CN101150658A - A bad point self-adapted grid noise elimination device and method - Google Patents
A bad point self-adapted grid noise elimination device and method Download PDFInfo
- Publication number
- CN101150658A CN101150658A CNA2007101205992A CN200710120599A CN101150658A CN 101150658 A CN101150658 A CN 101150658A CN A2007101205992 A CNA2007101205992 A CN A2007101205992A CN 200710120599 A CN200710120599 A CN 200710120599A CN 101150658 A CN101150658 A CN 101150658A
- Authority
- CN
- China
- Prior art keywords
- point
- current pixel
- pixel point
- value
- bad
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000008030 elimination Effects 0.000 title claims abstract description 37
- 238000003379 elimination reaction Methods 0.000 title claims abstract description 37
- 238000000034 method Methods 0.000 title claims description 35
- 230000008569 process Effects 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 9
- 230000006399 behavior Effects 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 abstract description 3
- 230000003139 buffering effect Effects 0.000 abstract 3
- 230000008859 change Effects 0.000 description 11
- 230000003287 optical effect Effects 0.000 description 6
- 238000004519 manufacturing process Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 239000003086 colorant Substances 0.000 description 3
- 230000002950 deficient Effects 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 238000012217 deletion Methods 0.000 description 2
- 230000037430 deletion Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 230000000295 complement effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 229910044991 metal oxide Inorganic materials 0.000 description 1
- 150000004706 metal oxides Chemical class 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Images
Landscapes
- Image Processing (AREA)
- Facsimile Image Signal Circuits (AREA)
Abstract
The invention discloses a noise eliminator of grid with adaptive dead pixels, comprising a data buffering module, an idneitification control module, a dead pixel compensation module and a noise elimination module. The data buffering module is used for buffering 2xN+1 rows of bel image data, wherein N is more than 2. The idneitification control module is used for judging color types of present pixels and is whether or not a dead pixel, and controlling the dead pixel compensation module and the noise elimination module according to judgement results. The dead pixel compensation module is used for compensating dead pixels of present pixel value according to surrounding homochromatic pixel value when present pixel is a dead pixel, if present pixel is a G color pixel, transmitting present pixel value which is compensated to the noise elimination module, if not, outputting present pixel value which is compensated directly. The noise elimination module is used for receiving compensating value of present pixel sent from the dead pixel compensation module when present pixel is a G color and dead pixel, eliminating noise to compensating value of present pixel according to surrounding G color pixel value, also eliminating noise to present pixel value according to surrounding G color pixel value when present pixel is a G color pixel but not a dead pixel, then outputting the value of which noise is eliminated.
Description
Technical field
The present invention relates to image processing field, relate in particular to bad point self-adapted grid noise elimination device and method.
Technical background
The optical pickocff of digital image pick-ups such as camera, digital camera and Digital Video adopts CCD (Charge-coupled device usually, charge coupled device) or CMOS (Complememary MetalOxide Semiconductor, complementary matal-oxide semiconductor) technology, a two-dimensional matrix of forming by the photo-sensitive cell of both direction dense arrangement (CCD or CMOS) anyhow, and CCD or cmos sensor can only be responded to light luminance, can not respond to color information.Therefore (ColorFiltered Array CFA) guarantees that each sensor pixel only can receive a kind of light of color: a kind of in normally red (R), green (G), blue (B) three kinds of colors must to use the color filter array.The color filter array can use different patterns, and the most frequently used is the color filter array of Bayer (Bel) pattern.The Bayer pattern is used alternatingly one group of red and green filter and one group of green and blue filter, and wherein green pixel point adds up to redness and blue pixel point sum.The line of pixels column format of original (raw) image (hereinafter to be referred as Bel's image) of Bayer pattern as shown in Figure 1.Image for Bel's image transitions is become can normally show need carry out cfa interpolation to it, promptly to each pixel, obtains other two kinds of color values of this pixel by the pixel value around it.For example, [m, n] locates in the position, has only the G value, utilizes the information of point on every side, can obtain the R value and the B value at this some place by interpolation.Through behind the cfa interpolation, can obtain R, G, B value on each pixel.
Because general natural scene image all is smooth, that is to say that color is slow gradual change, therefore, the G value (diagonal angle is adjacent) that Bel's image is adjacent should be very approaching.But, in the production process of digital image pick-up, because the production technology assembling alignment precision relatively poor or camera lens and optical pickocff of optical pickocff is not enough, make the Bayer image that generates at optical pickocff the differ greatly phenomenon of (being G value imbalance) of adjacent G value can occur.And will there be latticed noise in this Bel's image through the image that generates after the cfa interpolation, have a strong impact on picture quality.Fig. 2 A is the normal picture that does not have latticed noise, and Fig. 2 B is the image that has latticed noise.
Generally can adopt dual mode to eliminate latticed noise in the prior art.First kind of mode improved production technology exactly, with the quality of raising optical pickocff and the alignment precision of camera lens and optical pickocff, yet production technology improved often to be needed to drop into a large amount of funds and time, even and production technology carried out improving be difficult to also guarantee that each digital image pick-up of producing can both satisfy quality requirement.Second method is exactly with the mode of image processing Bel's image to be carried out grid noise to eliminate, promptly the G value pixel that each original G value pixel can be adjacent is weighted on average, and with the G value of this weighted average as this pixel, eliminate the unbalanced of G value like this, also just eliminated grid noise.
Yet, be bad point (dead pixel) if the G pixel is arranged on Bel's image, pass through so after the above-mentioned noise cancellation method, the grid noise that is caused by this evil idea point not only can not be eliminated also can spread on the contrary and come.Because go bad point on Bel's original image that above-mentioned noise cancellation method is a hypothesis to be imported, have only the unbalanced defective of G value.But the unbalanced defective of Bel's image of actual input existing G value of possibility also has bad point defect.So just not only need to carry out compensating bad point but also need to carry out the grid noise elimination.As shown in Figure 3, in the prior art can with two independently sequence of modules handle: at first damage a compensating module bad point compensated processing; Carrying out grid noise with the data of noise cancellation module after to compensating bad point then eliminates.So just both bad point can be compensated, latticed noise can be eliminated again.But such implementation, hardware costs are very big.
Please refer to shown in Fig. 4 A, for the compensating bad point module, if desired current pixel point P4 being carried out compensating bad point needs with reference to homochromy pixel P0～P3 and P5～P8 around the reference, since on Bel's original image with colored pixels point (except the G look) level, vertical, to the angular direction pixel of all being separated by, therefore to carry out the block of pixels that compensating bad point needs 5 * 5 sizes at least, and the typical hardware design is handled by row, so the compensating bad point module needs to preserve the view data of 5 row at least.Please refer to shown in Fig. 4 B, for noise cancellation module, owing to be according to adjacent G pixel the G value to be carried out equilibrium, thereby latticed noise eliminated, therefore the block of pixels that needs 3 * 3 sizes at least, so noise cancellation module needs to preserve the view data of 3 row at least.Like this, two modules need the internal memory of 8 row capacity to preserve the data that needs are handled at least, and the resource that needs is very big.
Summary of the invention
In view of this, one object of the present invention is to provide a kind of bad point self-adapted grid noise elimination device, and it can compensate bad point can eliminate grid noise again, and it can save hardware resource again simultaneously.
Another object of the present invention is to provide a kind of bad point self-adapted gridding noise elimination method, and it can compensate bad point can eliminate grid noise again, and it can save hardware resource again simultaneously.
In order to achieve the above object, the bad point self-adapted grid noise elimination device that provides according to an aspect of the present invention, be used for Bel's image is carried out compensating bad point and grid noise elimination, it includes data cache module, identification control module, compensating bad point module and noise cancellation module.Described data cache module is used for the capable Bel's view data of buffer memory 2*N+1, and wherein N is more than or equal to 2.Whether described identification control module is used to judge the color type of current pixel point and is bad point and according to the work of result of determination control compensating bad point module and noise cancellation module.Described compensating bad point module, be used for when current pixel point is bad point, value according to homochromy pixel around it is carried out compensating bad point to the value of current pixel point, if current pixel point is a G colour vegetarian refreshments, the value of the current pixel point after then will compensating flows to noise cancellation module, if current pixel point is not a G colour vegetarian refreshments, the value of the current pixel point after then will compensating is directly exported.Described noise cancellation module, be used for being the G look and receiving during for bad select from the offset of the current pixel point of compensating bad point module and according to the value of its G colour vegetarian refreshments on every side the offset of current pixel point is carried out noise removing in current pixel point, also be used for being the G look and not being according to the G color pixel point value around it value of current pixel point to be carried out noise removing when going bad point, then the value after the noise removing exported in current pixel point.Wherein the G color pixel point value on every side of the homochromy pixel point value on every side of the current pixel point of the value of current pixel point, compensating bad point needs and the current pixel point that noise removing needs all is stored in the data cache module.
Further, described identification control module is judging that current pixel point is R or B look and during for bad point, directly exports the value of current pixel point.
Further, described identification control module memory contains the bad some table of all bad points of marking image transducer.
Further, current pixel point being carried out compensating bad point is specifically as follows:
P4=(a1*P0+a2*P1+a3*P2+a4*P3+a5*P5+a6*P6+a7*P7+a8*P8)/(a1+a2+...+a7+a8), wherein P4 represents current pixel point, P0, P1, P2, P3, P5, P6, P7 and P8 are homochromy pixel around it, a1～a8 is the penalty coefficient of homochromy pixel, can be for more than or equal to 0 real number.
Further, when bad point is arranged among pixel P0, P1, P2, P3, P5, P6, P7 and the P8, the penalty coefficient of this evil idea point is changed to 0.
Further, the value of current pixel point or the offset process of carrying out noise removing is specially:
OUT＝(a*G+b*G
tl+c*G
tr+d*G
bl+e*G
br)/(a+b+c+d+e)，
Wherein G is the offset of current pixel point in current pixel point during for bad point, otherwise it is the initial value of current pixel point, and a, b, c, d, e are filter factor, b, c, d, e, can be for more than or equal to 0 real number, and a is the real number greater than 0, at G
Tl, G
Tr, G
Bl, G
BrIn when bad point is arranged, the filter factor that it is corresponding is changed to 0, OUT is the output valve of current pixel point G.
Further, the pending view data of N+1 behavior in the view data that the 2*N+1 of buffer memory is capable in the described data cache module, described subsequent processes is meant the compensating bad point of the judgement of identification control module, described compensating bad point module and the noise removing of described noise cancellation module, when carrying out above-mentioned processing, with treating image data processing is that the capable pixel of N+1 is handled in order one by one, and current pixel point is exactly the pixel of handling.
A kind of bad point self-adapted gridding noise elimination method that provides according to an aspect of the present invention, be used for Bel's image is carried out compensating bad point and grid noise elimination, it is applied to include on the device of data cache module, this method comprises: steps A, import Bel's view data with behavior unit to data cache module, described data cache module can the capable Bel's view data of buffer memory 2*N+1, described N is more than or equal to 2, pending view data be that N+1 is capable, the pixel of this pending row is handled in order one by one; Step B judges that whether current pixel point is bad point, if be bad point, then changes step C over to; If be not bad point, when the color type of current pixel point is R or B look, the value of current pixel point is directly exported, when being the G look, the color type of current pixel point changes step D over to; Step C, according to homochromy pixel point value around the current pixel point value of current pixel point is carried out compensating bad point, and when current pixel point is G colour vegetarian refreshments, changing step D over to, the current pixel point value after will compensating when current pixel point is not G colour vegetarian refreshments is directly exported; Step D carries out noise removing according to the G color pixel point value around the current pixel point to the offset or the initial value of current pixel point, with the value output of the current pixel point after the noise removing; Wherein the G color pixel point value on every side of the homochromy pixel point value on every side of the current pixel point of the value of current pixel point, compensating bad point needs and the current pixel point that noise removing needs all is stored in the data cache module.
Further, the value to current pixel point among the step C is carried out compensating bad point and is specially:
Suppose that P4 is a current pixel point, so current pixel point carried out compensating bad point and be specifically as follows:
P4=(a1*P0+a2*P1+a3*P2+a4*P3+a5*P5+a6*P6+a7*P7+a8*P8)/(a1+a2+...+a7+a8), wherein pixel P0, P1, P2, P3, P5, P6, P7 and P8 are the homochromy pixel on every side of current pixel point P4, a1～a8 is the penalty coefficient of homochromy pixel, for more than or equal to 0 real number, when bad point is arranged among pixel P0, P1, P2, P3, P5, P6, P7 and the P8, the penalty coefficient of this evil idea point is changed to 0.
Further, value or the offset to current pixel point of the step D process of carrying out noise removing is specially: OUT=(a*G+b*G
Tl+ c*G
Tr+ d*G
Bl+ e*G
Br)/(a+b+c+d+e),
Wherein G is the offset of current pixel point in current pixel point during for bad point, otherwise it is the initial value of current pixel point, and a, b, c, d, e are filter factor, b, c, d, e, can be for more than or equal to 0 real number, and a is the real number greater than 0, at G
Tl, G
Tr, G
Bl, G
BrIn when bad point is arranged, the filter factor that it is corresponding is changed to 0, OUT is the output valve of current pixel point G.
Like this, in the technical scheme that the present invention proposes, compensating bad point is handled and grid noise eliminates that handling walks abreast simultaneously carries out, and they can shared drive, can save hardware resource greatly like this.Filtering does not exert an influence to grid noise in order to guarantee bad point, according to the adaptive change filter factor of bad some situation, has reached good noise removing effect when noise removing.
Description of drawings
Fig. 1 is the line of pixels column format schematic diagram of Bel's image;
Fig. 2 A is the normal picture that does not have latticed noise;
Fig. 2 B is the image that has latticed noise;
Fig. 3 is a functional-block diagram of realizing the device of compensating bad point and grid noise elimination in the prior art simultaneously;
Pel array schematic diagram when Fig. 4 A is compensating bad point;
Fig. 4 B is the pel array schematic diagram that grid noise is eliminated;
Fig. 5 is the structured flowchart of an embodiment of the bad point self-adapted grid noise elimination device of the present invention; With
Fig. 6 is the schematic flow sheet of an embodiment of the bad point self-adapted gridding noise elimination method of the present invention.
Embodiment
Below in conjunction with Figure of description the specific embodiment of the present invention is described.
In the technical scheme that the present invention proposes, compensating bad point and grid noise are eliminated parallel simultaneously carrying out, and they can shared drive, can save hardware resource greatly like this.
Please refer to shown in Figure 5ly, it shows the structured flowchart of an embodiment of the bad point self-adapted grid noise elimination device of the present invention 1.Described bad point self-adapted grid noise elimination device 1 is used for the Bel's image with bad point and the unbalanced defective of G value is carried out compensating bad point and grid noise elimination, and described device 1 includes data cache module 10, identification control module 20, compensating bad point module 30 and noise cancellation module 40.
Described data cache module 10 is used for Bel's mode image data of buffer memory input, because typical hardware is handled by row, and three primary colors pixel pattern of rows and columns in view of Bel's mode image, G colour vegetarian refreshments around homochromy pixel and noise removing needed around compensating bad point needed, so described data cache module 10 needs the capable view data of buffer memory 2*N+1, wherein N is more than or equal to 2, that is to say that described data cache module 10 needs the view data of buffer memory 5 row at least, such as Bel's mode image of 400*600 size, to need the view data of buffer memory be 5*600 to data cache module 10 so.
Wherein, need in the view data that the 2*N+1 of buffer memory is capable in the described data cache module 10 to carry out subsequent processes pending view data be exactly the capable view data of N+1, described subsequent processes can refer to compensating bad point or noise removing processing etc.When handling, to this pending row is that the capable pixel of N+1 carries out subsequent processes in order one by one, for convenience, the pixel that will handle here is called current pixel point, the pixel that will just handle is called last pixel, pixel just to be processed is called next pixel.Handle this N+1 capable after, to described data cache module 10 input delegation view data, with previous the 1st line data deletion, the line number of other each row all subtracts 1, delegation's view data of just having imported is capable as 2*N+1, and then this capable view data of N+1 is handled in continuation.
Whether described identification control module 20 be used for judging its color type and being bad point according to the coordinate of current pixel point, if be R or B look and be not bad point, then the value of current pixel point directly exported; If be R or B look and for bad point, then notify the value of 30 pairs of current pixel point of compensating bad point module to handle; If be G look and, then notify the value of 40 pairs of current pixel point of noise cancellation module to handle for bad point; If be G look and for bad point, then notify the value of 30 pairs of current pixel point of compensating bad point module to handle.Wherein, the bad some table of the bad point of all of the underlined imageing sensor of described identification control module 20 stored, such as being stored in the EEROM (electrically-erasable read-only memory), whether it can mark each pixel with 1 bit data is bad point, such as 1 this pixel of expression is bad point, and 0 this pixel of expression is not a bad point.So whether not bad the concrete decision process of point can be that identification control module 20 is searched the value of relevant position according to the coordinate of current processed pixels point in bad some table, be that the current processed pixels point of 1 expression is a bad point, is that the current processed pixels point of 0 expression is not a bad point.In another embodiment, can only store the coordinate of bad point in the bad some table, when needs judge whether current pixel point is bad point, only need take the coordinate of current pixel point in bad some table, to inquire about, if in the bad some table this coordinate is arranged then illustrates that current pixel point is a bad point, if this coordinate then illustrate that current pixel point is not a bad point not in the bad some table.In addition, can not change,, just can determine the picture look type of current pixel point according to the coordinate of current pixel point as long as therefore know the arrangement model of Bel's mode image because in a single day the three primary colors RGB of Bel's mode image arranges very regular and just decides.
Described compensating bad point module 30 is used for when current pixel point is bad point, according to homochromy pixel point value around it value of current pixel point is carried out compensating bad point, if current pixel point is a G colour vegetarian refreshments, the current pixel point value after the compensation is flowed to noise cancellation module 40; If current pixel point is not a G colour vegetarian refreshments, the current processed pixels point value after the compensation is directly exported, wherein the homochromy pixel point value on every side of the current pixel point of compensating bad point needs all is buffered in the data memory module 10.In a specific embodiment, please refer to shown in Fig. 4 A, suppose that P4 is a current pixel point, so current pixel point being carried out compensating bad point is specifically as follows: P4=(a1*P0+a2*P1+a3*P2+a4*P3+a5*P5+a6*P6+a7*P7+a8*P8)/(a1+a2+...+a7+a8), pixel P0 wherein, P1, P2, P3, P4, P5, P6, P7 and P8 are buffered in the data cache module 10, a1～a8 is the penalty coefficient of a homochromy pixel, can be for more than or equal to 0 real number, the principle of compensation formula is that reference is on every side with the level and smooth definite badly value of a pixel of the value of color pixel.Certainly other any compensation method that can also adopt those of ordinary skill to expect compensates current pixel point.In a preferred embodiment, need simultaneously during compensating bad point to judge whether pixel P0, P1, P2, P3, P5, P6, P7 and P8 around the current pixel point P4 are bad point, if the coefficient that bad point then will be gone bad a little is changed to 0, also we can say the level and smooth definite badly value of a pixel of the value of homochromy non-bad pixel around the reference.
Described noise cancellation module 40 is used for being the G look and receiving during for bad point from the offset of the current pixel point of compensating bad point module 30 and according to the G color pixel point value around it offset of current pixel point is carried out noise removing in current pixel point, also be used for being the G look and according to the G color pixel point value around it value of current pixel point being carried out noise removing during for bad point, then with the value output of the current pixel point after the noise removing in current pixel point.Wherein the G color pixel point value on every side of the current pixel point of noise removing needs all is stored in the data cache module 10.In a specific embodiment, please referring to shown in Fig. 4 B, the G colour vegetarian refreshments around the current pixel point G has G
Tl, G
Tr, G
Bl, G
Br, they all are buffered in the data cache module 10, and the process of the value of current pixel point being carried out noise removing according to the G color pixel point value around the current pixel point G is specially: OUT=(a*G+b*G
Tl+ c*G
Tr+ d*G
Bl+ e*G
Br)/(a+b+c+d+e), wherein the G in the formula is the offset of current pixel point when current pixel point is bad point, otherwise it is the initial value of current pixel point, a, b, c, d, e are filter factor, b, c, d, e, can be for more than or equal to 0 real number, a is the real number greater than 0, at G
Tl, G
Tr, G
Bl, G
BrIn when bad point is arranged, the filter factor that it is corresponding is changed to 0, can change the size of other filter factors simultaneously as required, OUT is the output valve of current pixel point G.
At one more specifically among the embodiment, at G
Tl, G
Tr, G
Bl, G
BrWhen not being at bad, the output valve OUT of this pixel G is so:
out＝(4*G+Gtl+Gtr+Gbl+Gbr)/8；
At G
Tl, G
Tr, G
Bl, G
BrAny one during for bad point,
OUT=(3*G+G
C1+ G
C2+ G
C3)/6, wherein G
C1, G
C2, G
C3Representative is not three pixels of bad point;
Similar, at G
Tl, G
Tr, G
Bl, G
BrIn any two when being bad point,
OUT=(2*G+G
C1+ G
C2)/4, wherein G
C1, G
C2Representative is not two pixels of bad point;
Similar, at G
Tl, G
Tr, G
Bl, G
BrIn any three when being bad point,
OUT=(G+G
C1)/2, wherein G
C1Representative is not a pixel of bad point;
Similar, at G
Tl, G
Tr, G
Bl, G
BrWhen all being at bad,
OUT＝G。
Therefore, when current G colour vegetarian refreshments is carried out noise removing, only consider its non-badly some pixel on every side as can be seen, and ignore the bad some pixel around it, can eliminate badly a influence like this grid noise filtering.In bad point self-adapted grid noise elimination device 100 of the present invention, described noise cancellation module 40 and described compensating bad point module 30 shared data cache modules 10, saved hardware resource greatly, in addition, because compensating bad point and noise removing are carried out simultaneously, filtering does not exert an influence to grid noise in order to guarantee bad point, according to the adaptive change filter factor of bad some situation, has reached good noise removing effect when noise removing.
According to a further aspect in the invention, the invention provides a kind of bad point self-adapted gridding noise elimination method, it can be used for aforementioned bad point self-adapted grid noise elimination device 100, but might not be limited to said apparatus.
Please refer to shown in Figure 6ly, it shows the schematic flow sheet of an embodiment of the bad point self-adapted gridding noise elimination method of the present invention, and it specifically comprises the steps.
Described data cache module can the capable Bel's view data of buffer memory 2*N+1, wherein N is more than or equal to 2, need carry out subsequent processes pending view data be exactly that N+1 is capable, when handling, to this pending row is that the capable pixel of N+1 carries out subsequent processes in order one by one, for convenience, the pixel that will handle here is called current pixel point, the pixel that will just handle is called last pixel, pixel just to be processed is called next pixel.Handle N+1 capable after, view data is capable as 2*N+1 with it to described data cache module input delegation, will previous the 1st line data deletion, line numbers of other each row all subtract 1, and then continue this capable view data of N+1 of processing.
Step 700 is carried out noise removing according to the G color pixel point value around the current pixel point to the offset or the initial value of current pixel point, and step 800 is exported and changed over to the value of the current pixel point after the noise removing.
In a specific embodiment, the value to current pixel point in the step 300 is carried out compensating bad point and be please refer to shown in Fig. 4 A, supposes that P4 is a current pixel point, so current pixel point is carried out compensating bad point and is specifically as follows:
P4＝(a1*P0+a2*P1+a3*P2+a4*P3+a5*P5+a6*P6+a7*P7+a8*P8)/(a1+a2+...+a7+a8)，
Wherein pixel P0, P1, P2, P3, P4, P5, P6, P7 and P8 are buffered in the data cache module 10, a1～a8 is the penalty coefficient of a homochromy pixel, can be for more than or equal to 0 real number, the principle of compensation formula is that reference is on every side with the level and smooth definite badly value of a pixel of the value of color pixel.Certainly other any compensation method that can also adopt those of ordinary skill to expect compensates current pixel point.In a preferred embodiment, need simultaneously during compensating bad point to judge whether pixel P0, P1, P2, P3, P5, P6, P7 and P8 around the current pixel point P4 are bad point, if the coefficient that bad point then will be gone bad a little is changed to 0, also we can say the level and smooth definite badly value of a pixel of the value of homochromy non-bad pixel around the reference.
In a specific embodiment, offset or initial value to current pixel point in the step 700 are carried out noise removing please referring to shown in Fig. 4 B, and the G colour vegetarian refreshments around the current pixel point G has G
Tl, G
Tr, G
Bl, G
Br, they all are buffered in the data cache module 10, and the process of the value of current pixel point being carried out noise removing according to the G color pixel point value around the current pixel point G is specially: OUT=(a*G+b*G
Tl+ c*G
Tr+ d*G
Bl+ e*G
Br)/(a+b+c+d+e), wherein the G in the formula is the offset of current pixel point when current pixel point is bad point, otherwise it is the initial value of current pixel point, a, b, c, d, e are filter factor, b, c, d, e, can be for more than or equal to 0 real number, a is the real number greater than 0, at G
Tl, G
Tr, G
Bl, G
BrIn when bad point is arranged, the filter factor that it is corresponding is changed to 0, can change the size of other filter factors simultaneously as required, OUT is the output valve of current pixel point G.
Wherein the G color pixel point value on every side of the homochromy pixel point value on every side of the current pixel point of the value of current pixel point, compensating bad point needs and the current pixel point that noise removing needs all is stored in the data cache module.
It should be noted that, because bad point self-adapted grid noise elimination device and method has many similarities, therefore be not repeated in this description the content identical in the method part, but this can not hinder us by the content of device part the scheme of method to be understood with installing part.
The above only is preferred embodiment of the present invention, and is in order to restriction the present invention, within the spirit and principles in the present invention not all, any modification of being done, is equal to replacement etc., all should be included within protection scope of the present invention.
Claims (10)
1. a bad point self-adapted grid noise elimination device is used for Bel's image is carried out compensating bad point and grid noise elimination, and it includes data cache module, identification control module, compensating bad point module and noise cancellation module, it is characterized in that,
Described data cache module is used for the capable Bel's view data of buffer memory 2*N+1, and wherein N is more than or equal to 2;
Whether described identification control module is used to judge the color type of current pixel point and is bad point and according to the work of result of determination control compensating bad point module and noise cancellation module;
Described compensating bad point module, be used for when current pixel point is bad point, value according to homochromy pixel around it is carried out compensating bad point to the value of current pixel point, if current pixel point is a G colour vegetarian refreshments, the value of the current pixel point after then will compensating flows to noise cancellation module, if current pixel point is not a G colour vegetarian refreshments, the value of the current pixel point after then will compensating is directly exported;
Described noise cancellation module, be used for being the G look and receiving during for bad select from the offset of the current pixel point of compensating bad point module and according to the value of its G colour vegetarian refreshments on every side the offset of current pixel point is carried out noise removing in current pixel point, also be used for being the G look and not being according to the G color pixel point value around it value of current pixel point to be carried out noise removing when going bad point, then the value after the noise removing exported in current pixel point;
Wherein the G color pixel point value on every side of the homochromy pixel point value on every side of the current pixel point of the value of current pixel point, compensating bad point needs and the current pixel point that noise removing needs all is stored in the data cache module.
2. bad point self-adapted grid noise elimination device as claimed in claim 1 is characterized in that, described identification control module is judging that current pixel point is R or B look and during for bad point, directly exports the value of current pixel point.
3. bad point self-adapted grid noise elimination device as claimed in claim 1 is characterized in that, described identification control module memory contains the bad some table of all bad points of marking image transducer.
4. bad point self-adapted grid noise elimination device as claimed in claim 1 is characterized in that,
Current pixel point is carried out compensating bad point to be specifically as follows:
P4＝(a1*P0+a2*P1+a3*P2+a4*P3+a5*P5+a6*P6+a7*P7+a8*P8)/(a1+a2+...+a7+a8)，
Wherein P4 represents current pixel point, and P0, P1, P2, P3, P5, P6, P7 and P8 are homochromy pixel around it, and a1～a8 is the penalty coefficient of homochromy pixel, can be for more than or equal to 0 real number.
5. bad point self-adapted grid noise elimination device as claimed in claim 4 is characterized in that, when bad point is arranged among pixel P0, P1, P2, P3, P5, P6, P7 and the P8, the penalty coefficient of this evil idea point is changed to 0.
6. bad point self-adapted grid noise elimination device as claimed in claim 1 is characterized in that,
The process that the value of current pixel point or offset carry out noise removing is specially:
OUT＝(a*G+b*G
tl+c*G
tr+d*G
bl+e*G
br)/(a+b+c+d+e)，
Wherein G is the offset of current pixel point in current pixel point during for bad point, otherwise it is the initial value of current pixel point, and a, b, c, d, e are filter factor, b, c, d, e, can be for more than or equal to 0 real number, and a is the real number greater than 0, at G
Tl, G
Tr, G
Bl, G
BrIn when bad point is arranged, the filter factor that it is corresponding is changed to 0, OUT is the output valve of current pixel point G.
7. bad point self-adapted grid noise elimination device as claimed in claim 1, it is characterized in that, the pending view data of N+1 behavior in the view data that the 2*N+1 of buffer memory is capable in the described data cache module, described subsequent processes is meant the compensating bad point of the judgement of identification control module, described compensating bad point module and the noise removing of described noise cancellation module, when carrying out above-mentioned processing, with treating image data processing is that the capable pixel of N+1 is handled in order one by one, and current pixel point is exactly the pixel of handling.
8. a bad point self-adapted gridding noise elimination method is used for Bel's image is carried out compensating bad point and grid noise elimination, and it is applied to include on the device of data cache module, it is characterized in that, this method comprises:
Steps A, to data cache module input Bel view data, described data cache module can the capable Bel's view data of buffer memory 2*N+1 with behavior unit, and described N is more than or equal to 2, pending view data be that N+1 is capable, the pixel of this pending row is handled in order one by one;
Step B judges that whether current pixel point is bad point, if be bad point, then changes step C over to; If be not bad point, when the color type of current pixel point is R or B look, the value of current pixel point is directly exported, when being the G look, the color type of current pixel point changes step D over to;
Step C, according to homochromy pixel point value around the current pixel point value of current pixel point is carried out compensating bad point, and when current pixel point is G colour vegetarian refreshments, changing step D over to, the current pixel point value after will compensating when current pixel point is not G colour vegetarian refreshments is directly exported;
Step D carries out noise removing according to the G color pixel point value around the current pixel point to the offset or the initial value of current pixel point, with the value output of the current pixel point after the noise removing;
Wherein the G color pixel point value on every side of the homochromy pixel point value on every side of the current pixel point of the value of current pixel point, compensating bad point needs and the current pixel point that noise removing needs all is stored in the data cache module.
9. bad point self-adapted gridding noise elimination method as claimed in claim 8 is characterized in that, the value to current pixel point among the step C is carried out compensating bad point and is specially:
Suppose that P4 is a current pixel point, so current pixel point carried out compensating bad point and be specifically as follows:
P4＝(a1*P0+a2*P1+a3*P2+a4*P3+a5*P5+a6*P6+a7*P7+a8*P8)/(a1+a2+...+a7+a8)，
Wherein pixel P0, P1, P2, P3, P5, P6, P7 and P8 are the homochromy pixel on every side of current pixel point P4, a1～a8 is the penalty coefficient of homochromy pixel, for more than or equal to 0 real number, when bad point is arranged among pixel P0, P1, P2, P3, P5, P6, P7 and the P8, the penalty coefficient of this evil idea point is changed to 0.
10. bad point self-adapted gridding noise elimination method as claimed in claim 8 is characterized in that, the process that value or the offset to current pixel point of step D carry out noise removing is specially:
OUT＝(a*G+b*G
tl+c*G
tr+d*G
bl+e*G
br)/(a+b+c+d+e)，
Wherein G is the offset of current pixel point in current pixel point during for bad point, otherwise it is the initial value of current pixel point, and a, b, c, d, e are filter factor, b, c, d, e, can be for more than or equal to 0 real number, and a is the real number greater than 0, at G
Tl, G
Tr, G
Bl, G
BrIn when bad point is arranged, the filter factor that it is corresponding is changed to 0, OUT is the output valve of current pixel point G.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2007101205992A CN100563302C (en) | 2007-08-22 | 2007-08-22 | A kind of bad point self-adapted grid noise elimination device and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2007101205992A CN100563302C (en) | 2007-08-22 | 2007-08-22 | A kind of bad point self-adapted grid noise elimination device and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101150658A true CN101150658A (en) | 2008-03-26 |
| CN100563302C CN100563302C (en) | 2009-11-25 |
Family
ID=39250975
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2007101205992A Expired - Fee Related CN100563302C (en) | 2007-08-22 | 2007-08-22 | A kind of bad point self-adapted grid noise elimination device and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100563302C (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101715050A (en) * | 2009-11-20 | 2010-05-26 | 北京中星微电子有限公司 | Method and system for detecting dead pixels of image sensor |
| CN101754027B (en) * | 2008-12-02 | 2011-09-14 | 华晶科技股份有限公司 | Image compensation method and digital camera device thereof |
| CN104363391A (en) * | 2014-11-28 | 2015-02-18 | 广东欧珀移动通信有限公司 | Image defective pixel compensation method and system and photographing device |
| CN107424131A (en) * | 2017-07-14 | 2017-12-01 | 北京智慧眼科技股份有限公司 | Image based on deep learning removes grid method and device |
| CN107909555A (en) * | 2017-11-27 | 2018-04-13 | 北京大恒图像视觉有限公司 | A kind of gridding noise elimination method for keeping acutance |
| CN108765306A (en) * | 2018-04-17 | 2018-11-06 | 浙江华睿科技有限公司 | A kind of dead pixel points of images antidote, device, electronic equipment and medium |
| CN109740467A (en) * | 2018-12-24 | 2019-05-10 | 北京中安未来科技有限公司 | A kind of electronic certificate recognition methods, apparatus and system |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108234970B (en) * | 2018-01-04 | 2021-05-21 | 努比亚技术有限公司 | Image processing method, terminal and computer readable storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1288916C (en) * | 2004-12-27 | 2006-12-06 | 北京中星微电子有限公司 | Image dead point and noise eliminating method |
| CN100397907C (en) * | 2005-05-11 | 2008-06-25 | 北京中星微电子有限公司 | Bell pattern format based image processing method and apparatus |
| KR100723208B1 (en) * | 2005-07-22 | 2007-05-29 | 삼성전기주식회사 | Image processing apparatus for digital camera, and method thereof |
| CN100499825C (en) * | 2005-09-08 | 2009-06-10 | 凌阳科技股份有限公司 | Method and system for eliminating chromatic noise produced by interpolation |
-
2007
- 2007-08-22 CN CNB2007101205992A patent/CN100563302C/en not_active Expired - Fee Related
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101754027B (en) * | 2008-12-02 | 2011-09-14 | 华晶科技股份有限公司 | Image compensation method and digital camera device thereof |
| CN101715050A (en) * | 2009-11-20 | 2010-05-26 | 北京中星微电子有限公司 | Method and system for detecting dead pixels of image sensor |
| CN101715050B (en) * | 2009-11-20 | 2014-07-09 | 北京中星微电子有限公司 | Method and system for detecting dead pixels of image sensor |
| CN104363391A (en) * | 2014-11-28 | 2015-02-18 | 广东欧珀移动通信有限公司 | Image defective pixel compensation method and system and photographing device |
| CN107424131A (en) * | 2017-07-14 | 2017-12-01 | 北京智慧眼科技股份有限公司 | Image based on deep learning removes grid method and device |
| CN107424131B (en) * | 2017-07-14 | 2019-11-12 | 智慧眼科技股份有限公司 | Image based on deep learning removes grid method and device |
| CN107909555A (en) * | 2017-11-27 | 2018-04-13 | 北京大恒图像视觉有限公司 | A kind of gridding noise elimination method for keeping acutance |
| CN107909555B (en) * | 2017-11-27 | 2020-06-02 | 北京大恒图像视觉有限公司 | Sharpness-keeping grid noise elimination method |
| CN108765306A (en) * | 2018-04-17 | 2018-11-06 | 浙江华睿科技有限公司 | A kind of dead pixel points of images antidote, device, electronic equipment and medium |
| CN109740467A (en) * | 2018-12-24 | 2019-05-10 | 北京中安未来科技有限公司 | A kind of electronic certificate recognition methods, apparatus and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100563302C (en) | 2009-11-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100563302C (en) | A kind of bad point self-adapted grid noise elimination device and method | |
| EP1908273B1 (en) | Pixel binning and averaging based on illumination | |
| CN100366053C (en) | Method for compensating bad dots on digital images | |
| US8681250B2 (en) | Electronic device with two image sensors | |
| CN1988673B (en) | Image processing circuit and image processing method | |
| US8013914B2 (en) | Imaging apparatus including noise suppression circuit | |
| CN101304484B (en) | Image forming apparatus, flaw pixel correction apparatus, processing method of the device | |
| JP5055643B2 (en) | Image pickup device and image pickup apparatus | |
| US8417031B2 (en) | Image processing apparatus and method | |
| CN101272503B (en) | Gridding noise elimination method and device for remaining image definition | |
| US6757012B1 (en) | Color selection for sparse color image reconstruction | |
| CN1692630A (en) | Method and apparatus for real time identification and correction of pixel defects for image sensor arrays | |
| CN1835599A (en) | Method and apparatus for processing a bayer-pattern color digital image signal | |
| CN102959959B (en) | Solid-state imaging device driving method, solid state image pickup device and imaging device | |
| CN102595061A (en) | Image processing device, image processing method, and program | |
| US8922684B2 (en) | Imaging device, control method for imaging device, and storage medium storing a control program for imaging device | |
| US8970748B2 (en) | Imaging device, storage medium storing a control program for imaging device, and control method for imaging device | |
| US20220309712A1 (en) | Application processor including neural processing unit and operating method thereof | |
| CN109285113B (en) | Improved color image interpolation method based on gradient | |
| US20110170776A1 (en) | Image distortion correcting method and image processing apparatus | |
| US8036483B2 (en) | Couplet correction in image sensors | |
| US8938120B2 (en) | Image sensing device and image data processing method using the same | |
| AU2009357162B2 (en) | Determining color information using a binary sensor | |
| JP4464325B2 (en) | Image processing device | |
| CN103647957A (en) | Method for reducing real-time data transmission amount of color image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20091125 Termination date: 20200822 |