CN106341613B - Wide dynamic range image method - Google Patents
Wide dynamic range image method Download PDFInfo
- Publication number
- CN106341613B CN106341613B CN201510390387.0A CN201510390387A CN106341613B CN 106341613 B CN106341613 B CN 106341613B CN 201510390387 A CN201510390387 A CN 201510390387A CN 106341613 B CN106341613 B CN 106341613B
- Authority
- CN
- China
- Prior art keywords
- value
- image
- region
- pixel value
- pixel
- 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
Landscapes
- Image Processing (AREA)
Abstract
A kind of wide dynamic range image method, includes the following steps.Area maps are carried out to image through multiple regions mapping curve, to carry out incomplete the same gamma correction for multiple regions in image, wherein region at least includes a pixel.The step of area maps, is as follows.Obtain low frequency image.The reference value in each region of image is obtained according to low frequency image.The reference value in the region according to image selects one of them from multiple regions mapping curve.Area maps curve according to selection obtains corresponding gain curve, the pixel value in region or the yield value of brightness are obtained according to gain curve, gain adjustment value is obtained according to the corresponding brightness of pixel value, and pixel value is adjusted according to brightness, yield value and gain adjustment value.
Description
Technical field
The present invention relates to a kind of image methods, and especially a kind of wide dynamic range image (wide dynamic
Range imaging) method.
Background technique
The dynamic range for increasing image can often increase the details of image entirety, and the especially information at dark place and bright place can
To become apparent, and the dynamic range for increasing image may be roughly divided into the way of two major classes type: high dynamic-range image (high
Dynamic range imaging) method and wide dynamic range image method.
The image of the available better quality of high dynamic-range image method, but needed mostly in algorithm more than two
The image opened is as the available image output of input image, and furthermore this two input images can be difficult at the same time again between
It is arrived with capturing, therefore high dynamic-range image method is not easily applied to dynamic video signal and instant (real time) image very much
Output, that is, it has many use limitations.
Wide dynamic range image method only needs an input image to can be obtained by an image output, therefore it is whether
It is all easily real on software or hardware to do, but also because wide dynamic range image method only uses an input shadow
Picture, therefore its utilizable information is fewer than multiple input images, and its required algorithm is also therefore completely important.
General width dynamic image range method is easy to generate other adverse effects, such as: image incidental information becomes larger, is whole
Body contrast reduces, halation phenomenon and color is excessively bright-coloured or even colour cast.Even, the shadow that wide dynamic range image method generates
Picture, visual experience in the dark is not satisfactory, and especially when face is under backlighting environment, face can be excessively dark.
Most common wide dynamic range image method first is that reached using a universe mapping curve, such as use gal
Ma corrects (gamma correction) curve to be mapped, and formula is expressed as P'=N (P/N)1/γ, wherein P is indicated defeated
The pixel value (it can be red, green or blue pixel value) entered, P ' indicates to be corrected rear institute through universe mapping curve
The pixel value of generation, N indicates the quantization order (for example, 255) of pixel value, and γ is a static state or can dynamically adjust
Parameter, determine the amplitude of variation of gamma correction curve (variation is relatively slow or relatively steep).Such method is will be all in image
Pixel value is corrected using a universe mapping curve, therefore relatively simple in implementation.
However, the following disadvantage of the wide dynamic range image method person of having reached using a universe mapping curve.Because
Red, green or blue pixel value correspond to the different location of universe mapping curve, and these different locations be corresponding with it is different
Tangent slope, therefore red, green or blue pixel value yield value are not all the same, and generation the case where might have colour cast.In addition,
Because maxgain value is relevant to the maximum tangent slope of universe mapping curve, therefore for the details and view in dark portion region in image
Feel that the promotion of impression is very limited.In addition to this, although universe mapping curve can promote the brightness and details in somewhere in image,
But the details at other in image one can be additionally sacrificed, the loss of the details and overall contrast of image part is caused.
Another universal wide dynamic range image method is reached using multiple regions mapping curve.Image can be by
Be divided into multiple regions, wherein each region can correspond to one of area maps curve, with the pixel value to input into
Row correction.The wide dynamic range image method reached using multiple regions mapping curve, which can solve, can not be promoted in image secretly
The problem of details and visual experience in portion region, and not will cause the loss of the details and overall contrast of image part part.
However, the wide dynamic range image method reached using multiple regions mapping curve is that each region is corresponding
To one of area maps curve, and image is difficult correctly to divide region, therefore may have meeting halation existing between each region
The generation of elephant.Light especially in the case where not being available complicated hardware and further making edge filter to image, in image
Dizzy phenomenon can be clearly.Still it will appear even if edge filter has been used to handle image, between region unnatural
Lamination.
Summary of the invention
The embodiment of the present invention provides a kind of wide dynamic range image method comprising following step.Through multiple regions
Mapping curve carries out area maps to image, carries out incomplete the same gamma correction to be directed to multiple regions in image, wherein area
Domain at least includes a pixel.The step of area maps, is described as follows.The image is filtered, to obtain low frequency image.Foundation
The reference value in low frequency image acquisition each region of image.The reference value in the region according to image selects it from multiple regions mapping curve
One of correspond to image region.Area maps curve according to the region for corresponding to image obtains corresponding gain curve,
The pixel value in region or the yield value of brightness are obtained according to gain curve, obtain gain adjustment according to the corresponding brightness of pixel value
Value, and pixel value is adjusted according to brightness, yield value and gain adjustment value.
In summary, wide dynamic range image method provided by the embodiment of the present invention can increase image entirety
Details, and dark portion region and the details in highlights region in image can be allowed obviously to show simultaneously.
Feature and technology of the invention are further understood that enable the technical field of the invention to have usually intellectual
Content, reference should be made to the following detailed description and accompanying drawings of the present invention, but these explanations are intended merely to illustrate with institute accompanying drawings
The present invention, rather than make any limitation to interest field of the invention.
Detailed description of the invention
Fig. 1 is the schematic diagram of the process of the wide dynamic range image method of the embodiment of the present invention.
Fig. 2 is the schematic diagram of the universe mapping curve of the embodiment of the present invention.
Fig. 3 is the signal that the wide dynamic range image method of the embodiment of the present invention carries out the process of area maps to image
Figure.
Fig. 4 is that the wide dynamic range image method of the embodiment of the present invention lower to image the calculating difference of halation phenomenon
The schematic diagram of value.
When Fig. 5 is that the wide dynamic range image method of the embodiment of the present invention carries out attenuating halation phenomenon to image, calculated
The schematic diagram of the curve of difference value and halation value out.
When Fig. 6 is that the wide dynamic range image method of the embodiment of the present invention carries out attenuating halation phenomenon to image, calculated
The schematic diagram of the curve of maximum value and fine tuning rate out.
Fig. 7 is the schematic diagram of the multiple regions mapping curve of the embodiment of the present invention.
Fig. 8 is the schematic diagram of the selected corresponding gain curve of area maps curve of the embodiment of the present invention.
It is calculated bright when Fig. 9 is that the wide dynamic range image method of the embodiment of the present invention carries out gain amplification to image
The schematic diagram of degree and the curve of adjusted value.
When the wide dynamic range image method of Figure 10 embodiment of the present invention carries out gain amplification to image, to the pixel of input
Carry out the schematic diagram of gain amplification.
When Figure 11 is that the wide dynamic range image method of the embodiment of the present invention carries out noise reduction to image, used picture
The schematic diagram of element value and the curve of suppression noise value.
When Figure 12 is that the wide dynamic range image method of the embodiment of the present invention carries out noise reduction to image, it is flat to calculate pixel
The schematic diagram of slow value.
Figure 13 is that the wide dynamic range image method of the embodiment of the present invention carries out from fatigue resistance pixel adjusted image
The schematic diagram of the histogram of value.
Symbol description
S11~S14, S31~S36: steps flow chart
GC: universe mapping curve
R1~R3: region
W1, W2: low-pass filtering window
CP: pixel
LC1~LC3, LC: area maps curve
GNC: gain curve
MASK1~MASKN: mask
MOD: mode signal
MUX: multiplexer
CONV: volume integrator
Psmooth: gentle pixel value
HC1~HC3: histogram
α 1: halation value
α 3: gain adjustment value
α 1 ', R1 ', MA, mA, PI1, PI2, β 1, β 2, m1, m2: numerical value
R1: fine tuning rate
TH1~TH5: threshold value
R, R ': red luma value
G, G ': Green brightness value
B, B ': blue pixel value
GN, GN ': yield value
Y, Y ': brightness
U, U ': color difference
V, V ': concentration
DNS: suppression noise value
DIF: difference value
MAX: maximum value
Specific embodiment
The embodiment of the present invention provides a kind of wide dynamic range image method, and the wide dynamic range image method can be by image
Depth of exposure reduce, the details in overexposure or excessively bright highlights region in image can be rendered obvious by, and promote visual impression
By.Then, the wide dynamic range image method can carry out area maps to image through multiple regions mapping curve, to be directed to
Different zones carry out different gamma corrections, so that dark portion region and the details in highlights region can completely be in image
Existing, wherein region at least includes a pixel (including red, blue and green pixel;Alternatively, including two green pixels and two
A red pixel (bell pattern)).
It should also be noted that the step of carrying out area maps to image is described as follows.The wide dynamic range image method meeting
Low-pass filtering is carried out to image, then, generates reference value further according to the pixel value in the region of the image after low-pass filtered, and
Selected one of them as image from multiple regions mapping curve according to the reference value in region (it is not low-pass filtered)
The corresponding area maps curve in region.For each region of image (it is not low-pass filtered), reflected according to selected region
It penetrates curve and obtains corresponding gain curve, obtain the pixel value in region or the yield value of brightness according to the gain curve, so
Afterwards, then according to the corresponding brightness of pixel value gain adjustment value is obtained, and is adjusted according to brightness, yield value and gain adjustment value
Pixel value, to solve the problems, such as in image that color is excessively bright-coloured or colour cast whereby.
In other embodiments of the invention, in the step of carrying out area maps to image, the reference value in above-mentioned zone can
Be it is low-pass filtered after image region in pixel value or the reference value can be obtained according to following manner.?
It is above-mentioned to image carry out low-pass filtering after, the wide dynamic range image method can also calculate it is low-pass filtered after image it is every
Difference value of one pixel value under two different low-pass filtering windows, and according to difference value and one of low-pass filtering window
The calculated maximum value of lower institute determines halation phenomenon specific gravity, to determine that the pixel value is corresponding through this halation phenomenon specific gravity
Reference value.
In other embodiments of the invention, in the step of carrying out area maps to image, the wide dynamic range image side
Method more carries out area marking processing to image, for example, human face recognition (face detection), visual salient region mark
(salience map) or image cutting process (image segmentation), according to brightness, yield value and gain tune
Whole value is come after adjusting pixel value, further according to area marking processing as a result, carrying out again to the pixel value of specific region bright
Degree adjustment.For example, promoting the brightness in the region of face again.
In other embodiments of the invention, in the step of carrying out area maps to image, the wide dynamic range image side
Method is more after adjusting pixel value according to brightness, yield value and gain adjustment value, according to the region of the image after low-pass filtered
Reference value determine suppression noise value, and lower the noise of image according to suppression noise value and selected mask.
In other embodiments of the invention, the wide dynamic range image method more can carry out area maps to image
Before, first first image universe of progress is mapped using universe mapping curve, to make the primary overall situation for the brightness of image
Correction.In addition to this, the wide dynamic range image method can also carry out certainly image after carrying out area maps to image
Fatigue resistance adjustment, to increase the contrast of image.
Then, the details of above-mentioned steps will be described in detail with schema and text.It should also be noted that concept of the present invention may
It embodies in many different forms, and should not be construed as limited by exemplary embodiments set forth herein.In addition, in the drawings
Same reference numerical digit can be used to indicate similar element.
Firstly, please referring to Fig. 1, Fig. 1 is the schematic diagram of the process of the wide dynamic range image method of the embodiment of the present invention.Institute
Stating wide dynamic range image method includes step S11~S14, and can be implemented in the electronic device with image acquisition function,
Such as smart phone, digital still camera or drive recorder etc..
In step S11, the wide dynamic range image method can reduce the depth of exposure of image, to allow overexposure in image
Or the details in excessively bright highlights region can be rendered obvious by, and promote visual experience.For example, through stop down, reducing exposure
Time or the depth of exposure that image is reduced through the image processing techniques of software, to sum up, the present invention and unlimited reduction shadow
The practice of the depth of exposure of picture.
Then, in step S12, the wide dynamic range image method carries out universe to image through universe mapping curve
Mapping, to promote the brightness of whole image.Image can be divided into multiple regions, each region includes at least one pixel.If
Each region includes a plurality of pixels, then the average pixel value of the red in each region, green and blue pixel each
Red, green and the blue pixel value after correction as the input value of universe mapping curve, to generate the pixel in this region.If
Each region only includes a pixel, then the pixel value conduct of the red in each region, green and blue pixel each
The input value of universe mapping curve, red, green and the blue pixel value after correction to generate the pixel in this region.
Universe mapping curve is, for example, gamma correction curve, and formula is expressed as P'=N (P/N)1/γ, wherein P is indicated defeated
The pixel value (it can be red, green or blue pixel value or average pixel value) entered, P ' indicates to penetrate universe mapping curve
Pixel value caused by after being corrected, N indicate the quantization order (for example, 255) of pixel value, and γ be a static state or
The parameter that can dynamically adjust determines the amplitude of variation of gamma correction curve (variation is relatively slow or relatively steep).It should also be noted that ginseng
Number γ can be adjusted according to presentation content or demand, and universe mapping curve and it is non-limiting be only capable of for Gamma correction song
Line.
It is the schematic diagram of the universe mapping curve of the embodiment of the present invention referring to Fig. 1 and Fig. 2, Fig. 2.Briefly,
If image is divided into region R1~R3, region R1~R3 uses same universe mapping curve GC to carry out universe mapping, to mention
Rise the brightness of whole image.In Fig. 2, the corresponding longitudinal axis of universe mapping curve GC represents pixel value (such as the above-mentioned public affairs of output
The P ' of formula), and the corresponding horizontal axis of universe mapping curve GC represents the pixel value (such as P of above-mentioned formula) of input.
It please then continue to referring to Fig.1, universe mapping is being carried out to image, after the brightness to promote whole image, it is contemplated that
May there are different object and scene in image, the brightness for needing to be promoted corresponding to the region of different objects and scene may be different
Sample, therefore step S13 can be penetrated, area maps are carried out to image using multiple and different area maps curves, to not same district
Different promotions is made in the brightness in domain.
Then, the details that area maps are carried out to image is further explained.It is this referring to Fig. 1 and Fig. 3, Fig. 3
The wide dynamic range image method of inventive embodiments carries out the schematic diagram of the process of area maps to image.Step S13 can be into
It include one step step S31~S36, wherein step S31, S33 and S36 is not steps necessary, also can be according to demand certainly
It is selectively removed in step S13.
Firstly, in step S31, area marking processing carried out to image, that is, image is made to judge before something
Movement, for example, human face recognition, visual salient region mark or image cutting process, in subsequent step S35 according to brightness, gain
After value adjusts pixel value with gain adjustment value, further according to area marking processing as a result, again to the picture of specific region
Element value carries out brightness adjustment.For example, promoting the brightness in the region of face again.
It should also be noted that human face recognition, visual salient region mark are cut with image according to acceptable hardware complexity
It is performed with cutting the processing property of can choose.Preferably, human face recognition will be at least performed, because the brightness of face part sometimes may be used
Can be insufficient, therefore step S31 can be penetrated, by the area marking of face, in subsequent step S35 according to brightness, yield value and increasing
Beneficial adjusted value after adjusting pixel value, promotes the brightness in the region of face, so that the quality of image relatively meets user again
Demand.
Then, in step s 32, low-pass filtering is carried out to image, to take out the region of low frequency in image.Image is carried out
The practice of low-pass filtering for example has using gaussian filtering, median filtering or edge-protected filtering (edge preserved
Filtering) image is filtered, wherein edge-protected filtering may, for example, be bilateral filtering (bilateral again
Filtering) or guiding filtering (guided filtering), and the detailed practice of low-pass filtering is not to limit this hair
It is bright.
The pixel value in the region of low frequency image (image after low-pass filtered) can be used to generate reference value, according to area
Reference value in domain selects one of them corresponding as the region of image (it is not low-pass filtered) from multiple regions mapping curve
Area maps curve.It should also be noted that the reference value in the region of low frequency image is equal to low if omitting following step S33
The pixel value in the region of frequency image.If not omitting following step S33, then the pixel value according to the region of low frequency image executes
Step S33 calculates the reference value in the region of low frequency image.It below will further introduction step S33.
Because if different zones are likely to result in meeting between region using the adjustment that different zones mapping curve carries out brightness
There is halation phenomenon generation, therefore can use step S33 and low frequency image caused by step S32 is handled, to lower shadow
The halation phenomenon of picture.In addition, as described earlier, being removed from step S13 to the step S33 property of can choose, for example, scene meeting
Caused by halation phenomenon it is not serious, then can choose and do not execute step S33.
Image is subtracted referring to the wide dynamic range image method that Fig. 3 and Fig. 4, Fig. 4 are the embodiment of the present invention
The schematic diagram of the calculating difference value of low halation phenomenon.In step S33, the wide dynamic range image method can first be calculated through low
Difference value DIF, formula DIF=LPF of each pixel value of frequency image under two different low-pass filtering window W1, W2
{Pi|i∈W1}-LPF{Pi| i ∈ W2 }, wherein LPF { Pi| i ∈ W1 } indicate that point is according to low-pass filtering window centered on pixel CP
The calculated low frequency value of W1 institute, and LPF { Pi| i ∈ W2 } indicate that point is counted according to low-pass filtering window W2 centered on pixel CP
The low frequency value of calculating, wherein i is pixel index value.In general difference value DIF is bigger, and it is existing that expression is more likely to occur halation
As, and need to give biggish halation value α 1.
It should also be noted that the size of above-mentioned low-pass filtering window W1 and W2 are, for example, 33x33 and 3x3 pixel size, but
The present invention is not limited with the size of window W1 and W2.In addition to this, the type of above-mentioned low-pass filtering window W1 and W2 preferably that
This is identical, but actual conditions also can be with different from each other, to sum up, invention is not limited with the type of window W1 and W2.
Please image is subtracted referring next to the wide dynamic range image method that Fig. 3 and Fig. 5, Fig. 5 are the embodiment of the present invention
When low halation phenomenon, the curve of calculated difference value and halation value schematic diagram.Halation value α 1 is the numerical value of 0~α 1 ', α
1 ' is, for example, 1.When difference value DIF is less than or equal to negative threshold T H1, halation value α 1 is 0.When difference value DIF is between negative
When between threshold T H1 and positive threshold T H2, halation value α 1 and difference value DIF generally direct proportionality, that is, difference
Value DIF is bigger, then halation value α 1 is bigger.When difference value DIF is more than or equal to positive threshold T H2, halation value α 1 is α 1 '.
It then, is that the wide dynamic range image method of the embodiment of the present invention subtracts image with Fig. 6, Fig. 6 referring to figure 3.
When low halation phenomenon, the curve of calculated maximum value and fine tuning rate schematic diagram.Halation phenomenon is in addition to being relevant to above-mentioned difference
It is also related to the maximum value MAX in low-pass filtering window W2 except different value DIF, therefore, after obtaining halation value α 1, need according to
Halation value α 1 is finely adjusted according to the maximum value MAX in low-pass filtering window W2, to obtain halation specific gravity.Low-pass filtering window W2
Interior maximum value MAX is defined as the maximum value in the low-pass filtering window W2 put centered on pixel CP, that is, MAX=max { Pi
|i∈W2}。
In Fig. 6, fine tuning rate R1 is the numerical value of 0~R1 ', and R1 ' is, for example, 1.Maximum value in low-pass filtering window W2
MAX is less than or equal to positive threshold T H3, then fine tuning rate R1 is R1 '.When the maximum value MAX in low-pass filtering window W2 between
When positive threshold T H3 and TH4, then the fine tuning rate R1 and maximum value MAX in low-pass filtering window W2 is generally inversely proportional, that is,
Maximum value MAX in low-pass filtering window W2 is bigger, then fine tuning rate R1 is lower.Maximum value MAX in low-pass filtering window W2
More than or equal to positive threshold T H4, then fine tuning rate R1 is 0.In addition, according to MAX pairs of maximum value in low-pass filtering window W2
The formula that halation value α 1 is finely adjusted is expressed as α 2=1- (1- α 1) × R1, and wherein α 2 indicates halation specific gravity.
Halation specific gravity α 2 is bigger, then it represents that the halation phenomenon between imagery zone is more serious, on the contrary, halation specific gravity α 2 is got over
It is small, then it represents that the halation phenomenon between imagery zone is slighter.Then, the pixel value of low frequency image is repaired according to halation specific gravity
Just, to obtain the reference value of low frequency image, wherein the formula of the reference value of low frequency image is 2 × LPF of REF=α { Pi|i∈W1}+
(1-α2)×LPF{Pi| i ∈ W2 }, wherein REF indicates the reference value of low frequency image.Briefly, the reference value of low frequency image
REF is according to the calculated low frequency values of the institute of point foundation low-pass filtering window W1, W2 centered on respective pixel and low frequency image
Reference value REF is determined.
It please then continue to referring to Fig. 3, in step S34, the reference value REF in the region according to low frequency image selects image
The area maps curve in the region of (not low-pass filtered), wherein it is bent to can be gamma correction curve, index for area maps curve
Line or linearity curve, the present invention is not limitation with the type of area maps curve in a word.Furthermore, if region only has
One pixel is then selected corresponding to the region of image (not low-pass filtered) with pixel in region corresponding reference value REF
Area maps curve.If there is a plurality of pixels in region, with the corresponding reference value REF of pixels multiple in the region of low frequency image
Average value select area maps curve corresponding to the region of image (not low-pass filtered).
It is the schematic diagram of the multiple regions mapping curve of the embodiment of the present invention referring to Fig. 3 and Fig. 7, Fig. 7.Simply
It says, image (not low-pass filtered) is if be divided into region R1~R3, and in low frequency image, the reference value of region R1~R3 is averaged
Value is different, then region R1~R3 of image (not low-pass filtered) is carried out using different area maps curve LC1~LC3 respectively
Area maps, to promote the brightness of region R1~R3 of image (not low-pass filtered) respectively.In Fig. 2, area maps curve
The corresponding longitudinal axis of LC1~LC3 represents the pixel value of output, and the corresponding horizontal axis of area maps curve LC1~LC3 represents input
Pixel value.
In addition, the situation limited in hardware storage space, it is most slow bent with the area maps of steepest generally only to note down variation
The average value of line and corresponding reference value or reference value, and region corresponding to the average value of other reference values or reference value is reflected
Curve is penetrated, then can be obtained by changing the most slow area maps curve with steepest through the mode of interpolation.Come by taking Fig. 7 as an example
It says, if variation steepest and most slow area maps curve are LC3, LC1, only area maps curve is that LC3, LC1 are right with it
The average value of reference value or reference value is answered to be stored, and area maps curve LC2 is then to penetrate interpolation method to obtain.
Please continue to refer to Fig. 3, after the area maps curve corresponding to the region for having selected image, in step s 35, according to
According to the area maps curve that each region of image is selected, gain map is carried out to each region of image.Furthermore, in step
In S35, wide dynamic range image method can't directly using area mapping curve to the pixel value in the region of image into
Row correction, but corresponding gain curve is obtained according to selected area maps curve, then obtain area according to the gain curve
The yield value of pixel value or brightness in domain.Then, then according to the corresponding brightness acquisition gain adjustment value of pixel value, and according to
Brightness, yield value and gain adjustment value adjust pixel value, to solve the problems, such as in image that color is excessively bright-coloured or colour cast whereby.
It is the corresponding gain song of selected area maps curve of the embodiment of the present invention referring to Fig. 3 and Fig. 8, Fig. 8
The schematic diagram of line.In step S35, the practice explanation of corresponding gain curve is obtained such as according to selected area maps curve
Under, the selected area maps curve LC in some region of image can be corresponding divided by its horizontal axis by the output pixel value on its longitudinal axis
Input pixel value, to obtain the corresponding yield value GN of each input pixel value P whereby, to produce gain curve GNC.
Then, the yield value GN that the pixel value P in the region of image is obtained according to the gain curve, whereby by pixel
Value P is multiplied by yield value GN, to obtain amplified pixel value P ', also that is, P'=P × GN, wherein pixel value P is red, green
With blue pixel value R, G, B.Then, the corresponding brightness of pixel value P is calculated according to red, green and blue pixel value R, G, B
Y, wherein the calculation formula of brightness Y is Y=Cr × R+Cg × G+Cb × B, and wherein Cr, Cg and Cb indicate color gamut conversion coefficient.
Later, referring to Fig. 3 and Fig. 9, Fig. 9 be the embodiment of the present invention wide dynamic range image method to image into
When row gain is amplified, the schematic diagram of the curve of calculated brightness and adjusted value.Because to solve, image is excessively bright-coloured or colour cast
Problem, therefore, it is necessary to go to obtain a gain adjustment value α 3 according to the brightness Y calculated.As shown in figure 9, gain adjustment value α
3 in the corresponding brightness Y of pixel between 0~threshold T H5, generally inversely with brightness Y, wherein be in brightness Y
When 0, the maximum value MA that gain adjustment value α 3 defines for it, and when brightness Y is threshold T H5, gain adjustment value α 3 is fixed for it
The minimum value mA of justice.The physical significance that gain adjustment value α 3 is represented is to solve excessively bright-coloured or colour cast problem ratio, also that is, increasing
Beneficial adjusted value α 3 is bigger, then it represents that it is stronger to solve bright-coloured or colour cast problem ability, conversely, gain adjustment value α 3 is smaller, then and table
Show that bright-coloured or colour cast problem the ability of solution is weaker, pixel value adjusted can be closer to the pixel value before adjustment.
After obtaining gain adjustment value α 3, picture further is adjusted according to brightness Y, yield value GN and gain adjustment value α 3
Plain value P.The calculation formula of pixel value P " adjusted is probably expressed as follows, P "=P' × (1- α 3) 3 × Y' of+α, and wherein Y ' is to put
Brightness after big, that is, Y'=Y × GN.
Other than the above-mentioned practice for carrying out gain map to each region of image, in addition illustrate it is another to each region of image into
The practice of row gain map is as follows.Referring to Fig. 3 and Figure 10, the wide dynamic range image method of Figure 10 embodiment of the present invention
When carrying out gain amplification to image, the schematic diagram of gain amplification is carried out to the pixel of input.As shown in Figure 10, first by the picture of input
Element carries out color gamut conversion, is converted to the domain YUV from the domain RGB, wherein the formula of color gamut conversion is as previously described.
Then, corresponding brightness is obtained according to obtained corresponding gain curve (as shown in Figure 8) using brightness Y as input
The yield value GN of Y.Then, according to the curve (as shown in Figure 9) of brightness and gain adjustment value, the gain tune of corresponding brightness Y is obtained
Whole value α 3.Then, gain adjustment value α 3 is multiplied with yield value GN, to obtain yield value GN ' adjusted, that is, GN'=α 3
×GN。
Later, brightness Y is multiplied by yield value GN, and color difference U and concentration V is multiplied by yield value GN ' adjusted, to obtain
Brightness Y ', color difference U ' and concentration V ' adjusted are obtained, also that is, Y'=Y × GN, U'=U × GN', V'=V × GN'.Finally, will
Brightness Y ' adjusted, color difference U ' and concentration V ' are converted back from the domain YUV as the domain RGB, with obtain red, blue adjusted with it is green
Color pixel value R ', G ', B ', that is, R'=Y'+1.13983 × (V'-128), G'=Y'-0.39465 × (U'-128)-
0.58060 × (V'-128) and R'=Y'+2.03211 × (V'-128).In addition, the above-mentioned domain YUV can also use HIS or HSV
Domain replaces, to sum up, the present invention is not restricted to this.
It should also be noted that if step S31 has and is labeled processing to region before, step S35 more according to brightness, increase
After beneficial value adjusts pixel value with gain adjustment value, further according to area marking processing as a result, again to specific region
Pixel value carry out brightness adjustment.For example, promoting the brightness in the region of face again.
Promote the visibility in dark portion region in image due to, wide dynamic range image method, therefore darker region, it is past
Toward meeting at upper bigger yield value, so that the noise in image increases, therefore whole image may influenced in view of noise
When quality, need further to inhibit the noise of image, to promote visual experience degree.
Image is carried out referring to the wide dynamic range image method that Fig. 3 and Figure 11, Figure 11 are the embodiment of the present invention
When noise reduces, the schematic diagram of the curve of used pixel value and suppression noise value.Because of darker region, often Cheng Shangyue
Big yield value, therefore noise value DNS is then pressed down it is known that pixel value is bigger by the curve of the pixel value of Figure 11 and suppression noise value
It is smaller, on the contrary, pixel value is smaller, then it is bigger to press down noise value DNS.For example, when pixel value is PI1, press down noise value DNS
β 1 is corresponded to, when pixel value is PI2, suppression noise value DNS corresponds to β 2, and wherein PI1 is less than PI2, and β 1 is greater than β 2.
After obtaining the corresponding suppression noise value DNS of pixel value, according to gentle pixel value PsmoothWith suppression noise value DNS adjustment
Each pixel value P for the image that step S35 is generatedWDR, to obtain each pixel value P after lowering image incidental informationDenoised, operation public affairs
Formula can be expressed as PDenoised=DNSPsmooth+PWDR·(1-DNS).Above-mentioned gentle pixel value PsmoothIt is through nn mask
With pixel value PWDRCentered on through volume integral operation and obtain.
Image is carried out referring to the wide dynamic range image method that Fig. 3 and Figure 12, Figure 12 are the embodiment of the present invention
When noise reduces, the schematic diagram that pixel is gently worth is calculated.Wide dynamic range image method can correspond to scene or situation generates mode
Signal MOD, with control multiplexer MUX select corresponding multiple scenes or situation multiple mask MASK1~MASKN wherein it
One.Then, through volume integrator CONV, with pixel value PWDRCentered on take 55 pixel values use selection mask, such as
MASK1 carries out volume integral operation, to obtain gentle pixel value Psmooth.In addition, though above-mentioned mask MASK1~MASKN is with 55
Illustrate for mask, but the present invention does not limit the size of mask.
Image is carried out referring to the wide dynamic range image method that Fig. 1 and Figure 13, Figure 13 are the embodiment of the present invention
From the schematic diagram of the histogram of fatigue resistance pixel value adjusted.Since step S13 carries out area maps to image, shadow
The contrast of picture may change, therefore in the case where considering the universe contrast of image, wide dynamic range image method is more
Step S14 can be executed, image adjust from fatigue resistance.
In Figure 13, the histogram of original image is the histogram HC1 on the left side, and wherein the minimum brightness of image is m1.
After by step S13, the histogram of image is intermediate histogram HC2, and wherein the minimum brightness of image is m2.In step
After S13 is executed, the histogram of image can be offset, therefore will lead to the decline of the universe contrast of image.Therefore, step S14
Purpose is exactly to handle image, so that the histogram of image can become the histogram on the right from intermediate histogram HC2
HC3.The formula adjusted from fatigue resistance can be expressed as Pout=2n-1-(2n-1-PDenoised)·((2n-1-m1)/((2n-1-
M2)), wherein n be pixel value bit number, and PoutFor the pixel value of output.If it should also be noted that the S36 in step S13
It is removed, then above-mentioned pixel value PDenoisedPixel value P must be used insteadWDR。
Please continue to refer to Fig. 1, it should also be noted that above-mentioned steps S12 and S14 can also be removed in other embodiments.Letter
Singly say, the wide dynamic image method of the embodiment of the present invention can only include step S11 and S13, if only carry out step S11 with
S13 reduces the depth of exposure of image, and the image for exported after area maps to image has been able to meet need
It asks, then step S12 and S14 can be removed.It certainly, in other embodiments, can also only removing step S12 and S14 be wherein
One of.
To sum up, the embodiment of the present invention provides a kind of wide dynamic range image method, the wide dynamic range image side
Method can increase the details of image entirety, especially while can allow the details in dark portion region and highlights region in image can
Obviously show.In addition to this, the wide dynamic range image method more can solve traditional wide dynamic range image method and generate
Image incidental information become larger, overall contrast reduction, halation phenomenon, color are excessively bright-coloured in addition colour cast and the details in dark portion region not
Obviously the problems such as (especially face can be excessively dark in backlight).In addition, the computation complexity of the wide dynamic range image method
Less, therefore its runing time is fast, and hardware complexity is low, especially may be implemented in the electronic device with image acquisition function
It is interior.
The above, optimal specific embodiment only of the invention, only feature of the invention is not limited thereto, any ripe
Know this those skilled in the art within the field of the present invention, can think easily and changes or modifications, can all cover the patent in following this case
Range.
Claims (10)
1. a kind of wide dynamic range image method, comprising:
One area maps are carried out to an image through multiple regions mapping curve, to carry out incomplete one for multiple regions in image
The gamma correction of sample, wherein the region at least includes a pixel;
Wherein the area maps include:
The image is filtered, to obtain a low frequency image;
A reference value in each region of the image is obtained according to the low frequency image;
The reference value in the region according to the image corresponds to the image from one of those area maps Curve selections
The region;And
The area maps curve according to the region for corresponding to the image obtains a corresponding gain curve, according to gain song
Line obtains a yield value of a pixel value in the region or a yield value of a brightness, obtains according to the corresponding brightness of the pixel value
A gain adjustment value is obtained, and the pixel value is adjusted according to the brightness, the yield value and the gain adjustment value.
2. wide dynamic range image method according to claim 1, wherein penetrating those area maps curves to the shadow
As further including before carrying out the area maps:
One universe mapping is carried out to the image through a universe mapping curve.
3. wide dynamic range image method according to claim 1, wherein penetrating those area maps curves to the shadow
As further including after carrying out the area maps:
It carries out one to the image to adjust from fatigue resistance, to increase a contrast of the image.
4. wide dynamic range image method according to claim 1, wherein obtaining each area of the image according to the low frequency image
The reference value in domain includes:
Calculate a difference value of each pixel value in the region of the low frequency image under two different low-pass filtering windows;
A halation phenomenon specific gravity is determined according to the difference value and lower calculated maximum value of one of low-pass filtering window;
And
The corresponding reference value of the pixel value is determined through the halation phenomenon specific gravity.
5. wide dynamic range image method according to claim 1, wherein the area maps are further included:
One area marking processing is carried out to the image, to mark an at least region;And
After adjusting those pixel values, it is adjusted again for the pixel value in the region of mark.
6. wide dynamic range image method according to claim 1, wherein after adjusting those pixel values, the area maps
It further includes:
A suppression noise value is determined according to the reference value in the region;And
The noise for lowering the image according to the one of them being selected in the suppression noise value and multiple masks.
7. wide dynamic range image method according to claim 1, wherein obtaining being somebody's turn to do for the region according to the gain curve
The yield value is multiplied by the brightness and the pixel value to obtain the amplified brightness and the pixel by the yield value of pixel value
Value, and the pixel value is adjusted according to the gain adjustment value, the amplified brightness and the pixel value.
8. wide dynamic range image method according to claim 1, wherein a color gamut conversion will be carried out to the pixel value, with
The corresponding brightness of the pixel value, a color difference and a concentration are obtained, obtains being somebody's turn to do for the brightness in the region according to the gain curve
The yield value is multiplied by the brightness to obtain the amplified brightness, the gain adjustment value and the yield value is multiplied by by yield value
The color difference and the concentration, to obtain the amplified color difference and the concentration, to the amplified brightness, the color difference and the concentration into
Another color gamut conversion of row, to obtain the pixel value adjusted.
9. wide dynamic range image method according to claim 6, wherein should using selected centered on the pixel value
Mask carries out a roll of integral operation, to calculate a gentle pixel value, and according to the gentle pixel value, the pixel value and the suppression
Noise value adjusts the pixel value, to reduce the noise of the image.
10. wide dynamic range image method according to claim 3, wherein one when obtaining the image without any processing
First minimum luminance value, and obtain the image those pixel values be adjusted after one second minimum luminance value, Yi Jigen
The pixel value is adjusted according to a bit number of first minimum luminance value, second minimum luminance value and the pixel value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510390387.0A CN106341613B (en) | 2015-07-06 | 2015-07-06 | Wide dynamic range image method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510390387.0A CN106341613B (en) | 2015-07-06 | 2015-07-06 | Wide dynamic range image method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106341613A CN106341613A (en) | 2017-01-18 |
CN106341613B true CN106341613B (en) | 2019-05-14 |
Family
ID=57825963
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510390387.0A Active CN106341613B (en) | 2015-07-06 | 2015-07-06 | Wide dynamic range image method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106341613B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI659396B (en) * | 2017-08-31 | 2019-05-11 | Yuan Ze University | Method and image processing device for image reconstruction in multimodal noise suppression |
CN107591141B (en) * | 2017-09-28 | 2019-12-03 | 武汉华星光电技术有限公司 | Liquid crystal display driving method, system and computer-readable medium |
CN108510560B (en) * | 2018-04-11 | 2020-01-24 | 腾讯科技(深圳)有限公司 | Image processing method, image processing device, storage medium and computer equipment |
CN112488972A (en) * | 2020-11-27 | 2021-03-12 | 深圳卡多希科技有限公司 | Method and device for synthesizing green screen image and virtual image in real time |
CN113099200B (en) * | 2021-06-10 | 2021-09-03 | 深圳市火乐科技发展有限公司 | Method, apparatus, storage medium, and electronic device for processing video signal |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103295194A (en) * | 2013-05-15 | 2013-09-11 | 中山大学 | Brightness-controllable and detail-preservation tone mapping method |
CN103729873A (en) * | 2013-12-31 | 2014-04-16 | 天津大学 | Content-aware ambient light sampling method |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8203699B2 (en) * | 2008-06-30 | 2012-06-19 | Microsoft Corporation | System architecture design for time-of-flight system having reduced differential pixel size, and time-of-flight systems so designed |
TW201120402A (en) * | 2009-12-03 | 2011-06-16 | Ind Tech Res Inst | System and method for contructing high dynamic range images |
-
2015
- 2015-07-06 CN CN201510390387.0A patent/CN106341613B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103295194A (en) * | 2013-05-15 | 2013-09-11 | 中山大学 | Brightness-controllable and detail-preservation tone mapping method |
CN103729873A (en) * | 2013-12-31 | 2014-04-16 | 天津大学 | Content-aware ambient light sampling method |
Also Published As
Publication number | Publication date |
---|---|
CN106341613A (en) | 2017-01-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106341613B (en) | Wide dynamic range image method | |
US8131103B2 (en) | Image detail enhancement | |
US9076218B2 (en) | Method and image processing device for image dynamic range compression with local contrast enhancement | |
JP5105209B2 (en) | Image processing apparatus and method, program, and recording medium | |
US7020332B2 (en) | Method and apparatus for enhancing a digital image by applying an inverse histogram-based pixel mapping function to pixels of the digital image | |
RU2298226C1 (en) | Method for improving digital images | |
US7899267B2 (en) | Dynamic range compensation by filter cascade | |
US8374430B2 (en) | Apparatus and method for feature-based dynamic contrast enhancement | |
JP6602789B2 (en) | System and method for local contrast enhancement | |
TWI558207B (en) | Wide dynamic rage imaging method | |
CN101505421A (en) | Method of high dynamic range compression with detail preservation and noise constraints | |
JP6004757B2 (en) | Image processing apparatus and image processing method | |
AU767050B2 (en) | Contour correction device | |
CN111292269B (en) | Image tone mapping method, computer device, and computer-readable storage medium | |
CN109447912B (en) | Fluorescent image self-adaptive enhancement and noise reduction method of fluorescent navigation endoscope system | |
CN108280836B (en) | Image processing method and device | |
WO2006006157A2 (en) | Automatic adaptive gamma correction | |
JP4212741B2 (en) | Image processing device | |
Albu et al. | One scan shadow compensation and visual enhancement of color images | |
CN115239570A (en) | Image processing method, image processing apparatus, and storage medium | |
EP2446415A2 (en) | Digital image processing method and device for lightening said image | |
JP6335614B2 (en) | Image processing apparatus, control method thereof, and program | |
JP2006519447A (en) | Variable contrast mapping of digital images | |
KR20160025876A (en) | Method and apparatus for intensificating contrast in image | |
US11625817B2 (en) | Pyramid-based tone mapping |
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 |