CN105678718A - Method and device for image denoising - Google Patents
Method and device for image denoising Download PDFInfo
- Publication number
- CN105678718A CN105678718A CN201610186820.3A CN201610186820A CN105678718A CN 105678718 A CN105678718 A CN 105678718A CN 201610186820 A CN201610186820 A CN 201610186820A CN 105678718 A CN105678718 A CN 105678718A
- Authority
- CN
- China
- Prior art keywords
- pixel
- block
- weight
- described reference
- reference block
- 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
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000004364 calculation method Methods 0.000 claims abstract description 17
- 230000001105 regulatory effect Effects 0.000 claims description 57
- 230000000694 effects Effects 0.000 abstract description 15
- 238000005259 measurement Methods 0.000 abstract description 7
- 239000000654 additive Substances 0.000 description 8
- 230000000996 additive effect Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 230000000007 visual effect Effects 0.000 description 5
- 239000006185 dispersion Substances 0.000 description 4
- 230000006870 function Effects 0.000 description 2
- 238000012886 linear function Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 229910044991 metal oxide Inorganic materials 0.000 description 1
- 150000004706 metal oxides Chemical class 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 230000009711 regulatory function Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a device and a method of image denoising. The device comprises a selection module, a definition module, a first calculation module, and a second calculation module, wherein the selection module is used for selecting one pixel in an image as a target pixel, and pixels with the same color channel as the target pixel in a set range around the target pixel are selected as reference pixels; the definition module is used for defining a block with a set size with the target pixel as the center as a target block and blocks with a set size with the reference pixels as the center as reference blocks, and each reference pixel is corresponding to each reference block; the first calculation module is used for difference measurement on two kinds of blocks according to all pixels in the reference blocks and the target block and calculating weights of the reference pixels in the center of the reference blocks based on the difference of the two kinds of blocks; and the second calculation module is used for solving a weighted mean value on the target pixel and all reference pixels according to the weights of the target pixel and the reference pixels and the weighted mean value serves as a denoising result for the target pixel. The denoising result keeps the structural information of the original image, and the denoising effects of the image are enhanced.
Description
Technical field
The present invention relates to technical field of image processing, especially relate to a kind of image de-noising method and device.
Background technology
In prior art, major part imageing sensor such as complementary metal oxide semiconductors (CMOS) (CMOS, ComplementaryMetalOxideSemiconductor) when obtaining image, first obtain bayer data (i.e. the image of bayer form), then the method recycling interpolation obtains the data (namely image being carried out demosaicing process) of disappearance, finally gives complete RGB image. But, if image not being carried out denoising before interpolation, the RGB image finally given certainly will have substantial amounts of chromatic noise, and the removal one of chromatic noise is relatively difficult, two to carry out amount of calculation bigger, it is therefore desirable to first view data was carried out denoising before interpolation.
The substantially flow process of traditional image de-noising method is: first the three-channel pixel of R, G, B in the image of bayer form extracted respectively, the image that each composition one is secondary independent, and independent image is carried out denoising; Then denoising result is arranged by original picture format, obtain the view data after denoising. This denoising mode, owing to the separately performed denoising of R, G, B, have ignored the relation between R, G, B triple channel, it is easy to the denoising degree causing adjacent R, G, B is different, affects denoising effect. As shown in Figure 1, if the image of original bayer form has a very thin straight line, when adopting existing image de-noising method to carry out denoising, owing to have ignored the relation between R, G, the denoising degree causing R and G is different, the result images after denoising be likely to show as straight line is unsmooth or after interpolation color impure.
In sum, existing image de-noising method, image restoring degree is relatively low, and denoising effect is poor, thus have impact on final picture quality.
Summary of the invention
The main purpose of the embodiment of the present invention is in that to provide a kind of image de-noising method and device, it is intended to improving image restoring degree in image denoising process, promote the denoising effect of image.
To achieve these objectives, propose a kind of image denoising device, each pixel in described device traversing graph picture on the one hand, respectively each pixel is carried out denoising, including:
Choosing module, the pixel being used for choosing in described image, as object pixel, chooses the pixel identical with the Color Channel of described object pixel as reference pixel in the set point around described object pixel;
Definition module, is object block for the block being sized defined centered by described object pixel, and the block that definition is sized described in centered by described reference pixel is reference block, the corresponding reference block of each reference pixel;
First computing module, for described reference block and described object block being carried out Diversity measure according to described reference block with all pixels in described object block, weight based on described reference block with the described reference pixel at the described reference block center of diversity factor calculating of described object block, obtain the weight of each described reference pixel accordingly, and the weight assignment 1 to described object pixel, wherein said diversity factor and described weight negative correlation;
Second computing module, for seeking weighted mean according to the weight of described object pixel and described reference pixel to described object pixel and all described reference pixels, using the described weighted mean denoising result as described object pixel.
Further, described first computing module includes Diversity measure unit, and described Diversity measure unit is used for: the absolute value sum of the difference pixel of correspondence position in described reference block and described object block subtracted each other is as the diversity factor of described reference block Yu described object block.
Further, described first computing module includes weight calculation unit, and described weight calculation unit is used for:
According to formula w=exp (-r2/sig2) calculate the weight of the described reference pixel at described reference block center, wherein, r is the diversity factor of described reference block and described object block, and sig is empirical, and w is the weight of the described reference pixel at described reference block center.
Further, described first computing module also includes weight regulon, and described weight regulon is used for:
The absolute value of the difference described reference pixel at described reference block and described object block center and described object pixel subtracted each other is as the diversity factor regulated value of described reference block Yu described object block, the weight regulated value of the described reference pixel at described reference block center is calculated, using the weight of described reference pixel and the product of the weight regulated value of described reference pixel as the final weight of described reference pixel based on the diversity factor regulated value of described reference block and described object block.
Further, described weight regulon is used for:
According to formula ww=exp (-rr2/sig_p2) calculate the weight regulated value of the described reference pixel at described reference block center, wherein, rr is the diversity factor regulated value of described reference block and described object block, and sig_p is empirical, and ww is the weight regulated value of the described reference pixel at described reference block center.
A kind of image de-noising method is proposed on the other hand, including step:
Choose a pixel in image as object pixel, choose the pixel identical with the Color Channel of described object pixel in the set point around described object pixel as reference pixel;
The definition block being sized centered by described object pixel is object block, and the block that definition is sized described in centered by described reference pixel is reference block, the corresponding reference block of each reference pixel;
With all pixels in described object block, described reference block and described object block are carried out Diversity measure according to described reference block, weight based on described reference block with the described reference pixel at the described reference block center of diversity factor calculating of described object block, obtain the weight of each described reference pixel accordingly, and the weight assignment 1 to described object pixel, wherein said diversity factor and described weight negative correlation;
Described object pixel and all described reference pixels are sought weighted mean by the weight according to described object pixel and described reference pixel, using the described weighted mean denoising result as described object pixel;
Each pixel traveled through in described image repeats above-mentioned steps.
Further, described according to all pixels in described reference block and described object block, described reference block and described object block carried out Diversity measure and include:
The absolute value sum of the difference pixel of correspondence position in described reference block and described object block subtracted each other is as the diversity factor of described reference block Yu described object block.
Further, the described weight based on described reference block with the described reference pixel at the described reference block center of diversity factor calculating of described object block includes:
According to formula w=exp (-r2/sig2) calculate the weight of the described reference pixel at described reference block center, wherein, r is the diversity factor of described reference block and described object block, and sig is empirical, and w is the weight of the described reference pixel at described reference block center.
Further, also include after the step of the weight of the described reference pixel at the described described reference block center of diversity factor calculating according to described reference block and described object block:
The absolute value of the difference described reference pixel at described reference block and described object block center and described object pixel subtracted each other is as the diversity factor regulated value of described reference block Yu described object block, the weight regulated value of the described reference pixel at described reference block center is calculated, using the weight of described reference pixel and the product of the weight regulated value of described reference pixel as the final weight of described reference pixel based on the diversity factor regulated value of described reference block and described object block.
Further, the described weight regulated value based on described reference block with the described reference pixel at the described reference block center of diversity factor regulated value calculating of described object block includes:
According to formula ww=exp (-rr2/sig_p2) calculate the weight regulated value of the described reference pixel at described reference block center, wherein, rr is the diversity factor regulated value of described reference block and described object block, and sig_p is empirical, and ww is the weight regulated value of the described reference pixel at described reference block center.
A kind of image de-noising method that the embodiment of the present invention provides, by carrying out Diversity measure in denoising process in units of block, object pixel is carried out denoising, R is contained in each piece, G, the pixel of tri-Color Channels of B, therefore R is taken full advantage of, G, the three-channel relation information of B carries out denoising, denoising result maintains the structural information of original image, make what the image after denoising can be more true fine and smooth to show image detail, improve the reduction degree of image, improve the denoising effect of image, final image is made to have higher quality and better visual effect.
Accompanying drawing explanation
Fig. 1 is the dot structure schematic diagram of the image of original bayer form;
Fig. 2 is the flow chart of the image de-noising method first embodiment of the present invention;
Fig. 3 is the schematic diagram choosing object pixel and reference pixel in the embodiment of the present invention on image;
Fig. 4 is the schematic diagram defining object block and reference block in the embodiment of the present invention on image;
Fig. 5 is the flow chart of image de-noising method second embodiment of the present invention;
Fig. 6 is that the image de-noising method of the image sensor application embodiment of the present invention is to obtain the flow chart of the application example of image;
Fig. 7 be Fig. 6 application example in the design sketch of image of original bayer form that obtains of imageing sensor;
Fig. 8 be Fig. 6 application example in the design sketch of complete RGB image that finally obtains of imageing sensor;
Fig. 9 is the module diagram of the image denoising device first embodiment of the present invention;
Figure 10 is the module diagram of image denoising device second embodiment of the present invention.
The realization of the object of the invention, functional characteristics and advantage will in conjunction with the embodiments, are described further with reference to accompanying drawing.
Detailed description of the invention
Should be appreciated that specific embodiment described herein is only in order to explain the present invention, is not intended to limit the present invention.
Referring to Fig. 2, it is proposed to the image de-noising method first embodiment of the present invention, described image de-noising method comprises the following steps:
S11, the pixel chosen in image, as object pixel, choose the pixel identical with the Color Channel of object pixel as reference pixel in the set point around object pixel.
In the embodiment of the present invention, successively each pixel in image is carried out denoising, until all pixels in traversing graph picture. First choosing a pixel as object pixel, this object pixel is the pixel being about to carry out denoising; Then the pixel identical with the Color Channel of object pixel is chosen in the set point around object pixel as reference pixel, for instance: when the pixel that object pixel is R passage, then the pixel choosing R passage is reference pixel; When the pixel that object pixel is G passage, then the pixel choosing G passage is reference pixel; When the pixel that object pixel is channel B, then the pixel choosing channel B is reference pixel.
As it is shown on figure 3, the R pixel in the big square frame of central authorities is the object pixel chosen, around object pixel, the R pixel in little square frame is the reference pixel chosen.
The size of set point can be determined as required, and in theory, scope is the bigger the better. Alternatively, the scope finally determined is a region centered by object pixel, pixel identical for the Color Channel of all and object pixel dropped in this region is chosen for reference pixel, is illustrated in figure 3 the region of 9x9 size.
S12, the definition block being sized centered by object pixel are object block, and the definition block being sized centered by reference pixel is reference block.
Object block is identical with the size of reference block, and the size of the two can set as required, is the bigger the better in theory. As shown in Figure 4, schematically mark object block 10 and reference block 20, object block 10 and reference block 20 and be 5x5 size. In Fig. 4 being diagrammatically only by property marked a reference block 20, it practice, the corresponding reference block of each reference pixel, the quantity of reference block is equal with the quantity of reference pixel.
S13, according to all pixels in reference block and object block, two blocks are carried out Diversity measure, calculate the weight of the reference pixel at reference block center the weight assignment 1 to object pixel based on the diversity factor of two blocks.
In this step S13, need to calculating the weight of each reference pixel, the embodiment of the present invention calculates the weight of this reference pixel according to the diversity factor of reference block corresponding to reference pixel with object block. As seen from the figure, reference block and not only comprise the pixel of R passage in object block, also include the pixel of G passage and channel B, therefore the embodiment of the present invention takes full advantage of the three-channel relation information of R, G, B and carries out denoising. Owing to the weight of reference pixel is relative to object pixel, and object pixel is 1 relative to the weight of itself, the therefore weight assignment 1 to object pixel.
Alternatively, when carrying out Diversity measure (or similarity measurement), the absolute value sum of the difference that can the pixel of reference block with correspondence position in object block (as corresponding in the upper left corner of reference block and the upper left corner of object block) be subtracted each other is as the diversity factor of two blocks, that is, reference block is taken absolute value after performing subtraction with two pixels of each position in object block, again the operation result of all positions is performed additive operation, using additive operation result as diversity factor, its computing formula is as follows:
R=∑ | M (i, j)-N (i, j) |,
Wherein, M (i, j) and N (i, j) respectively coordinate position is that (i, two pixels j), r is the diversity factor of reference block and object block in reference block and object block. For the block of 5x5 size, having 25 positions, two block correspondences obtain 25 numerical value after subtracting each other and taking absolute value, and being added up by 25 numerical value is exactly the diversity factor of two blocks.
Alternatively, when carrying out Diversity measure, square sum of the difference that can also the pixel of correspondence position in reference block and object block be subtracted each other is as the diversity factor of two blocks, that is, two pixels of each position in reference block and object block are performed after subtractions again square, again the operation result of all positions being performed additive operation, using additive operation result as diversity factor, its computing formula is as follows:
R=∑ (M (i, j)-N (i, j))2。
Further, it is also possible to adopt alternate manner of the prior art to carry out Diversity measure (or similarity measurement), do not repeat them here.
After obtaining diversity factor, then can calculate weight according to the negative correlativing relation of diversity factor Yu weight.
It is alternatively possible to calculate the weight of reference pixel according to below equation:
W=exp (-r2/sig2),
Wherein, r is the diversity factor of reference block and object block, and sig is empirical, and w is the weight of the reference pixel at reference block center. Sig is equivalent to dispersion degree or the diversity of the weight of each reference pixel, and the more big then dispersion degree of sig or diversity are more little, namely then each reference pixel weight closer to. The size positive correlation of the size of sig and reference block and object block, namely reference block and object block are more big, then sig is more big, and when reference block and block that object block is 5x5 size, the size of sig is about about 200 (such as 180-220).
Alternatively, it is also possible to calculate the weight of reference pixel according to other functional relationship, for instance linear function, reciprocal function etc.
In certain embodiments, it is also possible to calculated the similarity of reference block and object block by Diversity measure (or similarity measurement), the positive correlation further according to similarity Yu weight calculates weight.
Object pixel and all reference pixels are sought weighted mean by S14, weight according to object pixel and reference pixel, using the weighted mean denoising result as object pixel.
Concrete, the Color Channel of hypothetical target pixel is R, then all of R pixel in set point around object pixel and all reference pixels and object pixel, the weighted mean that the denoising result of object pixel is around object pixel in set point all of R pixel, its computing formula is:
Wherein, T_out is the denoising result of object pixel, molecular moiety is all of R pixel R (i in set point around object pixel, j) with its weight w (i, j) the sum of products, denominator part is weight w (i, j) sum of all of R pixel in set point around object pixel.
After a processes pixel completes, it may be judged whether traveled through all pixels in image. When there is no all pixels in traversing graph picture, return step S21 and continue next pixel is processed, circulation step S11-S14, until all pixels processed in image; When all pixels in traversing graph picture, the denoising result of all pixels, process ends in output image.
The image de-noising method of the embodiment of the present invention, by carrying out Diversity measure in denoising process in units of block, object pixel is carried out denoising, R is contained in each piece, G, the pixel of tri-Color Channels of B, therefore R is taken full advantage of, G, the three-channel relation information of B carries out denoising, denoising result maintains the structural information of original image, make what the image after denoising can be more true fine and smooth to show image detail, improve the reduction degree of image, improve the denoising effect of image, final image is made to have higher quality and better visual effect.
Referring to Fig. 5, it is proposed to image de-noising method second embodiment of the present invention, said method comprising the steps of:
S21, the pixel chosen in image, as object pixel, choose the pixel identical with the Color Channel of object pixel as reference pixel in the set point around object pixel.
S22, the definition block being sized centered by object pixel are object block, and the definition block being sized centered by reference pixel is reference block.
S23, according to all pixels in reference block and object block, two blocks are carried out Diversity measure, calculate the weight of the reference pixel at reference block center the weight assignment 1 to object pixel based on the diversity factor of two blocks.
In the present embodiment, step S21-S23 is identical with the step S11-S13 in first embodiment respectively, does not repeat them here.
The weight of reference pixel is adjusted by S24, difference based on reference pixel Yu object pixel, using the weight after regulating as the final weight of reference pixel.
When two block similarities relatively big (or diversity factor is less), and the difference of pixel itself (object pixel at Ji Lianggekuai center and the difference of reference pixel) also big time, edge blurry effect can be produced, in order to avoid producing the problems referred to above, the weight of reference pixel has been regulated by the present embodiment based on the difference of reference pixel Yu object pixel.
Alternatively, the absolute value of the difference first reference pixel at reference block and object block center and object pixel subtracted each other is as the diversity factor regulated value of two blocks, it is then based on the weight regulated value of the reference pixel at the diversity factor regulated value calculating reference block center of two blocks, finally using the product of the weight of reference pixel and the weight regulated value of reference pixel as the final weight of reference pixel, wherein:
The computing formula of diversity factor regulated value is: rr=| M-N |, and wherein, M is the reference pixel at reference block center, and N is the object pixel in object block, and rr is the diversity factor regulated value of reference block and object block;
The computing formula of weight regulated value is: ww=exp (-rr2/sig_p2), wherein, rr is the diversity factor regulated value of reference block and object block, and sig_p is empirical, and value size is about the weight regulated value that about 10 (such as 8-12), ww are the reference pixel at reference block center;
The computing formula of the weight that reference pixel is final is: w=w0* ww, wherein, w0For the weight of the previous step S23 reference pixel calculated, ww is weight regulated value, and w is the weight that reference pixel is final.
Alternatively it is also possible to the difference that the reference pixel at reference block and object block center and object pixel are subtracted each other square as the diversity factor regulated value of two blocks, i.e. rr=(M-N)2。
Weight after use adjustment carries out the calculating of the denoising result of next step S25, would not produce edge blurry effect.
Object pixel and all reference pixels are sought weighted mean by S25, weight according to object pixel and reference pixel, using the weighted mean denoising result as object pixel.
This step S25 is identical with the step S14 in first embodiment, does not repeat them here.
After a processes pixel completes, it may be judged whether traveled through all pixels in image. When all pixels in not yet traversing graph picture, return step S21 and continue next pixel is processed, circulation step S21-S24, until all pixels processed in image; When all pixels in traversing graph picture, the denoising result of all pixels, process ends in output image.
The weight of reference pixel is adjusted by the present embodiment based on the difference of reference pixel Yu object pixel, avoid because of two block similarities compared with Datong District time pixel itself difference also relatively big and produce the problem of edge blurry effect, further increase the denoising effect of image.
As shown in Figure 6, obtain the application example of image for the image de-noising method of the image sensor application embodiment of the present invention, comprise the following steps:
S100, from internal storage data stream, obtain the image of a frame bayer form.
The image of the bayer form obtained is original view data, and its image effect is as shown in Figure 7.
S200, image is carried out denoising.
It is alternatively possible to according to the step S11-S14 in aforementioned first embodiment, each pixel in traversing graph picture carries out denoising.
Alternatively it is also possible to according to the step S21-S25 in aforementioned second embodiment, each pixel in traversing graph picture carries out denoising.
Adopt such scheme that image is carried out denoising, denoising is carried out owing to taking full advantage of the three-channel relation information of R, G, B, make what the image after denoising can be more true fine and smooth to show image detail, improve the reduction degree of image, improve the denoising effect of image so that final image has higher quality and better visual effect.
S300, the image after denoising is carried out demosaicing process, obtain complete RGB image.
When image being carried out demosaicing (Demosaic) and processing, it is possible to adopt existing Demosaic algorithm to process.
Alternatively it is also possible to come in the following ways image is carried out demosaicing process:
Obtain the first Horizontal interpolation result and the first vertical interpolation result of the pixel that R passage/channel B is corresponding in image; Obtain horizontal gradient and the vertical gradient of the pixel that R passage/channel B is corresponding in image; Judge that the absolute value of difference between horizontal gradient and the vertical gradient obtained is more than 0 and less than the first predetermined threshold value, rebuild the pixel value of the G passage of the pixel that R passage/channel B is corresponding in image with the weighted average of the first differential vertical result according to the first Horizontal interpolation result obtained; Rebuild the R passage of the pixel that G passage is corresponding in image and the pixel value of channel B; The pixel value of the G passage of the reconstruction according to the pixel corresponding with channel B of R passage in image rebuilds the pixel value of the channel B of the pixel that R passage is corresponding in image and the pixel value of the R passage of the pixel that in image, channel B is corresponding.
Adopt above-mentioned demosaicing processing scheme, when the absolute value of the difference between the horizontal gradient and the vertical gradient that obtain is more than 0 and less than the first predetermined threshold value, the pixel value of the G passage of the pixel that the first Horizontal interpolation result according to acquisition is corresponding with R passage in the weighted average of the first differential vertical result reconstruction image or channel B, decrease pseudo-colours and moire fringes, thus improve the visual quality of image.
As shown in Figure 8, for the final complete RGB image obtained after adopting the image de-noising method of the embodiment of the present invention to carry out denoising, the details performance of this image is true and enriches, lines smooth, and color is pure and full, is a high-quality RGB image.
Referring to Fig. 9, propose the image denoising device first embodiment of the present invention, each pixel in described device traversing graph picture, respectively each pixel is carried out denoising, described device includes choosing module, definition module, the first computing module and the second computing module, wherein:
Choose module: the pixel being used for choosing in image, as object pixel, chooses the pixel identical with the Color Channel of object pixel as reference pixel in the set point around object pixel.
Choosing pixel that module chooses in image successively as object pixel to carry out denoising, after a processes pixel completes, reselection next one pixel, until all pixels in traversing graph picture, all pixels all complete till denoising.
When choosing reference pixel, the size of set point can set as required, and in theory, scope is the bigger the better. Alternatively, choosing the scope that module finally determines is a region centered by object pixel, and pixel identical for the Color Channel of all and object pixel dropped in this region is chosen for reference pixel.
Definition module: being used for the block being sized defined centered by object pixel is object block, the definition block being sized centered by reference pixel is reference block.
The corresponding reference block of each reference pixel, therefore finally defines an object block and multiple reference block. Object block is identical with the size of reference block, and the size of the two can set as required, is the bigger the better in theory. For example, it is possible to object block and reference block to be defined as the block of 5x5 size.
First computing module: for two blocks being carried out Diversity measure according to reference block with all pixels in object block, the weight of reference pixel at reference block center is calculated based on the diversity factor of two blocks, obtain the weight of each reference pixel the weight assignment 1 to object pixel accordingly. Then the weight of the weight of the object pixel calculated and all reference pixels is sent to the second computing module.
First computing module need to calculate the weight of each reference pixel, and the embodiment of the present invention calculates the weight of this reference pixel according to the diversity factor of reference block corresponding to reference pixel with object block. Owing to reference block and object block containing the pixel of tri-passages of R, G, B, by two blocks are carried out Diversity measure, the relation between triple channel is fully associated so that final denoising result contains three-channel relation information. Owing to the weight of reference pixel is relative to object pixel, and object pixel is 1 relative to the weight of itself, the therefore weight assignment 1 to object pixel.
As shown in Figure 9, first computing module includes Diversity measure unit and weight calculation unit, Diversity measure unit is for carrying out Diversity measure with all pixels in object block to two blocks according to reference block, weight unit for calculating the weight of the reference pixel at reference block center based on the diversity factor (or similarity) of two blocks, and the weight assignment 1 to object pixel.
Alternatively, Diversity measure unit is when carrying out Diversity measure (or similarity measurement), the absolute value sum of the difference that can the pixel of correspondence position in reference block and object block be subtracted each other is as the diversity factor of two blocks, that is, reference block is taken absolute value after performing subtraction with two pixels of each position in object block, again the operation result of all positions being performed additive operation, using additive operation result as diversity factor, its computing formula is as follows:
R=∑ | M (i, j)-N (i, j) |,
Wherein, M (i, j) and N (i, j) respectively coordinate position is that (i, two pixels j), r is the diversity factor of reference block and object block in reference block and object block. For the block of 5x5 size, having 25 positions, two block correspondences obtain 25 numerical value after subtracting each other and taking absolute value, and being added up by 25 numerical value is exactly the diversity factor of two blocks.
Alternatively, when carrying out Diversity measure, square sum of the difference that the pixel of correspondence position in reference block and object block can also be subtracted each other by Diversity measure unit is as the diversity factor of two blocks, that is, two pixels of each position in reference block and object block are performed after subtractions again square, again the operation result of all positions being performed additive operation, using additive operation result as diversity factor, its computing formula is as follows:
R=∑ (M (i, j)-N (i, j))2。
Further, it is also possible to adopt alternate manner of the prior art to carry out Diversity measure (or similarity measurement), do not repeat them here.
After obtaining diversity factor, weight calculation unit then can calculate weight according to the negative correlativing relation of diversity factor Yu weight.
Alternatively, weight calculation unit can calculate the weight of reference pixel according to below equation:
W=exp (-r2/sig2),
Wherein, r is the diversity factor of reference block and object block, and sig is empirical, and w is the weight of the reference pixel at reference block center. Sig is equivalent to dispersion degree or the diversity of the weight of each reference pixel, and the more big then dispersion degree of sig or diversity are more little, namely then each reference pixel weight closer to. The size positive correlation of the size of sig and reference block and object block, namely reference block and object block are more big, then sig is more big, and when reference block and block that object block is 5x5 size, the size of sig is about about 200 (such as 180-220).
Alternatively, it is also possible to calculate the weight of reference pixel according to other functional relationship, for instance linear function, reciprocal function etc.
In certain embodiments, Diversity measure unit can also pass through Diversity measure (or similarity measurement) and calculate the similarity of reference block and object block, and weight calculation unit calculates weight further according to the positive correlation of similarity Yu weight.
Second computing module: for object pixel and all reference pixels being asked weighted mean according to the weight of object pixel and reference pixel, using the weighted mean denoising result as object pixel.
Concrete, the Color Channel of hypothetical target pixel is R, then all of R pixel in set point around object pixel and all reference pixels and object pixel, the weighted mean that the denoising result of object pixel is around object pixel in set point all of R pixel, its computing formula is:
Wherein, T_out is the denoising result of object pixel, molecular moiety is all of R pixel R (i in set point around object pixel, j) with its weight w (i, j) the sum of products, denominator part is weight w (i, j) sum of all of R pixel in set point around object pixel.
When, after all pixels in traversing graph picture, image denoising device exports the denoising result of all pixels.
The image denoising device of the embodiment of the present invention, by carrying out Diversity measure in denoising process in units of block, object pixel is carried out denoising, R is contained in each piece, G, the pixel of tri-Color Channels of B, therefore R is taken full advantage of, G, the three-channel relation information of B carries out denoising, denoising result maintains the structural information of original image, make what the image after denoising can be more true fine and smooth to show image detail, improve the reduction degree of image, improve the denoising effect of image, final image is made to have higher quality and better visual effect.
Referring to Figure 10, it is proposed to image denoising device second embodiment of the present invention, the present embodiment is that the first computing module adds weight regulatory function on the basis of first embodiment.
As shown in Figure 10, the first computing module includes Diversity measure unit, weight calculation unit and weight regulon, and Diversity measure unit is identical with first embodiment with weight calculation unit. After weight calculation unit calculates the weight of reference pixel, it is sent to weight regulon and is adjusted; The weight of reference pixel is adjusted by weight regulon based on the difference of reference pixel Yu object pixel, and as the weight that reference pixel is final, weight after regulating is sent to the second computing module.
Alternatively, the absolute value of the difference that first reference pixel at reference block and object block center and object pixel are subtracted each other by weight regulon is as the diversity factor regulated value of two blocks, it is then based on the weight regulated value of the reference pixel at the diversity factor regulated value calculating reference block center of two blocks, finally using the product of the weight of reference pixel and the weight regulated value of reference pixel as the final weight of reference pixel, wherein:
The computing formula of diversity factor regulated value is: rr=| M-N |, and wherein, M is the reference pixel at reference block center, and N is the object pixel in object block, and rr is the diversity factor regulated value of reference block and object block;
The computing formula of weight regulated value is: ww=exp (-rr2/sig_p2), wherein, rr is the diversity factor regulated value of reference block and object block, and sig_p is empirical, and value size is about the weight regulated value that about 10 (such as 8-12), ww are the reference pixel at reference block center;
The computing formula of the weight that reference pixel is final is: w=w0* ww, wherein, w0For the weight of the reference pixel that weight calculation unit calculates, ww is weight regulated value, and w is the weight that reference pixel is final.
Alternatively, the difference that the reference pixel at reference block and object block center and object pixel can also be subtracted each other by weight regulon square as the diversity factor regulated value of two blocks, i.e. rr=(M-N)2。
The weight of reference pixel is adjusted by the present embodiment based on the difference of reference pixel Yu object pixel, avoid because of two block similarities compared with Datong District time pixel itself difference also relatively big and produce the problem of edge blurry effect, further increase the denoising effect of image.
It should be understood that the image denoising device that above-described embodiment provides belongs to same design with image de-noising method embodiment, it implements process and refers to embodiment of the method, and the technical characteristic in embodiment of the method is all corresponding applicable in device embodiment, repeats no more here.
The image de-noising method of the embodiment of the present invention and device, be particularly suited for the image to bayer form or be similar to the image of bayer form and carry out denoising.
The image de-noising method of the embodiment of the present invention and device, it is possible to be applied to the computer equipment such as PC, terminating machine, it is also possible to be applied to the camera installation such as camera, camera, it is also possible to be applied to the mobile terminal such as mobile phone, flat board.
Through the above description of the embodiments, those skilled in the art is it can be understood that can add the mode of required general hardware platform by software to above-described embodiment method and realize, hardware can certainly be passed through, but in a lot of situation, the former is embodiment more preferably. Based on such understanding, the part that prior art is contributed by technical scheme substantially in other words can embody with the form of software product, this computer software product is stored in a storage medium (such as ROM/RAM, magnetic disc, CD), including some instructions with so that a station terminal equipment (can be mobile phone, computer, server, or the network equipment etc.) perform the method described in each embodiment of the present invention.
Should be understood that; these are only the preferred embodiments of the present invention; can not therefore limit the scope of the claims of the present invention; every equivalent structure utilizing description of the present invention and accompanying drawing content to make or equivalence flow process conversion; or directly or indirectly it is used in other relevant technical fields, all in like manner include in the scope of patent protection of the present invention.
Claims (10)
1. an image denoising device, it is characterised in that each pixel in described device traversing graph picture, carries out denoising to each pixel respectively, including:
Choosing module, the pixel being used for choosing in described image, as object pixel, chooses the pixel identical with the Color Channel of described object pixel as reference pixel in the set point around described object pixel;
Definition module, is object block for the block being sized defined centered by described object pixel, and the block that definition is sized described in centered by described reference pixel is reference block, the corresponding reference block of each reference pixel;
First computing module, for described reference block and described object block being carried out Diversity measure according to described reference block with all pixels in described object block, weight based on described reference block with the described reference pixel at the described reference block center of diversity factor calculating of described object block, obtain the weight of each described reference pixel accordingly, and the weight assignment 1 to described object pixel, wherein said diversity factor and described weight negative correlation;
Second computing module, for seeking weighted mean according to the weight of described object pixel and described reference pixel to described object pixel and all described reference pixels, using the described weighted mean denoising result as described object pixel.
2. image denoising device according to claim 1, it is characterized in that, described first computing module includes Diversity measure unit, and described Diversity measure unit is used for: the absolute value sum of the difference pixel of correspondence position in described reference block and described object block subtracted each other is as the diversity factor of described reference block Yu described object block.
3. image denoising device according to claim 1, it is characterised in that described first computing module includes weight calculation unit, and described weight calculation unit is used for:
According to formula w=exp (-r2/sig2) calculate the weight of the described reference pixel at described reference block center, wherein, r is the diversity factor of described reference block and described object block, and sig is empirical, and w is the weight of the described reference pixel at described reference block center.
4. the image denoising device according to any one of claim 1-3, it is characterised in that described first computing module also includes weight regulon, and described weight regulon is used for:
The absolute value of the difference described reference pixel at described reference block and described object block center and described object pixel subtracted each other is as the diversity factor regulated value of described reference block Yu described object block, the weight regulated value of the described reference pixel at described reference block center is calculated, using the weight of described reference pixel and the product of the weight regulated value of described reference pixel as the final weight of described reference pixel based on the diversity factor regulated value of described reference block and described object block.
5. image denoising device according to claim 4, it is characterised in that described weight regulon is used for:
According to formula ww=exp (-rr2/sig_p2) calculate the weight regulated value of the described reference pixel at described reference block center, wherein, rr is the diversity factor regulated value of described reference block and described object block, and sig_p is empirical, and ww is the weight regulated value of the described reference pixel at described reference block center.
6. an image de-noising method, it is characterised in that include step:
Choose a pixel in image as object pixel, choose the pixel identical with the Color Channel of described object pixel in the set point around described object pixel as reference pixel;
The definition block being sized centered by described object pixel is object block, and the block that definition is sized described in centered by described reference pixel is reference block, the corresponding reference block of each reference pixel;
With all pixels in described object block, described reference block and described object block are carried out Diversity measure according to described reference block, weight based on described reference block with the described reference pixel at the described reference block center of diversity factor calculating of described object block, obtain the weight of each described reference pixel accordingly, and the weight assignment 1 to described object pixel, wherein said diversity factor and described weight negative correlation;
Described object pixel and all described reference pixels are sought weighted mean by the weight according to described object pixel and described reference pixel, using the described weighted mean denoising result as described object pixel;
Each pixel traveled through in described image repeats above-mentioned steps.
7. image de-noising method according to claim 6, it is characterised in that described according to all pixels in described reference block and described object block, described reference block and described object block carried out Diversity measure and include:
The absolute value sum of the difference pixel of correspondence position in described reference block and described object block subtracted each other is as the diversity factor of described reference block Yu described object block.
8. image de-noising method according to claim 6, it is characterised in that the described weight based on described reference block with the described reference pixel at the described reference block center of diversity factor calculating of described object block includes:
According to formula w=exp (-r2/sig2) calculate the weight of the described reference pixel at described reference block center, wherein, r is the diversity factor of described reference block and described object block, and sig is empirical, and w is the weight of the described reference pixel at described reference block center.
9. the image de-noising method according to any one of claim 6-8, it is characterised in that also include after the step of the weight of the described reference pixel at the described described reference block center of diversity factor calculating according to described reference block and described object block:
The absolute value of the difference described reference pixel at described reference block and described object block center and described object pixel subtracted each other is as the diversity factor regulated value of described reference block Yu described object block, the weight regulated value of the described reference pixel at described reference block center is calculated, using the weight of described reference pixel and the product of the weight regulated value of described reference pixel as the final weight of described reference pixel based on the diversity factor regulated value of described reference block and described object block.
10. image de-noising method according to claim 9, it is characterised in that the described weight regulated value based on described reference block with the described reference pixel at the described reference block center of diversity factor regulated value calculating of described object block includes:
According to formula ww=exp (-rr2/sig_p2) calculate the weight regulated value of the described reference pixel at described reference block center, wherein, rr is the diversity factor regulated value of described reference block and described object block, and sig_p is empirical, and ww is the weight regulated value of the described reference pixel at described reference block center.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610186820.3A CN105678718B (en) | 2016-03-29 | 2016-03-29 | Image de-noising method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610186820.3A CN105678718B (en) | 2016-03-29 | 2016-03-29 | Image de-noising method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105678718A true CN105678718A (en) | 2016-06-15 |
| CN105678718B CN105678718B (en) | 2019-11-15 |
Family
ID=56224531
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610186820.3A Active CN105678718B (en) | 2016-03-29 | 2016-03-29 | Image de-noising method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105678718B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110517206A (en) * | 2019-09-05 | 2019-11-29 | 昆山锐芯微电子有限公司 | Colored moire fringes removing method and device |
| CN113240607A (en) * | 2021-05-26 | 2021-08-10 | Oppo广东移动通信有限公司 | Image denoising method and device, electronic equipment and storage medium |
| CN113596401A (en) * | 2021-07-29 | 2021-11-02 | 上海应用技术大学 | Image de-similarity transmission and restoration method based on ORB similarity judgment |
| CN117911405A (en) * | 2024-03-19 | 2024-04-19 | 济南科汛智能科技有限公司 | X-ray image processing method for dental pulp and dental body detection |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101464997A (en) * | 2009-01-21 | 2009-06-24 | 北京中星微电子有限公司 | Method and device for removing noise |
| US20110142368A1 (en) * | 2009-12-16 | 2011-06-16 | Samsung Electronics Co., Ltd. | Method and apparatus for block-based image denoising |
| CN102509269A (en) * | 2011-11-10 | 2012-06-20 | 重庆工业职业技术学院 | Image denoising method combined with curvelet and based on image sub-block similarity |
| CN103679639A (en) * | 2012-09-05 | 2014-03-26 | 北京大学 | Image denoising method and device based on non-local mean value |
| CN104376540A (en) * | 2014-12-01 | 2015-02-25 | 北京思比科微电子技术股份有限公司 | Bayer image denoising method |
| CN104978715A (en) * | 2015-05-11 | 2015-10-14 | 中国科学院光电技术研究所 | Non-local mean image denoising method based on filtering window and parameter self-adaption |
-
2016
- 2016-03-29 CN CN201610186820.3A patent/CN105678718B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101464997A (en) * | 2009-01-21 | 2009-06-24 | 北京中星微电子有限公司 | Method and device for removing noise |
| US20110142368A1 (en) * | 2009-12-16 | 2011-06-16 | Samsung Electronics Co., Ltd. | Method and apparatus for block-based image denoising |
| CN102509269A (en) * | 2011-11-10 | 2012-06-20 | 重庆工业职业技术学院 | Image denoising method combined with curvelet and based on image sub-block similarity |
| CN103679639A (en) * | 2012-09-05 | 2014-03-26 | 北京大学 | Image denoising method and device based on non-local mean value |
| CN104376540A (en) * | 2014-12-01 | 2015-02-25 | 北京思比科微电子技术股份有限公司 | Bayer image denoising method |
| CN104978715A (en) * | 2015-05-11 | 2015-10-14 | 中国科学院光电技术研究所 | Non-local mean image denoising method based on filtering window and parameter self-adaption |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110517206A (en) * | 2019-09-05 | 2019-11-29 | 昆山锐芯微电子有限公司 | Colored moire fringes removing method and device |
| CN110517206B (en) * | 2019-09-05 | 2021-10-15 | 锐芯微电子股份有限公司 | Method and device for eliminating color moire |
| CN113240607A (en) * | 2021-05-26 | 2021-08-10 | Oppo广东移动通信有限公司 | Image denoising method and device, electronic equipment and storage medium |
| CN113596401A (en) * | 2021-07-29 | 2021-11-02 | 上海应用技术大学 | Image de-similarity transmission and restoration method based on ORB similarity judgment |
| CN117911405A (en) * | 2024-03-19 | 2024-04-19 | 济南科汛智能科技有限公司 | X-ray image processing method for dental pulp and dental body detection |
| CN117911405B (en) * | 2024-03-19 | 2024-05-24 | 济南科汛智能科技有限公司 | X-ray image processing method for dental pulp and dental body detection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105678718B (en) | 2019-11-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106127688B (en) | A kind of super-resolution image reconstruction method and its system | |
| KR101634562B1 (en) | Method for producing high definition video from low definition video | |
| CN102103751B (en) | Foreground image extraction method and device | |
| CN102119401B (en) | Method and apparatus for banding artifact detection | |
| CN105678718A (en) | Method and device for image denoising | |
| CN104023225B (en) | Video quality evaluation without reference method based on Space-time domain natural scene statistical nature | |
| JPH0799660A (en) | Motion compensation predicting device | |
| CN106170093B (en) | Intra-frame prediction performance improving coding method | |
| CN108600783B (en) | Frame rate adjusting method and device and terminal equipment | |
| CN109360157B (en) | TV and wavelet regularization-based spatial variation blurred image restoration method | |
| CN110324617B (en) | Image processing method and device | |
| CN104063849A (en) | Video super-resolution reconstruction method based on image block self-adaptive registration | |
| Sandić-Stanković et al. | Fast blind quality assessment of DIBR-synthesized video based on high-high wavelet subband | |
| CN105279742B (en) | A kind of image de-noising method quickly based on piecemeal estimation of noise energy | |
| CN102223545A (en) | Rapid multi-view video color correction method | |
| CN112435182B (en) | Image noise reduction method and device | |
| CN105389825A (en) | Image processing method and system | |
| CN110443754B (en) | Method for improving resolution of digital image | |
| CN113422956B (en) | Image coding method and device, electronic equipment and storage medium | |
| CN113111770B (en) | Video processing method, device, terminal and storage medium | |
| CN101268623B (en) | Method and device for creating shape variable blocks | |
| JP4001629B2 (en) | Method and apparatus for processing pixels of an image segment by a computer | |
| CN113822937B (en) | Image correction method, device, equipment and storage medium | |
| CN110717875B (en) | High-definition image processing method | |
| CN108810317B (en) | True motion estimation method and device, computer readable storage medium and terminal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |