some light of target has dropped on and causes on the adjacent area with a certain photosensitive area, or because in fact some light induced electrons in a certain photosensitive area are captured by adjacent areas is caused.Certainly, it has caused certain loss to spatial resolution, and it mainly influences the image scene with high spatial frequency.But, the most important thing is, when relating to colorimetric method, it is very crucial that this optical crosstalk phenomenon between the adjacent area is proved to be: even the image-region that only has a low spatial frequency (for example, image-region with uniform redness) also might be affected---owing to the neighbor corresponding to other colors systematically detects not is to be the electronics of target with it, thereby quality of colour has been subjected to systematically reducing.The electronic image that is produced by even red area no longer is red, but comprises green and blue composition.

This has caused the optical crosstalk of deterioration of colorimetric method particularly responsive in the CMOS technology, because in this technology, crosstalk sources is from the big height (several microns) of the layer between colour filter and photosensitive area.

In color image sensor, also there are other colorimetric method problems, especially moire problem, this problem is to be handled by performed interpolation to cause, the purpose of carrying out this processing is to provide to some pixel the brightness value of first color, described pixel does not also correspond to this color, but they are between two pixels that show this color.

In patent FRA2829289, disclose a kind of structure, reduced colorimetric mass defect problem thus with substrate of attenuate.

Summary of the invention

The present invention proposes a kind of mode of improving colorimetric method, it both had been applicable to conventional structure, also was applicable to the structure of attenuate substrate, and was applicable to any this The Application of Technology (CMOS or non-CMOS) that adopted.

Theme of the present invention is a kind of color image sensor that is provided with optical system, described optical system be used for the projects that will will observe to the integrated same single block of semiconductor chip of described optical system on the photosensitive area array made, it is characterized in that, the photosensitive area array is divided at least two matrixes that can read separately and described optical system comprises several optical modules, the whole projects that each optical module is designed to will observe is to corresponding matrix, the even filter of two different colours places respectively on described each matrix, thereby makes the light of solid color lead to each matrix.

Therefore, be not the neighbor that adopts different colour filter covering sensors, but make each matrix in the middle of all matrixes be covered that described monochromatic filter is different from the monochromatic filter of adjacency matrix by monochromatic filter.

At present, knownly can produce the optical system with very short focal length, described optical system can project the focusedimage of scenery on the matrix with hundreds of thousands pixel, and each limit of described matrix has several millimeters size.Can be in the combination manufacture process of transducer with these system's manufacturings and be installed on the described transducer, that is to say, before silicon wafer is divided into independent transducer, with its manufacturing and be installed on the described wafer.Therefore, might gather to make and comprise the plate that has much relative to each other obtained the optical module of point-device location, and set manufacturing has been carried the silicon wafer of each transducer, make the matrix of described transducer obtain point-device location by the mode identical with described optical module, can and put described plate and described wafer, thereby make each optical module in the face of the corresponding photosensitive matrix; Afterwards, described wafer/plate sub-assembly is divided into single-sensor, each transducer comprises several optical modules and several matrix, and it can be with the corresponding monochromatic complete image that receives the scenery that will observe at each matrix.

In practice, adopt four matrixes, if described matrix is a square, so with these matrix arrangements in square, if described matrix is a rectangle, so with these matrix arrangements in rectangle, diagonal along described square or rectangle is placed two matrixes relevant with green filter, place two matrixes in addition along another diagonal, one of them is relevant with red filter, and another is relevant with blue filter.

To get up from the signal combination of the different matrixes of same sensor chip, thereby the aggregate performance of image will be provided with various color.Described binding operation just will be received from the luminous flux that is positioned at different matrixes and is in several (four) zone of identical relative position at each matrix simply, distribute to the same point (point comprises several chrominance components) of electronic color image, rather than the luminous flux that will be received from several adjacent photosensitive areas of same matrix is distributed to this point.Described relative position will be considered the location defective, and it is illustrated between the center of the center of optical module and corresponding matrix may exist skew, and described skew may be different for each different optical module.

The invention enables solution, solving the colorimetric method problem that prior art faced in some cases to small part becomes possibility.This is because crosstalking of generating between the present adjacent photosensitive area only acts between the same color pixel.Described in prior art, this is crosstalked the forfeiture by resolution is showed, and this is inevitably, and must cause contrasting the infringement of color method.

This solution hypothesis in theory, projects the image of the scenery that will observe on each matrix in an identical manner exactly, and that is exactly, and any one picture point of the scenery that observe is projected on four matrixes with identical relative position.But, if the accuracy of described optical module is not enough to guarantee the consistency of this relative position, will see so after a while, can find the electronic calibration measure that compensates these site errors to small part.

Particularly, described matrix is preferably greater than the image of the scenery that will observe.In this way, suppose because the relative positioning defective between described optical module and the described matrix the former thereby central point of the scenery that will observe is not projected exactly the center of four matrixes, might calibrate described position error so, the signal of each matrix output is taken place and the proportional electronics skew of viewed error, the entire image of the scenery of observing still is projected onto on four matrixes, though the center of described image drop on no longer exactly these matrixes in the heart.

Should be understood that, have only when comparing with the interval at the center of two optical modules, when the distance from the object to the transducer was very big, the center that is projected two object images on the matrix that is equal to by two optical pickocffs that are arranged side by side just can be positioned on two matrixes in the same way.If described object is placed non-infinite distant place, object leans on closely more so, project two images on the matrix just tend to more mutually away from.The function (for given optical module interval for) of described skew as object distance can be calculated, this skew can be taken into account, thereby the image that each matrix is provided become point-to-point correspondence.The relativity shift between the image on two adjacency matrixs and the direct ratio that is partitioned into of corresponding optical module are inversely proportional to the distance of object.

According to an important characteristic of the present invention, can calculate this skew under following situation, that is, described transducer comprises two matrixes at green, a matrix and the matrix at blueness at redness.In this case, (for example take very near object if wish, residing distance is less than N times of the matrix focal length, wherein, N is the quantity that projects the pixel of the delegation's image on the matrix), to utilize this fact so, promptly separate corresponding to the distance of two optical modules of green matrix identical with the distance of separating with the assembly of red and blue matrix correlation.Its target is to seek the overlapping skew of image optimum that makes two green matrixes by the image related algorithm, and must make the skew of different greens, blueness and the generation of red matrix by its deduction, thereby sets up the image correspondence between the pixel of four matrixes.

According to another characteristic of the invention, in order to improve colorimetric method, to no longer adopt two green matrixes, a red matrix and a blue matrix, be covered with four kinds of different colour filters, especially the matrix of redness, green, blueness and cyan filter and provide four.

Description of drawings

To become apparent by reading the following embodiment other features and advantages of the present invention that provide in conjunction with the accompanying drawings, wherein:

Fig. 1 shows the example of the known configuration of the color image sensor with photosensitive area array of arranging according to row and column in the mode of top view;

Fig. 2 is with less engineer's scale, show same transducer by sectional view, and described transducer has the integral image that will remain to be observed and projects optical system on the photosensitive area array;

Fig. 3 shows according to transducer of the present invention with the form of sectional view, and it has several with the optical module of same image projection to the matrix of several different colours;

Fig. 4 shows the top view of the transducer of Fig. 3;

Fig. 5 shows and is used to illustrate because the schematic diagram of the inherent drift that the relative positioning error between described optical module and the described matrix causes;

Fig. 6 shows a kind of modification, and wherein, four matrixes are four quadrants of bigger matrix;

Fig. 7 shows the one exemplary embodiment that has several lens on each matrix;

Fig. 8 shows the schematic diagram that is used to illustrate image shift, and described image shift is that two the optical module off center (off-centering) owing to the same image of observation that is arranged on non-infinite distant place cause; And

Fig. 9 shows practical matrix embodiment, and it allows the location of image is defined within the half-pixel, thereby improves the resolution that adopts two green matrixes.

Embodiment

Fig. 1 shows the example of the general layout of the pixel of the photosensitive matrix of color image sensor in the CMOS technology in the mode of top view.Only show several pixels among the figure, but should be appreciated that described matrix can comprise a hundreds of thousands pixel, for example, 1,200,000 pixels.In this example, the photosensitive area of these pixels has the basic octagonal shape that is.Control described pixel by being generally level with vertical lead (not shown), described lead is not straight fully, and it follows described octagonal outline line on the contrary.

Fig. 1 also shows the tissue of described color: letter r, G and the B relevant with each pixel represents that corresponding pixel is covered by the independent filter of red (R), green (G) or blue (B) respectively.Neighbor color alternate, the quantity of green pixel are the twices of red pixel and blue pixel.

Fig. 2 adopts the engineer's scale littler than Fig. 1, schematically shows the method that can produce the color image sensor that has following parts with the form of sectional view, and color image sensor described here is the thin silicon transducer:

-have and be electrically connected 12 substrate 10;

-the matrix of the photosensitive area of formation in the silicon substrate 20 of extremely thin (thickness) with about 10 micron dimensions;

-color filter matrix layer 30, wherein, neighbor is covered by filter as shown in Figure 1, different colours R, G, B alternate; And

-optical image forming system 40, it can project the entire image (passing through colour filter) of the scenery that remains to be observed on the photosensitive matrix.Described optical system is made of one or more transparent panels, adopts these plates to form an one or more glass or transparent plastic lens that are stacked on another, only shows single lens L here.Described optical system accurately must be positioned on the described photosensitive matrix along short transverse, thus with the image focusing of viewed scenery to the plane of described matrix.Fig. 2 shows the transparent separate layer 35 of the thickness with explication, and it is used to guarantee this accurate location.

Fig. 3 shows according to transducer of the present invention in the mode of sectional view, and Fig. 4 shows this transducer in the mode of top view.

The sensitivity speck matrix is divided into the several matrixes that are formed on the same integrated circuit (IC) chip, and described optical system is divided into as many optical module.The complete optical imagery of the whole scenery that each optical module will remain to be observed projects on the corresponding matrix.The single filter that employing has the homogeneous color covers each matrix.

In a preferred embodiment, four that represented by M1, M2, M3 and M4, square or rectangles are arranged and put matrix (two row, two matrixes of every row), described matrix is covered by corresponding filter F1, F2, F3, F4 and corresponding optical module L1, L2, L3, L4, wherein, each described optical module is expressed as lens.Matrix M 1 and M4 diagonal angle arrange that matrix M 2 and M3 are also like this.Two matrixes that the diagonal angle is arranged, filter F1 and the F4 for M1 and M4 is preferably green here.The filter of two matrixes is respectively red (F2 on the matrix M 2) and blue (F3 on the matrix M 3) in addition.Can only see filter F1 and F2, optical module L1 and L2 and matrix M 1 and M2 among Fig. 3.

Described optical module with essentially identical visible projects to each described matrix (only have little drift, as mentioned below, this owing to described matrix relative to each other lateral shift cause).

Among the figure described matrix is shown square matrices, but they also can be the rectangles of the picture of picked-up rectangular format.

In the pinpoint accuracy industry of being implemented by micromechanics and little packaging technology is made, it is identical described optical module can being considered as, have identical focal length, and the location with regard to regard to the distance of the common sides at described four photosensitive matrix places, having obtained accurately, be equal to.

In this case, suppose that the image that remains to be observed just very simply represents the white rectangle test pattern of the theoretical profile of scenery, this profile will be used as four rectangular profile that are equal to and project on each matrix so, shown in the dashed rectangle among Fig. 4.Here, take not have the simplification hypothesis of image fault, if there is distortion, it also all is the same for all optical modules.

Described dashed rectangle may not be located in the same way with respect to the photosensitive matrix that they will throw.Its reason at first is, described optical module " is not seen " described rectangle at definite same position, because relative to each other there is lateral shift in they, this will form proportional shifting with two images that the people saw of taking pictures along same direction of standing side by side along with the skew of camera be the same.Here, the skew of described optical module is several microns, and magnification ratio is multiply by in the skew that the skew to a certain degree of the rectangle that it causes equals transducer.Here, magnification ratio is defined as the ratio of the size of the size of images that projects the object on the matrix and object self, therefore for the scenery that is in the infinity, skew is zero.In addition, make the restriction of accuracy owing to when transferring to optical module on the matrix, there is industry, thereby described optical module may not obtain accurate located lateral with respect to matrix according to the requirement of necessity, thereby its center may not be vertical exactly with the center of matrix.Therefore, exist on the one hand because the skew that the near relatively proximity of object causes exists on the other hand owing to the undesirable intrinsic skew that cause of optical module with respect to the relative positioning of matrix.

But, if described optics is equal to, and with the surperficial equidistant setting of photosensitive area of these rectangles of projection, the size of the rectangle that is throwed all is accurately to equate for all rectangles so.

Therefore, might be in manufactory or subsequently by calibrating the intrinsic relatively skew of determining different matrixes at least.Be arranged on enough big distance (for the matrix that constitutes by the capable pixel of N, this distance is greater than the N of focal length doubly) locate will on each matrix, generate image such as the white test pattern of above-mentioned rectangle or by the simple image that several white points constitute, confirm corresponding relation by shift vector by described four width of cloth images, described afterwards corresponding relation will enough greatly four width of cloth images of the observed any scenery of distance be converted into point-to-point correspondence.

Fig. 5 shows the top view of the matrix M 1 of the calculating that is used to illustrate the intrinsic relativity shift between the matrix.

The photocentre of optical module of representing to be positioned at the vertical direction of matrix M 1 by O1.This assembly is in the center distance of infinity with respect to the central shaft O of transducer the rectangular image IM1 of object projects on the described matrix.The center of institute's image projected is vertical with some O1, therefore overlaps with this point in top view shown in Figure 5.Can think that matrix M 1 self has center C 1, if the position adjustment between optical module and the matrix is desirable, C1 also will overlap with an O1 so.For simplicity, can think that center C 1 is the central row of available part of matrix M 1 and the intersection point of central series.Owing to there is defective, center C 1 may be offset a little with respect to center O 1.Sign matrix IM1 and center O 1 are with respect to the accurate position of center C 1 in the coordinate system that pixel column and pixel column by matrix constitute.In factory's internal calibration process, carry out this sign operation.Also adopt image I M2, IM3 and the IM4 that projects same rectangle on other matrixes, that be in the infinity, carry out this operation at other matrixes.

With regard to matrix M 1, point O1 and C1 are along the alternate position spike of x axle and y axle, and for other matrixes, similarity O2 and C2, O3 and C3 and O4 and C4 will determine the skew along row and column that certainly leads to when reading the electronic signal of each matrix output along the alternate position spike of x axle and y axle, so that make each picture point of observed scenery corresponding with the corresponding points in each matrix, wherein, suppose that described scenery is in the infinity.

With these offset storage in calibration register.Therefore, these registers will contain the information corresponding to the relativity shift between the center of the center of optics and each matrix.Afterwards, the electronic signal of four matrixes being exported according to the content of these registers combines, and is offset to compensate these, and rebulids correspondence accurate, system between the image of each matrix output.

(but simultaneously preferred) collects the electronic signal by each matrix output respectively.

Matrix M 1 and matrix M 4 will provide the signal of the green component of the image of representing observed scenery.Matrix M 2 and matrix M 3 will provide the red component and the blue component of image respectively.

Matrix shown in Figure 4 is an adjacency not.Might lay electronic driver in the slit between matrix as required like this.But, as shown in Figure 6, also can preferably have such layout, wherein, and put four matrixes, and make it only to form single big matrix M continuously.Therefore, this matrix must be divided into four quadrants that can read separately in theory.Therefore, preferably will be divided into four elements with the reading circuit of described matrix correlation, as shown in Figure 6, it has the form with readout register R1, R2, R3 and the R4 of four corresponding quadrant dependences of matrix M.

No matter with regard to regard to the structure shown in Figure 4 or with regard to structure shown in Figure 6, in both cases, in order to ensure the relative positioning error of each optical module being taken into account, the active surface of wishing matrix can be than big at least tens row and tens row of maximum area of the image of observed scenery on area.Like this, can consider intrinsic position error better and the skew of the image that causes by captured being in the approximated position of object, get rid of simultaneously because image spills into the risk of losing a part of observed scenery outside the matrix.

(for example suppose wish to obtain resolution and be 1,200,000 pixels, comprise 1000 row and 1200 row pixels) image (pixel is corresponding to the picture point of particular color), to adopt the matrix of four juxtaposed different colours to implement the present invention so, each matrix comprises about 500 row, 600 row, and the pixel of these matrixes has identical size and spacing with the pixel of three colour moment battle arrays of available technology adopting.

Suppose that the length on the limit of the matrix of different colours is half (for equal resolution, same pixel area and identical sensitivity) of length of the whole matrix of prior art, the optical module of projects images also has 1/2nd lateral dimension so.As a result, the focal length of optical module becomes 1/2nd.Therefore, not only thickness is littler for they, and the present position is more near the surface of photosensitive matrix.Correspondingly reduced the gross thickness of transducer.Sensor thickness in the manufacturing data of compact image transducer in occupation of the status that becomes more and more important.The improvement of this parameter that obtains by the present invention is very significant.

The transducer that adopts the manufacturing of set manufacture process to have such formation, that is to say, not only on same integrated circuit (IC) wafer, make a plurality of imageing sensors, also make the plate that comprises a plurality of optical systems be attached to this wafer, afterwards, described wafer/plate sub-assembly is cut into single-sensor, and each transducer comprises the silicon integrated with optical system.

Described optical module is made of the transparent panel that has formed lens on it.These plates can be made of glass, and it is carried out mold treatment.Described assembly can comprise one or more stacked lens.If several lens (assemble and/or divergent lens) are arranged,, that is, make slit between the lens corresponding to the optical function parts that will form so by the stacked described plate of such mode.Can adopt in the same way, form fixed aperture by the transparent panel that is coated with opaque layer (no matter whether having carried out molded with the form of lens to it), wherein, described opaque layer is opened wide around the optical axis of each optical module.

Fig. 7 shows a kind of imageing sensor, and wherein, each optical module comprises two stacked lens that are formed in the stacked

molded glass plate

41 and 42, and described

glass plate

41 and 42 separates by demarcation strip 43, to adjust the expection perpendicular separation between the lens.One of described lens are coated with opaque layer 44 (for example, aluminium lamination), and described opaque layer 44 is local opening wide four positions, thereby define (fixing) diaphragm 45 on each matrix, will pass described diaphragm from the light of observed scenery.

Also can on another plate that is positioned on the plate 42 and keeps at a certain distance away with it, define diaphragm.Described diaphragm can have the effect of field stop, and in this case, it preferably has rectangular shape (identical with the shape of matrix), thereby prevents that matrix is subjected to the illumination of the optics relevant with adjacency matrix.

Transducer according to the present invention is particularly useful for the situation that observed scenery and imageing sensor keep enough distance, thereby makes image shift (rather than because intrinsic skew that manufacturing defect causes) become and can ignore.

For scenery nearby, (for example will cause comparing the skew of can not ignore with Pixel Dimensions, for the N row matrix, the residing position of scenery is less than N focal length doubly), at this moment can adjust described skew according to the degree of closeness of scenery---the skew of required compensation is directly proportional with the interval d of optical module, is inversely proportional to the distance of observed target.

With regard to transducer that constitute, that comprise two green matrixes by four matrixes, might utilize such actual conditions, if the skew that promptly has intrinsic skew or caused by the proximity of target, two green matrixes should be seen identical image so.We can see, can calibrate intrinsic skew in factory, and in the mode of system it be paid attention to afterwards, thereby when considering remaining explanation, only need the skew that the proximity by target causes is calibrated.Therefore, will think that hereinafter photocentre O1, O2, O3 and the O4 of optical module overlaps with center C 1, C2, C3 and the C4 of matrix M 1 to M4 respectively.

When captured target is in non-infinite distant place D, the image of this target of the opticses projection that is equal to apart from two of d by Center Gap is with respect to the central cross skew of described optics, offset distance and d/D are proportional, perhaps more precisely, described offset distance equals dF/D, wherein, F is a focal length.

In other words, with reference to figure 8:

-in Fig. 8 a, the single optical module L that imagination has center O, its image (vector vs) that will be in the object (vector V S) at distance D place projects on the described matrix.The center of object VS and image vs is provided with respect to the imaginary centres C of matrix, and this center is vertical with photocentre O.

-in Fig. 8 b, adopt identical with optical module L among Fig. 8 a, but lateral shift substitutes described simple optic apart from two optical module L1 and the L4 of d each other, that is to say they center O 1 and O4 in fact with respect to the center O lateral shift+d/2 and the d/2 of single optical module.Their observe with Fig. 8 a in identical target (still representing) with vector V S, the center of the set that the center of this object still constitutes with respect to two optical modules is provided with, that is to say that its center still is arranged on the center O, in fact its representative comprises the central line shaft of the imageing sensor of two optical modules.Therefore, vector V S is in center O 1 off-centre (off-center) of a side with respect to first optical module, in center O 4 off-centre of opposite side with respect to second optical module.Introduced parallax effect by non-infinite these actual conditions of D.

Image vs1 that is provided by two optical modules and vs4 be offset distance d under the influence of spacing distance d between the optics and each other not only, and because each optical module carries out lateral observation to described target, exist eccentricly, thereby described image also will be offset extra apart from dF/D mutually.

Suppose that C1 is the center of matrix M 1, its center O 1 with the optical module L1 that is positioned at the left side is vertical, and C4 is the center of matrix M 4, its center O 4 with the optics that is positioned at the right side is vertical, C1 and C4 offset distance d that is to say C1 with respect to C skew+d/2, C4 skew-d/2.

The dF/2D such as distance that the image vs1 that is provided by optics L1 is offset to the left with respect to center C 1.In other words, the center of the image of described target is not to overlap with center C 1, but offset d F/2D to the left, and similarly, the image vs4 that is provided by optics L4 is with respect to center C 4 distance of offset d F/2D to the right.

With regard to the system that comprises four projection optics assemblies according to the present invention, such actual conditions have been utilized, promptly suppose along two green matrixes (M1 and M4) of diagonal layout and see identical scenery, but, as long as the magnification ratio non-zero, this scenery will be offset on described two matrixes.

Employing to from the image of matrix M 1 with seek shift value from the related calculating between the image of the skew in succession of matrix M 4, this deviant can provide optimal relevance, and therefore representative projects the total drift that exists between two images on the matrix.The intrinsic skew of supposing not exist the bad location by optical module to cause perhaps deducts the calibration value corresponding to intrinsic skew from total drift, the offset d F/D that so remaining is exactly causes owing to the proximity of target.This skew is a vector, and this vector place direction is the diagonal that connects the recalibration center of green matrix.

This skew searching algorithm is simple, because because the direction of the skew that the proximity of captured object causes is known---extract the picture strip (image band) that probably contains observed main target from matrix M 1 and matrix M 4.After considering calibration, with respect to center sign (referenced) these picture strips of each matrix---thereby, identify these picture strips with respect to the center O 1 and the O4 of optical module.This sign all is the same for two matrixes, that is to say, if on the center O that is centered close to matrix M 11 of described picture strip, the center of the correspondence band in the matrix M 4 then is positioned on the O4 so.Described picture strip is the sample of entire image, because it is related to need not to adopt entire image to set up.

Picture strip from M1 and M4 is subtracted each other each other.Determine and store the illumination (this value can be the very simply average illumination of image that is obtained by described subtraction) of the representative image that obtains by described subtraction.

Afterwards, make described image pattern be offset a distance increment (this means the skew of an one-row pixels and a row pixel) along the diagonal at the center that connects described matrix, thereby to its correct.Therefore, obtained two other picture strip now, its center with respect to the off-centring of matrix one increment.Repeat described subtraction, and determine the illumination (illumination) of the image that obtains by described subtraction.

Repeat this operation in succession, it increases progressively the relative position of described band at every turn with respect to the center of matrix.Carry out described increment operation along a certain direction, wherein, take further to separate two picture strips of (but not close mutually) in succession along cornerwise direction at the center that connects described matrix.Can understand this operation by check Fig. 8 b: when captured gtoal setting imageing sensor, described image will be further towards the skew of cornerwise outside.A kind of more easy way is detected the existence of the skew of continuous increase gradually from being in infinite distant place corresponding to target, not having the ideal situation of skew.

Its target is to find a certain skew, and this skew has provided minimum brightness value for the image that is obtained by described subtraction.Above-mentioned dF/2D value is represented in this skew, because but minimum brightness value is the brightness value corresponding to the overlay chart image-tape, as can be seen from Figure, for the distance D of captured object, after all mutual in opposite direction offset d F/2D of each image, described image becomes can be overlapping.

In case found this best dF/2D value, still needed so and adopt this value:

-at first, determine the normal green image that each green matrix will provide, just as described green matrix is positioned at the center O place of transducer just, and off-centring d/2 is not the same thus.Therefore, draw close dF/2D by the image that makes matrix M 1 to center C 1 image of matrix M 1 is offset, and draw close dF/2D by the image that makes matrix M 4 to center C 4 image of matrix M 4 is offset; And

-secondly, especially derive blue and the required skew of red matrix.The described skew that is caused by the proximity of captured object is identical between two green matrixes, make the amount of the red image of matrix M 2, make the blue image of matrix M 3 just enough along the identical amount of this diagonal skew along the diagonal offset d F/2D between the center of matrix M 2 and M3.The offset d F/2D of described redness and blue matrix is in rightabout along this diagonal, and is in and makes image more near the direction at the center of each matrix because the layout of described optical module tend to make they more away from.

Afterwards, adopt the skew of calculating at each matrix (green, blue and red) to determine compound sequencing (recombination sequencing), it makes each picture point corresponding with the respective point in each matrix, finally obtains complete coloured image.

In the algorithm of seeking best offset value, can carry out more meticulous increment operation, for example, only one of two picture strips are increased a pixel, afterwards another picture strip is increased a pixel.Also may follow increases a pixel, but does not increase along row, and waits for that the edge row increase a pixel, but does not follow the next increment computing of increase.

Thereby, the color image sensor with improved colorimetric method has been described, even it also can provide accurate image at close target here.

Also may be and improve the apparent resolution of transducer by the existence of two green matrixes.In order to realize this purpose, described green matrix is relative to each other accurately located, make project one on the matrix image and project the relativity shift that has half-pix between the image on another matrix (for the purpose of simplifying the description, suppose to consider and proofreaied and correct the intrinsic skew that the non-accurate location by the center that is in the optical module on the described matrix causes).Therefore, in green channel, there is bigger sampling.Described skew is the half-pixel that follows and along the half-pixel of row.Institute's image projected is also inevitable to be offset half-pixel mutually along described diagonal.

A kind ofly two images are carried out half-pixel adopt such matrix in the possible mode of interior location, constitute 1/2nd of the size of pixel of this matrix and expection Pixel Dimensions that spacing is described matrix, and the electric charge on the execution neighbor add and, thereby the pixel of reconstruction of standard size.Before reading, on each pixel, carry out this and add and operate, and regardless of being at CCD or in the CMOS technology.Have normal size according to it being carried out the adjacent than small pixel of add operation, forming, the center is in the pixel within the half-pixel, and described two matrixes can have different drivers, thereby realize effective half-pix skew between them.

In the CCD technology, transmit gate control mutually by four and in readout register, carry out and describedly add and operate (being also referred to as " warehouse entry (binning) ").

In the CMOS technology, carry out at the electric charge storage node place relevant and to add and operate with four little adjacent photodiode, by transmission gate described node and these photodiodes are separated, thereby by these suitable control being selected which adjacent photodiode will see that its electric charge is transferred to the designated store intranodal.Adopt nine adjacent photodiode of pressing the row and column tissue, described photodiode might be divided into the group that is made of four adjacent photodiode, electric charge storage node then is positioned at the central authorities of each group.Four kinds of possible groups are arranged, and these groups have constituted along row or column or have been spaced from each other the bigger pixel of half-pixel along diagonal.

Therefore, if a matrix is organized its pixel according to certain mode, another matrix can be organized its pixel according to another mode, is along cornerwise half-pixel thereby make two skews between the matrix.

Fig. 9 shows a specific embodiment, has provided the octagon photodiode among this embodiment, the gross area of described octagon photodiode equal for the expection pel spacing the photodiode size that should have 1/4th.Arrow shows from four photodiodes and discharges to memory node, and described memory node is surrounded by these four photodiodes.The selection of the transmission gate that activates at photodiode determines that this photodiode will be to described memory node discharge.In Fig. 9 a, by the excitation transmission gate four diodes are made up according to certain mode, and in Fig. 9 b, then described diode is centered around around the memory node in a different manner by activating transmission gate.Make the memory node (the first green matrix) that adopts among Fig. 9 a all with respect to the memory node that adopts among Fig. 9 b (the second green matrix) skew half-pixel.

If colorimetric method is than the important parameters all of the accuracy in resolution and the image overlay, perhaps the user wishes to see the image of the far plane scenery that under any circumstance accurately superposes, can provide transducer so with four matrixes, described four matrixes are by the filter of four different colours, promptly red, green, blueness and cyan filter cover, each filter all with a matrix correlation.Certainly, like this will be no longer may determine and calibrate parallax, so the user be confined to watch distant view by handling from the image of two matrixes.

Claims

1, a kind of color image sensor that is provided with optical system, the image projection that described optical system is used for the scenery that will will observe to the integrated same single block of semiconductor chip of described optical system on the photosensitive area array made, it is characterized in that, described optical system comprises several optical module (L1, L2, L3, L4) and described photosensitive area array be divided at least two matrix (M1 that can read separately, M2, M3, M4), the whole projects that each optical module is designed to will observe to corresponding matrix, the even filter (F1 of two different colours, F2, F3 F4) places respectively on described each matrix, thereby makes the light of solid color lead to each matrix.

2, imageing sensor according to claim 1, it is characterized in that, with described photosensitive area array partition is four matrixes that are arranged in the square, two matrixes of arranging along described foursquare diagonal are relevant with green filter, two other of being arranged in the matrix on another diagonal is relevant with red filter, and another is relevant with blue filter.

3, imageing sensor according to claim 1 and 2 is characterized in that, it comprises at least one calibration register, and described calibration register contains the relativity shift between the center of the center of described optical module and described matrix.

4, according to a described imageing sensor in the aforementioned claim, it is characterized in that, described matrix is greater than the image of the scenery that will observe, thereby can be at the relative positioning error of considering under the situation of not losing image segments between described optical module and the described matrix, even the center of the entire image of the scenery of observing no longer accurately be positioned at these matrixes in the heart, described entire image still is projected onto on described four matrixes.

5, according to a described imageing sensor in the claim 1 to 4, it is characterized in that, it comprises the distance according to the object in the scenery that will observe, the distance that image relativity shift by making two adjacency matrixs and the distance that is partitioned into direct ratio and described object of corresponding optical module are inversely proportional to, and make the device of the point-to-point correspondence of image of described each matrix.

6, imageing sensor according to claim 2, it is characterized in that, it comprises and is used for seeking the overlapping skew of image optimum that makes two green matrixes by the image related algorithm, and the skew that different greens, blueness and red matrix are taken place by its deduction, thereby between the pixel of four matrixes, set up the device of image correspondence.

7, imageing sensor according to claim 1 is characterized in that, it comprises four by the filter of four kinds of different colours, especially the matrix of redness, green, blueness and cyan filter covering.