CN105430357B - The demosaicing methods and device of imaging sensor - Google Patents
The demosaicing methods and device of imaging sensor Download PDFInfo
- Publication number
- CN105430357B CN105430357B CN201510844323.3A CN201510844323A CN105430357B CN 105430357 B CN105430357 B CN 105430357B CN 201510844323 A CN201510844323 A CN 201510844323A CN 105430357 B CN105430357 B CN 105430357B
- Authority
- CN
- China
- Prior art keywords
- brightness
- point
- estimate
- weight
- signal
- 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.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/10—Cameras or camera modules comprising electronic image sensors; Control thereof for generating image signals from different wavelengths
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Color Television Image Signal Generators (AREA)
Abstract
The invention discloses a kind of demosaicing methods of imaging sensor and device, this method includes：Calculate the brightness initial estimate of current point and neighborhood pixels, calculate current point and gradient of the vicinity points in all directions, the weight in respective direction is obtained, weight and the brightness initial estimate that is obtained are weighted, Strehl ratio low frequency signal estimate is obtained；Pixel and the luminance difference and colour difference of central point in neighborhood are calculated, and luminance difference and colour difference weighting are obtained into Strehl ratio high-frequency signal estimate；Low frequency signal estimate is added with high-frequency signal estimate, Strehl ratio enhancing signal is obtained；The luminance color component signal between the different passages of each point and brightness is obtained using brightness enhancing signal, initial data and weight calculation；Strengthen the Color component signals that signal obtains each point with luminance color component signal using brightness, the present invention can solve the problems such as output image resolution ratio after demosaicing is low, light tone aliasing is serious.
Description
Technical field
The present invention relates to picture signal process field, more particularly to a kind of figure being based on 2 × 2 for the minimum sampling period
As the demosaicing methods and device of sensor.
Background technology
It is most common using imaging sensor that 2x2 is the minimum sampling period as Bayer format, it is on each location of pixels
Only sampling red green blue tricolor in one kind.Sample rate of the three kinds of colors of RGB in Bayer format sensor is respectively 1/
4,1/2 and 1/4.By many color channel images of undressed (RAW) image interpolation of the output of imaging sensor to full resolution
Process is referred to as demosaicing.In general, usual first with green for the interpolation algorithm of Bayer format imaging sensor
The first interpolation of the high characteristic of sample rate goes out the green of full resolution, recycles green interpolation red and blueness.
In actual applications, in order to lift the low light effect of imaging sensor, a green in Bayer format can be led to
Road changes white channel into, and now imaging sensor is RGBW forms.And some application requirement imaging devices can provide quilt simultaneously
Take the photograph the RGB image and infrared image of scene, on existing Bayer format sensor integration a kind of simplest improved method be by
A green filter on the minimum sampling period diagonal of existing 2x2 changes infrared filter into.Now imaging sensor is RGBIR lattice
Formula.Subsequently for convenience of description, it is the minimum sampling period using 2x2 and comprises at least the image of three kinds of colors of RGB by this
Sensor is referred to as RGBX.
And be not difficult to find out, a green channel of Bayer format sensor has been substituted for infrared channel or white is logical
After road, the sample rate of four passages is consistent, is 1/4.So for Bayer format this first interpolation high sampling rate it is green
The demosaicing methods of the chrominance channel red blue channel of interpolation low sampling rate again are obviously no longer desirable for RGBX imaging sensors.
The common interpolating methods such as bilinearity bilinear, bicubic bicubic are respectively acting on RGBIR sensors
Four passages can complete RGBX demosaicing operation, export the image of four full resolutions.But it is this kind of by each passage
The way of independent interpolation does not utilize the information of interchannel completely, and the effect of output image can have resolution ratio after demosaicing
Low, sawtooth is serious, light tone aliasing the problems such as.
Patent No. US6476865 United States Patent (USP)《Sparsely Sampled Image Sensing Device
With Color and Luminance Photosites》In the demosaicing methods mentioned outside interpolation current sampling point
Three components when only judged the direction of respective components and adaptively selected interpolation method on current point diagonal, other two
The interpolation of component can only be fixed using correlation RAW interpolation of datas on horizontally or vertically direction, and this method inevitably results in interpolation
The resolution ratio of output image and the reduction of subjective quality.
Application No. CN201410855583.6 Chinese patent application《The demosaicing methods and dress of imaging sensor
Put》In in the method mentioned interpolation go out customized brightness L, chromatic component C1, C2, C3, then helped point by matrixing
The color output of resolution, this demosaicing methods make use of the information between passage, but its output performance is largely
Dependent on the selection of filter parameter, and the original RAW numbers that the brightness of its definition and chromatic component are exported with imaging sensor
According to and in the absence of direct corresponding relation, therefore the demosaicing methods output of the final output relative ideal after matrixing
There may be the phenomenons such as colour cast, false side.
The content of the invention
To overcome the shortcomings of that above-mentioned prior art is present, the purpose of the present invention is that a kind of imaging sensor of offer goes horse
Gram method and device is matched, it can be good at compatible multiple format imaging sensor, solve output image after demosaicing and differentiate
The problems such as rate is low, light tone aliasing is serious.
In view of the above and other objects, the present invention proposes a kind of demosaicing methods of imaging sensor, including following step
Suddenly：
Step one, the brightness initial estimate of current point and neighborhood pixels is calculated, current point is calculated and exists with vicinity points
Gradient in all directions, obtains the weight in respective direction, and weight and the brightness initial estimate that is obtained are weighted, obtained
To the low frequency signal estimate of Strehl ratio；
Step 2, calculates specific pixel and the luminance difference and colour difference of central point in neighborhood, and by luminance difference and colour difference
Weighting obtains the high-frequency signal estimate of Strehl ratio；
Step 3, the low frequency signal estimate of acquisition is added with high-frequency signal estimate, and the brightness for obtaining central point increases
Strong signal；
Step 4, using brightness strengthen signal, initial data and weight calculation obtain the different passages of each point and brightness it
Between brightness-colour difference signal；
Step 5, the Color component signals that signal obtains pointwise with brightness-colour difference signal are strengthened using brightness.
Further, the brightness initial estimate is done by FIR filter and the initial data of current point surrounding pixel point
Convolution algorithm is obtained, or adaptively by the RAW data of weight and the surrounding pixel point obtained by level, vertical gradient
Ranking operation is obtained.
Further, step one includes：
Convolution algorithm is done according to initial data around predefined interpolated parameter and current point, at the beginning of obtaining the brightness of current point
Beginning estimate, and use same procedure obtains the brightness initial estimate of neighborhood pixels；
Gradient of the current point on level, vertical and tilted direction is calculated, the weight obtained in respective direction of tabling look-up will be weighed
Weight is weighted with brightness initial estimate, obtains the low frequency estimate of Strehl ratio, and obtain neighborhood pixels using same procedure
Brightness low frequency estimate.
Further, step 2 includes：
The brightness low frequency estimation for calculating sampling channel identical with central point RAW in neighborhood respectively is estimated with Strehl ratio low frequency
The difference of meter；
The point and the colour difference of central point of sampling channel identical with central point RAW in neighborhood are calculated respectively；
Luminance difference and colour difference are weighted respectively, then added up by predefined Lagrange factor, Strehl ratio is obtained
High-frequency signal estimate.
Further, in step 4, during the difference Weighted Interpolation for obtaining brightness enhancing signal and RAW data is obtained currently
Brightness-colour difference signal of heart point.
Further, it is poor using the luma-chroma for obtaining brightness enhancing signal with interpolation in step 4 goes out in step 5
Signal subtraction obtains the carrier chrominance signal of all missings of current point.
Further, the acquisition of the weight comprises the following steps：
Four passages are in level, vertical and diagonal 1, the side of diagonal 2 in NxN neighborhoods of the calculating centered on current pixel
Upward partial gradient value；
Grad of the aforementioned four passage on level, vertical, diagonal 1, diagonal 2 is added up respectively and takes average, is obtained
Grad ▽ H, ▽ V, ▽ D1, ▽ D2 on to current pixel neighborhood four direction, send Grad into default look-up table and obtain
Corresponding weight.
To reach above-mentioned purpose, the present invention also provides a kind of demosaicing device of imaging sensor, including：
Low-frequency brightness computing unit, current point and the brightness initial estimation of neighborhood pixels are calculated according to default interpolated parameter,
Gradient of the current point with vicinity points on level, vertical and tilted direction is calculated, and obtains the weight in respective direction, will
Weight is weighted with the brightness initial estimation obtained, obtains the low frequency signal estimate of Strehl ratio；
Brightness strengthens computing unit, specific pixel and the luminance difference and colour difference of central point in neighborhood is calculated, by luminance difference
And colour difference weighting obtains the high-frequency signal estimate of Strehl ratio, and it is low that the brightness high fdrequency component of acquisition is added into brightness
On frequency component, enhanced luminance signal is obtained；
Chrominance distortion unit, strengthens the output of computing unit using the brightness and original RAW data is lacked by interpolation
The carrier chrominance signal of mistake；
Output unit, exports the image in final full width face.
Further, the low-frequency brightness computing unit includes：
Data before original RAW image input block, input interpolation；
Weight calculation unit, weight is then obtained for counting current point and neighborhood territory pixel in the gradient of all directions；
Brightness initial value computing unit；Convolution algorithm is done according to predefined interpolated parameter with RAW data around current point to obtain
To brightness initial estimation；
First weighted accumulation output unit, the power exported by the brightness initial estimate on periphery and by weight calculation unit
Weighting obtains the brightness low frequency component estimation of current point again.
Further, brightness enhancing computing unit further comprises：
Luminance difference computing unit, passes through luminance difference meter by point brightness low frequency estimation around and the estimation of current point brightness low frequency
Calculate unit and try to achieve luminance difference；
Colour difference computing unit, the point of sampling channel identical with central point RAW in neighborhood and central point RAW data are passed through
Colour difference computing unit tries to achieve colour difference；
The weight of point around weight calculation unit, output；
Second weighted accumulation output unit, luminance difference, colour difference and weight is weighted the brightness increasing for obtaining current point
Estimate after strong.
Compared with prior art, the demosaicing methods and device of a kind of imaging sensor of the invention introduce luminance channel
Concept, by the way that brightness is divided into low frequency and high fdrequency component, first interpolation low frequency component, then with neighbor information estimated brightness high frequency
Component, interpolation chromatic component is removed using luminance component and RAW data, solves output image resolution ratio after demosaicing low, bright
The problems such as color contamination is folded, the output of the demosaicing device of the present invention can't change the RAW data sampled, will not be extra
Colour cast problem is introduced, and being capable of compatible multiple format imaging sensor.
Brief description of the drawings
Fig. 1 is a kind of step flow chart of the demosaicing methods of imaging sensor of the invention；
Fig. 2 is a kind of flow chart of the specific embodiment of the demosaicing methods of imaging sensor of the invention；
Fig. 3 is the gradient direction schematic diagram of 5x5 neighborhoods centered on X in present pre-ferred embodiments；
Fig. 4 is a kind of structural representation of the demosaicing device of imaging sensor of the invention；
Fig. 5 is the structural representation of low-frequency brightness computing unit 400 in present pre-ferred embodiments；
Fig. 6 is the structural representation of brightness enhancing computing unit 401 in present pre-ferred embodiments；
Fig. 7 provides the detailed construction schematic diagram of the demosaicing device specific embodiment of imaging sensor for the present invention.
Embodiment
Below by way of specific instantiation and embodiments of the present invention are described with reference to the drawings, those skilled in the art can
Understand the further advantage and effect of the present invention easily by content disclosed in the present specification.The present invention also can by it is other not
Same instantiation is implemented or applied, and the various details in this specification can also not carried on the back based on different viewpoints and application
Various modifications and change are carried out under spirit from the present invention.
Fig. 1 is a kind of step flow chart of the demosaicing methods of imaging sensor of the invention.As shown in figure 1, of the invention
A kind of demosaicing methods of imaging sensor, comprise the following steps：
Step 101, current point and the brightness initial estimation of neighborhood pixels are calculated according to default interpolated parameter, calculates current point
With gradient of the vicinity points on level, vertical and tilted direction, and the weight in respective direction is obtained, by weight with being obtained
The brightness initial estimation obtained is weighted, and obtains the low frequency signal estimate of Strehl ratio.
Assume initially that there is monochrome information L, it can be expressed as the shape of brightness low frequency component and brightness high fdrequency component sum
Formula, coordinate is shown in the monochrome information such as formula (1) on the point of (i, j).
L (i, j)=Ll(i,j)+Lh(i,j) (1)
Wherein, low frequency component Ll(i, j) can use the original intensity estimate L' of current point surrounding pixel pointnWeighting put down
Come approximately to obtain.
Ll(i,j)≈ΣwnL'n (2)
In the present invention, above-mentioned original intensity estimate can pass through one 2 dimension FIR filter and current point surrounding pixel
The RAW data of point are done convolution algorithm and obtained, and can also adaptively pass through the weight obtained by level, vertical gradient and week
The RAW data weighting computings for enclosing pixel are obtained.N be centered on current point coordinates (i, j) in mxm neighborhoods with central point RAW
The pixel coordinate of data same channels.
Above-mentioned normalized weight wnCan by mxm neighborhoods with original intensity estimate L'nThe gradient of corresponding all directions is obtained
Arrive, can also be by illumination estimate value L'nObtained with the similarity degree of central point initial estimate, the present invention is not limited.
Step 102, specific pixel and the luminance difference and colour difference of central point in neighborhood are calculated, and by luminance difference and colour difference
Weighting obtains the high-frequency signal estimate of Strehl ratio.
The difference for defining neighboring pixel signal and central point signal is difn(i, j), its weighted results Σ wndifn(i, j) can
To be equivalent to the high fdrequency component of current point, then the brightness high fdrequency component L of current pointh(i, j) can be usedWithWeighting sum it is approximate, point RAW data paths centered on wherein C may be RAW numbers according to the difference of coordinate
In in all paths it is any all the way.
Step 103, the low frequency signal estimate of acquisition is added with high-frequency signal estimate, the brightness for obtaining central point increases
Strong signal.
The estimate of final luminance signal can be write as (brightness enhancing signal)：
Step 104, using brightness strengthen signal, original (RAW) data and weight calculation obtain the different passages of each point with
Brightness-colour difference signal between brightness.
Obtain after luminance signal estimation, starting with RAW data around, (neighborhood is) with each self-corresponding brightness it
Brightness-colour difference signal of poor interpolation current point.
It is above-mentionedIt can be obtained using the gradient of periphery RAW data by tabling look-up
Step 105, it is final to strengthen the Color component signals that signal obtains pointwise with brightness-colour difference signal using brightness.
Specifically, after obtaining brightness-colour difference signal, subtracted each other using obtained brightness enhancing signal is calculated before with it
Obtain the Color component signals outside original (RAW) the data sampling path of current point.
Herein it should be noted that, the demosaicing methods of the present invention are applicable not only to include by the minimum sampling period of 2x2
RGB trichromatic RGBIR, RGBW or Bayer format imaging sensor, while being also applied for any adopting by minimum of 2x2
Sample cycle, and the imaging sensor in each minimum sampling period in the presence of two, three or four different filters, such as
CYMG imaging sensors.
Fig. 2 is a kind of flow chart of the specific embodiment of the demosaicing methods of imaging sensor of the invention.It is specific at this
In embodiment, the demosaicing methods of the imaging sensor of the present invention comprise the following steps：
A convolution algorithm) is done according to RAW data around predefined interpolated parameter and current point, the brightness of current point is obtained
Initial estimation.
B) similar step A, obtains the brightness initial estimation of neighborhood pixels.
C gradient of the current point on level, vertical and tilted direction) is calculated, the weight obtained in respective direction of tabling look-up will
Weight is weighted with brightness initial estimation, obtains the low frequency estimation of Strehl ratio.
D) similar step C, obtains the brightness low frequency estimation of neighborhood pixels.
E the estimation of brightness low frequency and the Strehl ratio low frequency of sampling channel identical with central point RAW in neighborhood) are calculated respectively
The difference of estimation.
F the point of sampling channel identical with central point RAW and central point colour difference (RAW data) in neighborhood) are calculated respectively.
G) luminance difference in step E and step F is weighted respectively, colour difference weighting, then by predefined Lagrange because
Son is cumulative, obtains Strehl ratio high-frequency signal estimate.The weight can be obtained by method in similar step C.
H) estimation of brightness high fdrequency component is added with the estimation of brightness low frequency signal, obtains Strehl ratio enhancing signal.
I the carrier chrominance signal for) assuming Current central point missing is on C1, C2, C3, the position that C1 pixels are sampled in neighborhood
Current central is obtained using the difference Weighted Interpolation of highlight signal and the original C1 data of the counted respective point of A-H steps to light
Degree-C1 is poor.Similarly, Strehl ratio can be obtained and the carrier chrominance signal of all missings is poor.
J) the 3 luma-chroma difference signals gone out using interpolation in brightness enhancement value and last step, which are subtracted each other, obtains current point
The carrier chrominance signal of all missings.
Weight in above-mentioned steps C may be referred to Fig. 3, calculates obtain as follows：
1) four passages are calculated in 5x5 neighborhoods centered on current pixel in level (H), vertical (V) and diagonal
Partial gradient value on (D1, D2) direction.
▽ Hg=(abs (G24-G22)+abs(G44-G42))/2 ▽ Vg=(abs (G22-G42)+abs(G24-G44))/2
▽ D1g=abs (G22-G44)
▽ D2g=abs (G24-G42)
▽ Hr=(abs (R12-R14)+abs(R32-R34)+abs(R52-R54))/3
▽ Vr=(abs (R12-R32)+abs(R32-R52)
+abs(R14-R34)+abs(R34-R54))/4
▽ D1r=(abs (R12-R34)+abs(R32-R54))/2 ▽ D2r=(abs (R14-R32)+abs(R34-R52))/2
▽ Hir=(abs (X11-X13)+abs(X15-X13)
+abs(X31-X33)+abs(X35-X33)
+abs(X51-X53)+abs(X55-X53))/6
▽ Vir=(abs (X11-X31)+abs(X51-X31)
+abs(X13-X33)+abs(X53-X33)
+abs(X15-X35)+abs(X55-X35))/6
▽ D1ir=(abs (X11-X33)+abs(X55-X33))/2 ▽ D2ir=(abs (X15-X33)+abs(X51-X33))/2
▽ Hb=(abs (B21-B23)+abs(B25-B23)
+abs(B41-B43)+abs(B45-B43))/4
▽ Vb=(abs (B21-B41)+abs(B23-B43)+abs(B25-B45))/3
▽ D1b=(abs (B21-B43)+abs(B23-B45))/2
▽ D2b=(abs (B23-B41)+abs(B25-B43))/2,
Wherein abs represents the computing that takes absolute value, and ▽ represents gradient, and ▽ Hg then represent the gradient of g passages in the horizontal direction
(the gradient method for expressing of rest channels in different directions is similar therewith)；G24Etc. representing pixel value in 5x5 neighborhoods.
It should be noted that when this specific embodiment is using central point as IR 5x5 neighborhoods gradient calculation method.Central point
Neighborhood gradient calculation method is similar with the above method during for other elements, will not be described here.
2) Grad of the aforementioned four passage on level, vertical, diagonal 1, diagonal 2 is added up respectively and takes average,
Obtain Grad ▽ H, ▽ V, ▽ D1, the ▽ D2 on current pixel neighborhood four direction.The default look-up table of Grad feeding is obtained
To corresponding weight.
Fig. 4 is a kind of structural representation of the demosaicing device of imaging sensor of the invention.As shown in figure 4, of the invention
A kind of demosaicing device of imaging sensor, including：Low-frequency brightness computing unit 400, brightness enhancing computing unit 401, color
Spend computing unit 402 and output unit 403.
Wherein, low-frequency brightness computing unit 400 is according at the beginning of the brightness that default interpolated parameter calculates current point and neighborhood pixels
Begin to estimate, calculate gradient of the current point with vicinity points on level, vertical and tilted direction, and obtain in respective direction
Weight, weight and the brightness initial estimation that is obtained are weighted, the low frequency signal estimate of Strehl ratio is obtained；It is bright
Degree enhancing computing unit 401, calculates specific pixel and the luminance difference and colour difference of central point in neighborhood, by luminance difference and colour difference
Weighting obtains the high-frequency signal estimate of Strehl ratio, and the brightness high fdrequency component of acquisition is added into brightness low frequency component
On, obtain enhanced luminance signal；Chrominance distortion unit 402 strengthens the output of computing unit 401 and original RAW numbers using brightness
According to the carrier chrominance signal lacked by interpolation, and export by output unit 403 image in final full width face.
Fig. 5 is the structural representation of low-frequency brightness computing unit 400 in present pre-ferred embodiments.Wherein original RAW figures
The data inputted as input block 500 before interpolation.Weight calculation unit 501 is used to count current point with neighborhood territory pixel in all directions
Gradient then obtain weight, brightness initial value computing unit 502 is according to RAW data around predefined interpolated parameter and current point
Do convolution algorithm and obtain brightness initial estimation, the weight that the brightness initial estimate on periphery is exported by weight calculation unit 501
The brightness low frequency component estimation of current point is obtained by the first weighted accumulation output unit 503.
Fig. 6 is the structural representation of brightness enhancing computing unit 401 in present pre-ferred embodiments, and low-frequency brightness estimation is defeated
Enter the input current point of unit 601 and the brightness low frequency component of surrounding point is estimated, the brightness low frequency estimation of surrounding point and current point brightness
Low frequency estimation tries to achieve luminance difference by luminance difference computing unit 603, and the point of the interior sampling channel identical with central point RAW of neighborhood is with
Heart point RAW data try to achieve the weight of point around colour difference, the output of weight calculation unit 602, warp by colour difference computing unit 604
Cross the enhanced estimate of brightness that the second weighted accumulation output unit 605 obtains current point.
Fig. 7 provides the detailed construction schematic diagram of the demosaicing device specific embodiment of imaging sensor for the present invention.Its
In data before original RAW image input block 701 input interpolation.Luma-chroma difference computing unit 703 is tired using the second weighting
Plus the Initial R AW data highlighted in signal and neighborhood of current point and the surrounding point obtained in output unit 605 ask poor
The luma-chroma around put is poor.Weight calculation unit 704 be used for count current point and neighborhood all directions gradient then
Obtain weight, be weighted calculating after cumulative output unit 705 obtain current point missing passage luma-chroma it is poor.By in
The luma-chroma difference that brightness enhancing estimation and the current point of heart point lack passage is exported after interpolation output unit 706 seeks difference
Obtain the colourity that current point lacks passage.
In summary, the demosaicing methods and device of a kind of imaging sensor of the invention introduce the general of luminance channel
Read, by the way that brightness is divided into low frequency and high fdrequency component, first interpolation low frequency component, then with the high frequency division of neighbor information estimated brightness
Amount, interpolation chromatic component is removed using luminance component and RAW data, solves that output image resolution ratio after demosaicing is low, light tone
The problems such as aliasing, the output of the demosaicing device of the present invention can't change the RAW data sampled, will not additionally draw
Enter colour cast problem, and being capable of compatible multiple format imaging sensor.
The above-described embodiments merely illustrate the principles and effects of the present invention, not for the limitation present invention.Any
Art personnel can be modified above-described embodiment and changed under the spirit and scope without prejudice to the present invention.Therefore,
The scope of the present invention, should be as listed by claims.
Claims (10)
1. a kind of demosaicing methods of imaging sensor, comprise the following steps：
Step one, the brightness initial estimate of current point and neighborhood pixels is calculated, current point is calculated with vicinity points in each side
Upward gradient, obtains the weight in respective direction, weight and the brightness initial estimate that is obtained is weighted, in obtaining
The low frequency signal estimate of heart point brightness；
Step 2, calculates specific pixel and the luminance difference and colour difference of central point in neighborhood, and luminance difference and colour difference are weighted
Obtain the high-frequency signal estimate of Strehl ratio；
Step 3, the low frequency signal estimate of acquisition is added with high-frequency signal estimate, obtains the brightness enhancing letter of central point
Number；
Step 4, strengthens signal, initial data and weight calculation using brightness and obtains between the different passages of each point and brightness
Brightness-colour difference signal；
Step 5, the Color component signals that signal obtains pointwise with brightness-colour difference signal are strengthened using brightness.
2. a kind of demosaicing methods of imaging sensor as claimed in claim 1, it is characterised in that：The brightness initial estimation
Value is done convolution algorithm with the RAW data of current point surrounding pixel point by FIR filter and obtained, or adaptively by by water
The RAW data weighting computings of the flat, weight that vertical gradient is obtained and surrounding pixel point are obtained.
3. a kind of demosaicing methods of imaging sensor as claimed in claim 2, it is characterised in that step one includes：
Convolution algorithm is done according to initial data around predefined interpolated parameter and current point, the brightness for obtaining current point is initially estimated
Evaluation, and use same procedure obtains the brightness initial estimate of neighborhood pixels；
Calculate gradient of the current point on level, vertical and tilted direction, the weight obtained in respective direction of tabling look-up, by weight with
Brightness initial estimate is weighted, and obtains the low frequency estimate of Strehl ratio, and obtain the bright of neighborhood pixels using same procedure
Spend low frequency estimate.
4. a kind of demosaicing methods of imaging sensor as claimed in claim 2, it is characterised in that step 2 includes：
The brightness low frequency estimation for calculating sampling channel identical with central point RAW in neighborhood respectively estimates it with Strehl ratio low frequency
Difference；
The point and the colour difference of central point of sampling channel identical with central point RAW in neighborhood are calculated respectively；
Luminance difference and colour difference are weighted respectively, then added up by predefined Lagrange factor, Strehl ratio high frequency is obtained
Signal estimate.
5. a kind of demosaicing methods of imaging sensor as claimed in claim 4, it is characterised in that in step 4, will
Acquisition brightness enhancing signal and the difference Weighted Interpolation of initial data obtain brightness-colour difference signal of Current central point.
6. a kind of demosaicing methods of imaging sensor as claimed in claim 5, it is characterised in that：In step 5, profit
Brightness-the colour difference signal gone out with acquisition brightness enhancing signal with interpolation in step 4 subtracts each other the colourity for obtaining all missings of current point
Signal.
7. a kind of demosaicing methods of imaging sensor as claimed in claim 1, it is characterised in that the acquisition of the weight
Comprise the following steps：
Four passages are on level, vertical and diagonal 1, the direction of diagonal 2 in NxN neighborhoods of the calculating centered on current pixel
Partial gradient value；
Grad of the aforementioned four passage on level, vertical, diagonal 1, diagonal 2 is added up respectively and takes average, is worked as
Grad ▽ H, ▽ V, ▽ D1, ▽ D2 on preceding pixel neighborhood four direction, send Grad into default look-up table and obtain correspondence
Weight.
8. a kind of demosaicing device of imaging sensor, including：
Low-frequency brightness computing unit, current point and the brightness initial estimation of neighborhood pixels are calculated according to default interpolated parameter, are calculated
Gradient of the current point with vicinity points on level, vertical and tilted direction, and the weight in respective direction is obtained, by weight
It is weighted with the brightness initial estimation that is obtained, obtains the low frequency signal estimate of Strehl ratio；
Brightness strengthens computing unit, specific pixel and the luminance difference and colour difference of central point in neighborhood is calculated, by luminance difference and color
Degree difference weighting obtains the high-frequency signal estimate of Strehl ratio, and the brightness high fdrequency component of acquisition is added into brightness low frequency point
In amount, enhanced luminance signal is obtained；
Chrominance distortion unit, what the output for strengthening computing unit using the brightness was lacked with original RAW data by interpolation
Carrier chrominance signal；
Output unit, exports the image in final full width face.
9. the demosaicing device of a kind of imaging sensor as claimed in claim 8, it is characterised in that the low-frequency brightness is calculated
Unit includes：
Data before original RAW image input block, input interpolation；
Weight calculation unit, weight is then obtained for counting current point and neighborhood territory pixel in the gradient of all directions；
Brightness initial value computing unit；Convolution algorithm is done with RAW data around current point obtain bright according to predefined interpolated parameter
Spend initial estimation；
First weighted accumulation output unit, the weight exported by the brightness initial estimate on periphery and by weight calculation unit adds
Power obtains the brightness low frequency component estimation of current point.
10. a kind of demosaicing device of imaging sensor as claimed in claim 9, it is characterised in that brightness enhancing meter
Unit is calculated to further comprise：
Luminance difference computing unit, passes through luminance difference computing unit by point brightness low frequency estimation around and the estimation of current point brightness low frequency
Try to achieve luminance difference；
Colour difference computing unit, passes through colourity by the point of sampling channel identical with central point RAW in neighborhood and central point RAW data
Poor computing unit tries to achieve colour difference；
The weight of point around weight calculation unit, output；
Second weighted accumulation output unit, luminance difference, colour difference and weight is weighted after the brightness enhancing for obtaining current point
Estimate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510844323.3A CN105430357B (en) | 2015-11-26 | 2015-11-26 | The demosaicing methods and device of imaging sensor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510844323.3A CN105430357B (en) | 2015-11-26 | 2015-11-26 | The demosaicing methods and device of imaging sensor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105430357A CN105430357A (en) | 2016-03-23 |
| CN105430357B true CN105430357B (en) | 2017-07-18 |
Family
ID=55508256
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510844323.3A Active CN105430357B (en) | 2015-11-26 | 2015-11-26 | The demosaicing methods and device of imaging sensor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105430357B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR3078852B1 (en) * | 2018-03-07 | 2020-04-03 | Centre National De La Recherche Scientifique (Cnrs) | METHOD FOR RECONSTRUCTING A COLOR IMAGE ACQUIRED BY A SENSOR COVERED WITH A MOSAIC OF COLOR FILTERS |
| CN111724321B (en) * | 2020-06-19 | 2023-04-28 | 上海富瀚微电子股份有限公司 | Image filtering method and system |
| CN111784603B (en) * | 2020-06-29 | 2024-01-26 | 珠海全志科技股份有限公司 | RAW domain image denoising method, computer device and computer readable storage medium |
| KR102739817B1 (en) | 2020-07-17 | 2024-12-09 | 에스케이하이닉스 주식회사 | Demosaic operation circuit, image sensing device and operation method thereof |
| CN112104847B (en) * | 2020-09-17 | 2021-07-23 | 北京理工大学 | A Color Reconstruction Method of SONY-RGBW Array Based on Residual and High Frequency Replacement |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7006686B2 (en) * | 2001-07-18 | 2006-02-28 | Hewlett-Packard Development Company, L.P. | Image mosaic data reconstruction |
| US7071978B2 (en) * | 2001-07-18 | 2006-07-04 | Hewlett-Packard Development Company, L.P. | Image mosaic data reconstruction |
| KR101081000B1 (en) * | 2003-10-23 | 2011-11-09 | 소니 가부시키가이샤 | Image processing apparatus and image processing method, and recording medium |
| JP5045421B2 (en) * | 2007-12-21 | 2012-10-10 | ソニー株式会社 | Imaging apparatus, color noise reduction method, and color noise reduction program |
| FR2964490B1 (en) * | 2010-09-06 | 2012-09-28 | Commissariat Energie Atomique | METHOD FOR DEMOSAICING DIGITAL RAW IMAGE, COMPUTER PROGRAM, AND CORRESPONDING IMAGING OR GRAPHICS CIRCUIT |
-
2015
- 2015-11-26 CN CN201510844323.3A patent/CN105430357B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN105430357A (en) | 2016-03-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1812592B (en) | Method and apparatus for processing image data of a color filter array | |
| CN105430357B (en) | The demosaicing methods and device of imaging sensor | |
| JP5118047B2 (en) | System and method for high performance color filter mosaic array | |
| CN102870417B (en) | Color imaging device | |
| CN103905802B (en) | A kind of demosaicing methods based on P pattern colour filter array and device | |
| CN106504281B (en) | Image quality enhancing and filtering method applied to cmos image sensor | |
| EP1484743A2 (en) | System for improving display resolution | |
| CN110430403B (en) | Image processing method and device | |
| CN111510691B (en) | Color interpolation method and device, equipment and storage medium | |
| CN101867832B (en) | Apparatus and method for determining threshold value | |
| TW201618034A (en) | Deinterlacing interlaced high dynamic range images by using YUV interpolation | |
| CN107967668A (en) | A kind of image processing method and device | |
| EP1151617A1 (en) | Color image pickup device | |
| CN108122201A (en) | A kind of Bayer interpolation slide fastener effect minimizing technology | |
| CN104717474A (en) | Image processing method and module | |
| CN104680484B (en) | A kind of method and device of image enhaucament | |
| TW202021330A (en) | Image zooming method and device | |
| CN108307162B (en) | Efficient and flexible color processor | |
| CN101356830A (en) | image processing device | |
| CN1964500A (en) | An interpolation method and device for color filter array | |
| CN103747222B (en) | A kind of method that color artifact of image suppresses and device | |
| CN112422940A (en) | Self-adaptive color correction method | |
| US20150365643A1 (en) | Method of color processing using a color and white filter array | |
| CN103259960B (en) | The interpolation method of data and device, image output method and device | |
| CN117274060B (en) | Unsupervised end-to-end demosaicing method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |