CN103578081B - For improving the image processing method and system of picture quality - Google Patents
For improving the image processing method and system of picture quality Download PDFInfo
- Publication number
- CN103578081B CN103578081B CN201210280809.5A CN201210280809A CN103578081B CN 103578081 B CN103578081 B CN 103578081B CN 201210280809 A CN201210280809 A CN 201210280809A CN 103578081 B CN103578081 B CN 103578081B
- Authority
- CN
- China
- Prior art keywords
- pixel
- noise
- image
- filter
- influence
- 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
- 238000003672 processing method Methods 0.000 title claims abstract description 14
- 238000000034 method Methods 0.000 claims abstract description 40
- 238000001914 filtration Methods 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims description 92
- 230000002708 enhancing effect Effects 0.000 claims description 42
- 238000000605 extraction Methods 0.000 claims description 26
- 238000010276 construction Methods 0.000 claims description 24
- 238000012014 optical coherence tomography Methods 0.000 claims description 23
- 239000000284 extract Substances 0.000 claims description 15
- 238000012360 testing method Methods 0.000 claims description 2
- 230000001965 increasing effect Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 230000002146 bilateral effect Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000009792 diffusion process Methods 0.000 description 2
- 230000005484 gravity Effects 0.000 description 2
- 239000004615 ingredient Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012805 post-processing Methods 0.000 description 2
- 238000011946 reduction process Methods 0.000 description 2
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 230000000116 mitigating effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10101—Optical tomography; Optical coherence tomography [OCT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20004—Adaptive image processing
- G06T2207/20012—Locally adaptive
Abstract
It provides a kind of for improving the image processing method of picture quality, the described method includes: the different degrees of influence of noise that (a) generates each pixel based on the image pixel intensities away from the pixel at each pixel different distance in image, constructs noise filter for each pixel according to the line noise model of image；(b) noise reduction filtering is carried out to each pixel respectively by using the noise filter for the building of each pixel, to obtain noise-reduced image.
Description
Technical field
This application involves image processing techniques, more particularly, are related to a kind of for improving the image procossing of picture quality
Method.
Background technique
Recently, as the continuous development of image processing techniques and user are to high quality, the demand of high-definition image
Increase, it has been proposed that a variety of image processing methods for improving picture quality.In these image processing methods, how effectively to go
Except the various noises in image and improves image processing speed and have become research emphasis.
For example, optical coherence tomography (OCT) has become one kind with micron-sized resolution in existing image processing techniques
The new method that the intracorporal institutional framework of people is imaged in rate.This technology has been successfully applied for ophthalmology, but for
There is problem in application in the compact tissues such as dermatology: due to higher optical attenuation and Multiple Scattering, so that can
Biggish noise can be introduced.This noise will lead to Optical coherence tomography with lower dynamic range and lower letter
It makes an uproar and compares, and make it difficult to differentiate the details ingredient of the OCT image obtained from dense tissue.In this case it is necessary to a kind of
The image processing method of the noise in OCT image can be effectively reduced.In addition it is also necessary to which one kind can enhance the optics phase
The image processing method of the picture quality of dried layer analysis system, to provide more tissue information.
Noise in OCT image generally includes two kinds of noise: noncoherent noise and speckle noise.General pre-
In processing method, noncoherent noise can by system optimization be suppressed, and speckle noise due to include tissue information and
The elimination of speckle noise needs other hardware and scan process, it is thus possible to lead to soft edge and can reduce imaging
Speed.Therefore, be the influence for eliminating speckle noise, propose many post-processing approach (such as using a large amount of digital filter)
To avoid the soft edge effect for reducing image taking speed and mitigation occurs in the pretreatment of image.In the post-processing side
Filter used in method generally includes linear filter, nonlinear filter, diffusion filter and wavelet filter.Linear filter
Wave device and wavelet filter are to be applied on the basis of assuming that speckle noise model has multiplication form, therefore usually may be used
The influence of noncoherent noise can be ignored and set 0 for noncoherent noise.And nonlinear filter and diffusion filter are then bases
Inhibit noise in local image characteristics.These four existing filters are able to suppress noise.However, in existing filter
In still have following limitation: (1) need weighed between the noise reduction effect and processing speed of filter, it is effective
Despeckle filter may be not suitable for carrying out scan picture, and (2) are using the physical model of these filters usually by non-phase
Dry noise is considered as 0 and does not consider the influence of noncoherent noise, and (3) can make image one using the noise reduction process of these filters
Determine to thicken in degree and influence is presented on the details of image.
Summary of the invention
The purpose of the present invention is to provide one kind can effectively remove the influence of noise in image to improve picture quality
Image processing system and method.
According to an aspect of the present invention, it provides a kind of for improving the image processing method of picture quality, the method
It include: that (a) generates not each pixel based on the image pixel intensities away from the pixel at each pixel different distance in image
With the influence of noise of degree, noise filter is constructed for each pixel according to the line noise model of image；(b) by using
Noise reduction filtering is carried out to each pixel respectively for the noise filter of each pixel building, to obtain noise-reduced image.
The influence of noise that image pixel intensities away from the closer pixel of each pixel generate each pixel is bigger, away from
The influence of noise that the image pixel intensities of the remoter pixel of each pixel generate each pixel is smaller.
In step (a), for the filter f of the pixel i building in imageiMeeting can following equation (1):
Wherein, giIndicate the actual pixels intensity of the pixel i directly acquired from described image,WithIt respectively indicates and passes through meter
The different degrees of influence of noise that the pixel for pixel i in the preset range centered on pixel i is generated is calculated, acquisition is directed to picture
The first noise weighting coefficient and the second noise weighting coefficient of plain i.
Equation (2) can will be used calculatedIt is determined as the first noise weighting coefficient:
Equation (3) can will be used calculatedIt is determined as the second noise weighting coefficient:
Wherein, ωiAnd ωjRespectively indicate the picture of all pixels in the region of the predefined size centered on pixel i and j
Element set, | ωi| and | ωj| respectively indicate pixel set ωiAnd ωjIn pixel quantity, gkExpression is directly acquired from image
Pixel k actual pixels intensity,It indicates in pixel set ωjIn all pixels actual pixels intensity mean value,It indicates in pixel set ωjIn all pixels actual pixels intensity variance, ε is a in order to preventjIt is excessive and preparatory
Determining regularization parameter, βjIt is to indicate ajAnd bjIt is rightWithNoise contribution weighting coefficient and can be determined in advance.
The method may also include that (c) by the high fdrequency component in enhancing noise-reduced image to obtain enhancing image.
Step (c) can include: (c1) in a frequency domain, is divided by using the noise filter for the building of each pixel
Other each pixel in noise-reduced image is further filtered, and the low frequency component of the noise-reduced image is extracted；(c2) it is based on
The low frequency component of the noise-reduced image and the noise-reduced image of extraction, extracts the high fdrequency component of the noise-reduced image；(c3)
High fdrequency component and low frequency component to the noise-reduced image of extraction carry out linear combination and increase the weight of high fdrequency component, to be increased
Strong image.
In step (c2), the high fdrequency component p of noise-reduced image can be extracted by following equation (4)high:
phigh=p-plow (4)
Wherein, p indicates the noise-reduced image in the frequency domain obtained in step (b), plowIndicate the institute extracted in step (c1)
State the low frequency component of noise-reduced image.It, can be according to following equation (5) to the low frequency point extracted in step (c1) in step (c3)
Amount and the high fdrequency component extracted in step (c2) carry out linear combination, enhance image p to obtainenhanced:
penhanced=λ phigh+plow (5)
Wherein, λ is predetermined weighted value and is greater than 1.
λ can be predefined by testing.
Described image can be optical coherence tomography (OCT) image.
According to another aspect of the present invention, it provides a kind of for improving the image processing system of picture quality, the system
System includes: filter construction unit, for based on the image pixel intensities away from the pixel at each pixel different distance in image to institute
The different degrees of influence of noise that each pixel generates is stated, constructs noise reduction for each pixel according to the line noise model of image
Filter；Noise reduction processing unit, for by using the noise filter for being directed to the building of each pixel by filter construction unit
Noise reduction filtering is carried out to each pixel respectively, to obtain noise-reduced image.
The influence of noise that image pixel intensities away from the closer pixel of each pixel generate each pixel is bigger, away from
The influence of noise that the image pixel intensities of the remoter pixel of each pixel generate each pixel is smaller.
By filter construction unit for the filter f of the pixel i building in imageiFollowing equation (1) can be met:
Wherein, giIndicate the actual pixels intensity of the pixel i directly acquired from described image,WithIt respectively indicates and passes through
The different degrees of influence of noise that the pixel for pixel i in the preset range centered on pixel i is generated is calculated, acquisition is directed to
The first noise weighting coefficient and the second noise weighting coefficient of pixel i.
Equation (2) can will be used calculatedIt is determined as the first noise weighting coefficient:
Equation (3) can will be used calculatedIt is determined as the second noise weighting coefficient:
Wherein, ωiAnd ωjRespectively indicate the picture of all pixels in the region of the predefined size centered on pixel i and j
Element set, | ωi| and | ωj| respectively indicate pixel set ωiAnd ωjIn pixel quantity, gkExpression is directly acquired from image
Pixel k actual pixels intensity,It indicates in pixel set ωjIn all pixels actual pixels intensity mean value,It indicates in pixel set ωjIn all pixels actual pixels intensity variance, ε is a in order to preventjIt is excessive and preparatory
Determining regularization parameter, βjIt is to indicate ajAnd bjIt is rightWithNoise contribution weighting coefficient and can be determined in advance.
The system may also include that picture quality enhancement unit, for by enhancing noise-reduced image in high fdrequency component come
Obtain enhancing image.
Described image quality enhancement unit can include: low frequency component extraction unit, in a frequency domain by using by filtering
Wave device construction unit carries out each pixel in noise-reduced image for the noise filter of each pixel building further
Filtering, extract the low frequency component of the noise-reduced image；High fdrequency component extraction unit, for being based on the noise-reduced image and mentioning
The low frequency component of the noise-reduced image taken extracts the high fdrequency component of the noise-reduced image；Image processing unit, for extraction
Noise-reduced image high fdrequency component and low frequency component carry out linear combination and increase the weight of high fdrequency component, to obtain enhancing figure
Picture.
The high fdrequency component extraction unit can extract the high fdrequency component p of noise-reduced image by following equation (4)high:
phigh=p-plow (4)
Wherein, p indicates the noise-reduced image in the frequency domain obtained by noise reduction processing unit, plowExpression is mentioned by low frequency component
The low frequency component for the noise-reduced image for taking unit to extract.Image processing unit can be according to following equation (5) to by low frequency component
The low frequency component and carry out linear combination by the high fdrequency component that high fdrequency component extraction unit extracts that extraction unit extracts, are increased with obtaining
Strong image penhanced:
penhanced=λ phigh+plow (5)
Wherein, λ is predetermined weighted value and is greater than 1.
Beneficial effect
By using image processing system and method for the invention, the speckle noise in image can be more efficiently removed
And noncoherent noise.Additionally it is possible to farthest retain the details of image.In addition, with existing Typical Digital filter
Compare (wavelet filter, two-sided filter, Wiener filter, median filter etc.), on the basis of guaranteeing treatment effect,
It also has extraordinary performance in image processing speed, is advantageously implemented the real-time processing to image.In addition, of the invention
Image processing system and method can obtain better effect when handling OCT image.
Detailed description of the invention
By the description carried out with reference to the accompanying drawing, above and other purpose of the invention and feature will become more clear
Chu, in which:
Fig. 1 is the block diagram for showing the image processing system of an exemplary embodiment of the present invention；
Fig. 2 is the block diagram for showing image processing system in accordance with an alternative illustrative embodiment of the present invention；
Fig. 3 A to Fig. 3 C is shown using image processing system according to an exemplary embodiment of the present invention to original input picture
The processing result image for being handled and being obtained；
Fig. 4 A to Fig. 4 K show will use image processing system according to an exemplary embodiment of the present invention to image at
Reason and the comparison diagram for handle to image processing result image obtained using existing filter；
Fig. 5 is the flow chart for showing the image processing method of an exemplary embodiment of the present invention；
Fig. 6 is the flow chart for showing image processing method in accordance with an alternative illustrative embodiment of the present invention.
Specific embodiment
Hereinafter, the exemplary embodiment that the present invention is more fully described with reference to the accompanying drawings, its example is shown in the drawings.So
And it can be embodied in many different forms exemplary embodiment, and the present invention should not be construed as being limited to illustrate herein
Exemplary embodiment.On the contrary, these embodiments are provided so that the disclosure will be thorough and complete, and will fully will be exemplary
The range of embodiment is communicated to those skilled in the art.In the accompanying drawings, identical label indicates identical part.
Fig. 1 is the block diagram for showing the image processing system 100 of an exemplary embodiment of the present invention.
As shown in Figure 1, the image processing system 100 of an exemplary embodiment of the present invention includes filter construction unit
110 and noise reduction processing unit 120.
Referring to Fig.1, filter construction unit 110 can be used for based in image away from the pixel at each pixel different distance
The different degrees of influence of noise that image pixel intensities generate each pixel, according to the line noise model of image for each
Pixel constructs noise filter.An exemplary embodiment of the present invention, the pixel away from the closer pixel of each pixel are strong
It is bigger to spend the influence of noise generated to each pixel, the image pixel intensities away from the remoter pixel of each pixel are to described every
The influence of noise that a pixel generates is smaller.Noise reduction processing unit 120 can be used for by using the noise reduction for the building of each pixel
Filter carries out noise reduction filtering to each pixel respectively, to obtain noise-reduced image.
In particular, filter construction unit 110 according to the present invention is based on following design come for each of image
Pixel constructs noise filter: (1) the noise size at specific pixel is not only related with the image pixel intensities of the pixel itself, also
It is related with the image pixel intensities of pixel in surrounding neighbors；(2) pixel around specific pixel is made an uproar to caused by the specific pixel
During sound shadow is rung, specific gravity shared by the influence of noise that the image pixel intensities away from the closer pixel of the specific pixel generate the specific pixel
It is bigger, and specific gravity shared by the influence of noise that the image pixel intensities away from the remoter pixel of the specific pixel generate the specific pixel is more
It is small.Therefore, filter construction unit according to the present invention, can be based in image when constructing noise filter for each pixel
The image pixel intensities of pixel around each pixel and distance away from each pixel are that each pixel constructs noise filter, from
And can it is more efficient and targetedly remove image in the influence of noise due to caused by neighborhood pixels (for example, speckle is made an uproar
Sound).
Method illustrated below for each pixel building noise filter in image.
In general, speckle noise belongs to Multiplicative random noise, noncoherent noise belongs to Complex-valued additive random noise.Therefore, false
If fiIndicate the ideal pixel intensity of the pixel i in image, giThe actual pixels intensity for indicating the pixel i in image, then according to figure
The line noise model of picture, fiAnd giThere can be following relationship:
gi=fiui+ni (1)
Wherein, uiIt is the speckle noise in pixel i, niIt is the noncoherent noise in pixel i and is additivity, static Gauss
Noise.It can be seen that from equation (1), the process being filtered to image is practical to be can be considered as each pixel in image,
From the actual pixels intensity of the pixel (for example, being directed to the actual pixels intensity g of pixel ii) in extract ideal, not Noise
Ideal pixel intensity (for example, be directed to pixel i ideal pixel intensity fi) process.It is clear that can be held according to equation (1)
Easily obtain the ideal pixel intensity f for pixel ii, as shown in equation (2):
fi=aigi+bi (2)
Wherein, ai=1/ui, bi=-ni/ui。
Therefore pixel i building noise filter can be directed to according to equation (2).That is, for the drop of pixel i building
Make an uproar filter function fiMeet equation (2), to work as the actual pixels intensity g of pixel iiPass through the noise filter as input
When, the output of the noise filter is the ideal pixel intensity f of required pixel ii.In addition, can be by aiAnd biIt is respectively seen as
For the filter function f of pixel i buildingiThe first noise coefficient and the second noise coefficient.Due to giIt can directly be obtained from image
It and is known, therefore be really to determine for pixel i building for the process of the pixel i building noise filter in image
Filter fiIn coefficient aiAnd biProcess.
In the prior art, the noise level for having proven to speckle noise is usually proportional to local gray level grade.That is,
Brighter region is easier to be influenced by speckle noise than darker region.This characteristic of speckle noise makes general figure
As processing method is more complicated.But it under reasonable hypothesis, can make an uproar from the intensity that the local strength of noisy acoustic image obtains speckle
Sound.Therefore, the noise filter for pixel i can be constructed according to the region of predefined size is unit, that is, consider with pixel
Influence of noise that other pixel for pixel i in the region of predefined size centered on i is generated constructs the noise reduction for pixel i
Filter.
Assuming that ωiIt is the region of the predefined size centered on pixel i (for example, the predefined size centered on pixel i
Square area) in all pixels pixel set.It can be based on pixel set ωiIn all pixels actual pixels intensity
To determine the noise filter f for pixel i buildingiIn noise coefficient aiAnd bi。
In an exemplary embodiment in accordance with the present invention, as an example, in pixel set ωiIn domain, it can make following
E (a in equation (3)i,bi) it is minimum when aiAnd biIt is determined as the noise filter f according to the present invention for pixel iiIt is used
Noise coefficient aiAnd bi:
Wherein, ε is a in order to preventiExcessive and predetermined regularization parameter.It can be made by linear regression to calculate
E (a in formula (3)i,bi) it is minimum when aiAnd bi。
As shown in equation (4) and equation (5), | ωi| indicate pixel set ωiIn pixel quantity,It indicates
Pixel set ωiIn all pixels actual pixels intensity mean value, gkIndicate the pixel set ω directly acquired from imagei
The actual pixels intensity of middle pixel k,It indicates in pixel set ωiIn all pixels actual pixels intensity variance.From
Equation (4) and equation (5) are it is clear that coefficient aiAnd biIt can actually reflect the pixel in the preset range centered on pixel i
(that is, pixel set ωiIn pixel) to pixel i generate influence of noise.
It is found that pixel set ω from equation (4) and equation (5)iIn all pixels for linear coefficient aiAnd biHave
Identical weighted influence.However, in pixel set ωiIn, the image pixel intensities away from pixel i closer pixel generate pixel i
Influence of noise can be bigger, and the image pixel intensities away from pixel i remoter pixel can be smaller to the pixel i influence of noise generated.Therefore, it is
The different degrees of influence of noise that the pixel for pixel i away from pixel i different distance is generated is embodied, it can be to pixel set ωiIn it is every
Weighted calculation is done in the influence of noise that a pixel for pixel i is generated.Firstly, for ωiIn some pixel j (j ≠ i), can be with j
Centered on form new and ωiThe identical or different neighborhood territory pixel set ω of sizej, and according to formula (4) in this neighborhood
(5) the noise coefficient a for pixel j is acquiredj、bj.Pixel set ω is directed in this manner it is achieved that can getiInterior each pixel
Noise coefficient, then will be directed to ωiInterior each pixel and the noise coefficient acquired are weighted, to be directed to
The noise filter f of pixel i buildingiThe first noise weighting coefficientWith the second noise weighting coefficient
As shown in equation (6) and equation (7), equation (6) can will be used calculatedIt is determined as the first noise weighting system
Number:
Equation (7) can will be used calculatedIt is determined as the second noise weighting coefficient:
Wherein, βjIt is to indicate ajAnd bjIt is rightWithNoise contribution weighting coefficient and can be determined in advance.This implementation
In example, for convenience of explanation, β is takenj=1 is illustrated as example, and in this case, it is triangle weighting that the present embodiment, which uses,
Method is weighted processing.It should be understood that different methods of weighting also can be used in the present invention, for example, Gauss weighting method.Because from
On equation (6) and (7) in it is clear that when calculate for pixel i building noise filter fiThe first noise weighting system
NumberWith the second noise weighting coefficientWhen, the actual pixels intensity away from pixel i remoter pixel is counted in above-mentioned read group total
The number of calculation is fewer (that is, shared weight is lower), so as to reflect that the pixel for pixel i away from pixel i different distance is generated
Different degrees of influence of noise.
Therefore, the filter construction unit 110 of an exemplary embodiment of the present invention, image processing system 100 can be pressed
The noise filter f of pixel i is directed to according to following equation (8) buildingi:
Wherein, giIndicate the actual pixels intensity for the pixel i that can be directly acquired from described image,WithIt respectively indicates logical
It crosses and calculates the different degrees of influence of noise that the pixel for pixel i in the preset range centered on pixel i is generated, the needle of acquisition
To the first noise weighting coefficient and the second noise weighting coefficient of pixel i.
After for each pixel building noise filter in image, an exemplary embodiment of the present invention can
Noise reduction according to equation (8) for each pixel building in image is used by the noise reduction processing unit 120 of image processing system 100
Filter carries out noise reduction filtering to each pixel respectively in a frequency domain, to obtain noise-reduced image.
However, it will be appreciated by those skilled in the art that filter construction unit 110 constructs the tool of noise filter
Gymnastics is not limited to mode used in the description, can construct noise reduction filtering by various modeling methods known in the art
Device, for example, can be estimated based on the intensity of optical transport model etc..Further, it is contemplated that away from the picture at specific pixel different distance in image
Element is weighted the different degrees of influence of noise that the specific pixel generates come the coefficient of the noise filter to building
Method is also not necessarily limited to the above method, and also such as the methods of Gauss weighting can be used to be weighted place for filter construction unit 110
Reason.
Fig. 2 is the block diagram for showing image processing system 200 in accordance with an alternative illustrative embodiment of the present invention.The figure of Fig. 2
As processing system 200 can be used for obtaining the enhancing image that details/contrast enhances.
As shown in Figure 2, image processing system 200 according to the present invention includes filter construction unit 210, noise reduction process
Unit 220 and picture quality enhancement unit 260.At the filter construction unit 210 and noise reduction of the image processing system 200 of Fig. 2
It is corresponding with the filter construction unit 110 of the image processing system of Fig. 1 100 and noise reduction processing unit 120 respectively to manage unit 220,
And serve identical, therefore, it will be not described in detail herein.Picture quality enhancement unit 260 can be used for passing through increasing
High fdrequency component in strong noise-reduced image enhances image (that is, image of picture quality enhancing) to obtain.In particular, due to image
In details ingredient and image border display effect it is usually related to the high fdrequency component of image, therefore can pass through enhancing noise reduction figure
High fdrequency component as in improves details in noise-reduced image and contrast performance, so that image is more clear, to be had
There is the output image of more high image quality.
Referring to Fig. 2, in an exemplary embodiment of the present invention, picture quality enhancement unit 260 may include that low frequency component mentions
Take unit 230, high fdrequency component extraction unit 240 and image processing unit 250.
An exemplary embodiment of the present invention can pass through after obtaining noise-reduced image via noise reduction processing unit 220
Low frequency component extraction unit 230, high fdrequency component extraction unit 240 and image processing unit in picture quality enhancement unit 260
250 pairs of images carry out further picture quality enhancings processing, with obtain picture quality enhancing enhancing image (for example, details/
The image of contrast enhancing).Compared with the image processing system 100 shown in Fig. 1, image processing system 200 shown in Figure 2
The picture quality enhancement unit 260 for enhancing picture quality is additionally increased, therefore, the image processing system 200 of Fig. 2 can
It is considered to be the enhanced of the image processing system 100 of Fig. 1.
In particular, only as an example, low frequency component extraction unit 230 can in a frequency domain by using in the present invention by filtering
Wave device construction unit 210 carries out each pixel in noise-reduced image p for the noise filter of each pixel building further
Filtering, to extract the low frequency component p of the noise-reduced imagelow。
High fdrequency component extraction unit 240 can the low frequency based on the noise-reduced image p and the extracted noise-reduced image
Component plow, extract the high fdrequency component p of the noise-reduced imagehigh。
The high fdrequency component p of noise-reduced image can be extracted by following equation (9)high:
phigh=p-plow (9)
Wherein, p indicates the noise-reduced image in the frequency domain obtained by noise reduction processing unit 220, plowIt indicates by low frequency component
The low frequency component for the noise-reduced image p that extraction unit 230 extracts.
The low of noise-reduced image p is being extracted by low frequency component extraction unit 230 and high fdrequency component extraction unit 240 respectively
Frequency component plowWith high fdrequency component phighLater, can by 250 pairs of image processing unit extract noise-reduced image high fdrequency component and
Low frequency component carries out linear combination, increases the weight of high fdrequency component, to obtain the enhancing image p of picture quality enhancingenhanced.?
It can be as shown in following equation (10) to the low frequency component p extracted by low frequency component extraction unit 230 in the present inventionlowWith by high frequency
The high fdrequency component p that component extraction unit 240 extractshighLinear combination is carried out, enhances image p to obtainenhanced:
penhanced=λ phigh+plow (10)
Wherein, λ is predetermined weighted value and can be predefined by the method for experiment, in the present invention, considers
To that can be performed better than by increasing the weight of the high fdrequency component in noise-reduced image come the details and contrast for making noise-reduced image, therefore
λ is normally provided as being greater than 1.
By using image processing method according to the present invention, the quality of image can be effectively improved.However, should manage
Solution, the method that the enhancing image of picture quality enhancing is obtained by the high fdrequency component in enhancing noise-reduced image are not limited to above retouch
The method stated, picture quality enhancement unit 260 also can be used various method for enhancing image quality well known in the art defeated to improve
The picture quality of image out.
It is illustrated respectively by using image processing system 100 and 200 of the invention and conventional filter pair below
The Comparative result obtained that OCT image is handled.
Fig. 3 A to Fig. 3 C be show image processing system using an exemplary embodiment of the present invention to OCT image into
The processing result image of row processing and acquisition.
Fig. 3 A shows the OCT image being originally inputted, and Fig. 3 B is shown to be carried out using image processing system 100 shown in Fig. 1
The noise-reduced image of processing and acquisition, Fig. 3 C show the increasing for being handled and being obtained using image processing system 200 shown in Figure 2
Strong image.
Fig. 3 A is compared with Fig. 3 B, it is therefore apparent that the noise (including speckle noise) in image is effectively suppressed.So
And compared with the OCT image being originally inputted shown in Fig. 3 A, the edge and details of image are also omited during going speckle to handle
It is micro- smooth.This is common fault present in digital filtering method.
As a comparison, by using the image processing system 200 of an exemplary embodiment of the present invention shown in Figure 2
Further picture quality enhancing processing is carried out to the image of noise reduction, can get the enhancing image of details and contrast enhancing, such as
Shown in Fig. 3 C.
Four kinds of indexs can be passed through: PSNR (Y-PSNR), SNR (signal-to-noise ratio), SSIM (structural similarity) and SNRs are (single
Width signal noise ratio (snr) of image) evaluate picture quality.It is shown in table 1 for the OCT image, noise-reduced image and enhancing image being originally inputted
The indices value of estimation.
Table 1
By noise-reduced image and the SNRs value for enhancing image compared with the SNRs value for the OCT image being originally inputted, it can be seen that
After noise reduction filtering and picture quality enhancing processing, noise is effectively suppressed.Although compared with noise-reduced image, in image matter
The enhancing noise in image obtained after amount enhancing processing increases and SNRs is reduced, but other indexs PSNR, SNR and SSIM is equal
It is further augmented.This means that the picture quality enhancing processing that image processing system 200 uses can effectively restore to original
Begin the image detail lost during the OCT image inputted progress noise reduction.In addition, for for estimating marginal definition and more
The parameter SSIM for focusing on human eye vision, in enhancing image, SSIM is enhanced to 0.9, it means that is passing through picture quality
After enhancing processing, it can be reserved for more image details and marginal portion and make the visuality of enhancing image more preferably.
Fig. 4 A to Fig. 4 K, which is shown, will use the image processing system 100 and 200 pair of an exemplary embodiment of the present invention
OCT image carries out processing and shows with the comparison for handle processing result image obtained to OCT image using existing filter
Figure.
Fig. 4 A shows the OCT image being originally inputted, Fig. 4 B and Fig. 4 C be shown respectively using wiener (Wiener) filter and
The noise-reduced image and enhancing image for being handled image and being obtained, Fig. 4 D and Fig. 4 E are shown respectively to be filtered using intermediate value (Median)
The noise-reduced image and enhancing image, Fig. 4 F and Fig. 4 G that wave device handles image and obtains are shown respectively using homomorphism
(Homomorphic) noise-reduced image and enhancing image, Fig. 4 H and Fig. 4 I that filter handles image and obtains show respectively
Out using the noise-reduced image and enhancing image, Fig. 4 J and figure that bilateral (Bilateral) filter handles image and obtains
4K, which is shown respectively, to be handled image using image processing system 100 and image processing system 200 of the invention and is obtained
Noise-reduced image and enhancing image.
Although existing Wiener filter, median filter shown in Fig. 4 B to Fig. 4 I, homomorphic filter and bilateral filter
Wave device etc. can inhibit the noise in image, but these filters also put down the detail section of image
It is sliding.Fig. 4 B to Fig. 4 K is compared with the OCT image of Fig. 4 A being originally inputted respectively, image processing system according to the present invention
100 and image processing system 200 have and better image quality reinforcing effect and save more image details (although Fig. 4 E
Picture quality also relatively preferably, but compared with the OCT image being originally inputted, there are some bright spots).
The carry out image using above-mentioned existing filter and image processing system 100 and 200 of the invention is shown in table 2
Handle the comparison of the indices value of obtained image.
Table 2
As can be seen from Table 2, it is carried out in the step of noise reduction is to obtain noise-reduced image to image, at image of the invention
Reason system 100 and 200 and homomorphic filter and Wiener filter have good realization on the processing time, thus be suitble into
Row processing in real time.In addition, having also obtained highest SNRs and SNR value using image processing system 100 and 200 of the invention.?
Noise-reduced image is further processed to obtain in the image quality enhancement steps of enhancing image, in addition to the processing time is slightly lower
In homomorphic filter and Wiener filter, on other indices, image processing system 100 and 200 of the invention is all had
Optimal representation.Therefore, it can not only more efficiently be removed in image using image processing system 100 or 200 of the invention
Speckle noise and noncoherent noise, additionally it is possible to farthest retain the details of image, and also have in image processing speed
Extraordinary performance.
Fig. 5, which is shown, is handled image using an exemplary embodiment of the present invention image processing system 100
The flow chart of method.
It can be by using the filter construction unit 110 of image processing system 100, based on figure in step 501 referring to Fig. 5
The different degrees of noise shadow that the image pixel intensities away from the pixel at each pixel different distance generate each pixel as in
It rings, constructs noise filter for each pixel according to the line noise model of image.
Then in step 503, by 120 use of noise reduction processing unit of image processing system 100 by filter construction unit
110 carry out noise reduction filtering to each pixel respectively for the noise filter that each pixel constructs, to obtain noise-reduced image.
It is described in detail with reference to Fig. 1 and is filtered using filter construction unit 110 for each pixel building noise reduction in front
The processing of wave device and the processing that noise reduction is carried out using noise reduction processing unit 120, therefore for simplicity, will no longer carry out weight herein
Multiple description.
Fig. 6 is shown using image processing system 200 in accordance with an alternative illustrative embodiment of the present invention to image
The flow chart of the method for reason.
It as shown in Figure 6, can be by using the filter construction unit 210 of image processing system 200, base in step 601
Image pixel intensities away from the pixel at each pixel different distance in image make an uproar to each the different degrees of of pixel generation
Sound shadow is rung, and constructs noise filter for each pixel according to the line noise model of image.
Then in step 603, by 220 use of noise reduction processing unit of image processing system 200 by filter construction unit
110 carry out noise reduction filtering to each pixel respectively for the noise filter that each pixel constructs, to obtain noise-reduced image.
Above step 601 to 603 and the step 501 in Fig. 5 are to 503 identical, and for the building noise reduction filter of each pixel
Wave device and the processing for executing noise reduction for each pixel have been described in detail above, therefore for simplicity, herein will not
Repeated description is carried out again.In step 605 to 609, the picture quality enhancement unit 260 of image processing system 200 can be used, pass through
Enhance the high fdrequency component in the noise-reduced image obtained in step 603 to obtain the enhancing image of picture quality enhancing.
In particular, in step 605, by the low frequency component extraction unit 230 of picture quality enhancement unit 260, in frequency
It is used in domain and the noise filter of each pixel building is directed to each pixel in noise-reduced image by filter construction unit 210
It is further filtered, extracts the low frequency component of the noise-reduced image.
The noise reduction figure is based on by the high fdrequency component extraction unit 240 of picture quality enhancement unit 260 in step 607
The low frequency component of picture and the noise-reduced image, extracts the high fdrequency component of the noise-reduced image.
Finally, in step 609, by the image processing unit 250 of picture quality enhancement unit 260 come the noise reduction to extraction
The high fdrequency component and low frequency component of image carry out linear combination and increase the weight of high fdrequency component, to obtain picture quality enhancing
Enhance image (for example, image of details and contrast enhancing).
It is described in detail above with reference to Fig. 2 using low frequency component included in picture quality enhancement unit 260
Extraction unit 230, high fdrequency component extraction unit 240 and image processing unit 250 enhance the method for processing to carry out picture quality,
Therefore for simplicity, will no longer carry out repeated description herein.
Although being particularly shown and describing the present invention, those skilled in the art referring to its exemplary embodiment
It should be understood that in the case where not departing from the spirit and scope of the present invention defined by claim form can be carried out to it
With the various changes in details.
Claims (9)
1. a kind of for improving the image processing method of picture quality, which comprises
(a) difference that each pixel is generated based on the image pixel intensities away from the pixel at each pixel different distance in image
The influence of noise of degree constructs noise filter for each pixel according to the line noise model of image；
(b) noise reduction filtering is carried out to each pixel respectively by using the noise filter for the building of each pixel, to obtain
Noise-reduced image,
Wherein, in step (a), for the noise filter f of the pixel i building in imageiMeet following equation (1):
Wherein, giIndicate the actual pixels intensity of the pixel i directly acquired from described image,WithRespectively indicate by calculate with
The different degrees of influence of noise that pixel for pixel i in preset range centered on pixel i is generated, acquisition are directed to pixel i
The first noise weighting coefficient and the second noise weighting coefficient,
Wherein, equation (2) will be used calculatedIt is determined as the first noise weighting coefficient:
Equation (3) will be used calculatedIt is determined as the second noise weighting coefficient:
Wherein, ωiAnd ωjRespectively indicate the set of pixels of all pixels in the region of the predefined size centered on pixel i and j
It closes, | ωi| and | ωj| respectively indicate pixel set ωiAnd ωjIn pixel quantity, gkIndicate the picture directly acquired from image
The actual pixels intensity of plain k,It indicates in pixel set ωjIn all pixels actual pixels intensity mean value,Table
Show in pixel set ωjIn all pixels actual pixels intensity variance, ε is a in order to preventjIt is excessive and predetermined
Regularization parameter, βjIt is to indicate ajAnd bjIt is rightWithNoise contribution weighting coefficient and can be determined in advance.
2. the image pixel intensities the method for claim 1, wherein away from the closer pixel of each pixel are to described each
The influence of noise that pixel generates is bigger, and the image pixel intensities away from the remoter pixel of each pixel generate each pixel
Influence of noise is smaller.
3. the method as described in any one in claim 1 to 2, further includes:
(c) enhancing image is obtained by the high fdrequency component in enhancing noise-reduced image.
4. method as claimed in claim 3, wherein step (c) includes:
(c1) in a frequency domain, by using the noise filter for the building of each pixel respectively to every in noise-reduced image
A pixel is further filtered, and the low frequency component of the noise-reduced image is extracted；
(c2) low frequency component of the noise-reduced image based on the noise-reduced image and extraction, extracts the height of the noise-reduced image
Frequency component；
(c3) high fdrequency component to the noise-reduced image of extraction and low frequency component carry out linear combination and increase the weight of high fdrequency component,
Enhance image to obtain.
5. method as claimed in claim 4, wherein in step (c2), extract noise-reduced image by following equation (4)
High fdrequency component phigh:
phigh=p-plow (4)
Wherein, p indicates the noise-reduced image in the frequency domain obtained in step (b), plowIndicate the drop extracted in step (c1)
It makes an uproar the low frequency component of image,
Wherein, in step (c3), according to following equation (5) to the low frequency component extracted in step (c1) and in step (c2)
The high fdrequency component of middle extraction carries out linear combination, enhances image p to obtainenhanced:
penhanced=λ phigh+plow (5)
Wherein, λ is predetermined weighted value and is greater than 1.
6. method as claimed in claim 5, wherein predefine λ by testing.
7. the method for claim 1, wherein described image is optical coherence tomography (OCT) image.
8. it is a kind of for improving the image processing system of picture quality, the system comprises:
Filter construction unit, for based on the image pixel intensities away from the pixel at each pixel different distance in image to described every
The different degrees of influence of noise that a pixel generates constructs noise reduction filtering for each pixel according to the line noise model of image
Device；
Noise reduction processing unit, for being distinguished by using the noise filter by filter construction unit for the building of each pixel
Noise reduction filtering is carried out to each pixel, to obtain noise-reduced image,
Wherein, by filter construction unit for the noise filter f of the pixel i building in imageiMeet following equation (1):
Wherein, giIndicate the actual pixels intensity of the pixel i directly acquired from described image,WithRespectively indicate by calculate with
The different degrees of influence of noise that pixel for pixel i in preset range centered on pixel i is generated, acquisition are directed to pixel i
The first noise weighting coefficient and the second noise weighting coefficient,
Wherein, equation (2) will be used calculatedIt is determined as the first noise weighting coefficient:
Equation (3) will be used calculatedIt is determined as the second noise weighting coefficient:
Wherein, ωiAnd ωjRespectively indicate the set of pixels of all pixels in the region of the predefined size centered on pixel i and j
It closes, | ωi| and | ωj| respectively indicate pixel set ωiAnd ωjIn pixel quantity, gkIndicate the picture directly acquired from image
The actual pixels intensity of plain k,It indicates in pixel set ωjIn all pixels actual pixels intensity mean value,Table
Show in pixel set ωjIn all pixels actual pixels intensity variance, ε is a in order to preventjIt is excessive and predetermined
Regularization parameter, βjIt is to indicate ajAnd bjIt is rightWithNoise contribution weighting coefficient and can be determined in advance.
9. system as claimed in claim 8, wherein the image pixel intensities away from the closer pixel of each pixel are to described each
The influence of noise that pixel generates is bigger, and the image pixel intensities away from the remoter pixel of each pixel generate each pixel
Influence of noise is smaller.
Priority Applications (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210280809.5A CN103578081B (en) | 2012-08-08 | 2012-08-08 | For improving the image processing method and system of picture quality |
| KR1020130092599A KR102106073B1 (en) | 2012-08-08 | 2013-08-05 | Method and apparatus for processing images |
| US13/961,973 US9639915B2 (en) | 2012-08-08 | 2013-08-08 | Image processing method and apparatus |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210280809.5A CN103578081B (en) | 2012-08-08 | 2012-08-08 | For improving the image processing method and system of picture quality |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103578081A CN103578081A (en) | 2014-02-12 |
| CN103578081B true CN103578081B (en) | 2018-12-28 |
Family
ID=50049807
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210280809.5A Active CN103578081B (en) | 2012-08-08 | 2012-08-08 | For improving the image processing method and system of picture quality |
Country Status (2)
| Country | Link |
|---|---|
| KR (1) | KR102106073B1 (en) |
| CN (1) | CN103578081B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6512738B2 (en) | 2014-02-28 | 2019-05-15 | キヤノン株式会社 | Image processing apparatus, image processing method and program |
| CN106910166B (en) * | 2016-08-31 | 2020-06-26 | 湖南拓视觉信息技术有限公司 | Image processing method and device |
| KR102062372B1 (en) | 2017-12-13 | 2020-01-03 | 한국생산기술연구원 | Apparatus for optical coherence tomography and method for image processing thereof for improving image quality based on region segmentation |
| KR102005822B1 (en) * | 2018-01-26 | 2019-07-31 | 성신여자대학교 연구 산학협력단 | Apparatus for detecting colloid cyst in brain by processing image, method thereof and computer recordable medium storing the method |
| CN109712078A (en) * | 2018-07-23 | 2019-05-03 | 永康市巴九灵科技有限公司 | Cabinet security protection control platform |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101396279A (en) * | 2007-09-29 | 2009-04-01 | 深圳市蓝韵实业有限公司 | Method of removing real-time ultrasound pattern speckle noise |
| US20090219417A1 (en) * | 2006-11-10 | 2009-09-03 | Takao Tsuruoka | Image capturing system and computer readable recording medium for recording image processing program |
| US20100194933A1 (en) * | 2009-02-02 | 2010-08-05 | Olympus Corporation | Image processing apparatus and image processing method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100344807B1 (en) * | 2000-01-12 | 2002-07-20 | 엘지전자주식회사 | Apparatus for correcting of image signal and method for the same |
| US8160381B2 (en) * | 2006-08-30 | 2012-04-17 | Micron Technology, Inc. | Method and apparatus for image noise reduction using noise models |
| JP5637730B2 (en) * | 2010-05-14 | 2014-12-10 | キヤノン株式会社 | Imaging apparatus and imaging method thereof |
-
2012
- 2012-08-08 CN CN201210280809.5A patent/CN103578081B/en active Active
-
2013
- 2013-08-05 KR KR1020130092599A patent/KR102106073B1/en active IP Right Grant
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090219417A1 (en) * | 2006-11-10 | 2009-09-03 | Takao Tsuruoka | Image capturing system and computer readable recording medium for recording image processing program |
| CN101396279A (en) * | 2007-09-29 | 2009-04-01 | 深圳市蓝韵实业有限公司 | Method of removing real-time ultrasound pattern speckle noise |
| US20100194933A1 (en) * | 2009-02-02 | 2010-08-05 | Olympus Corporation | Image processing apparatus and image processing method |
Non-Patent Citations (3)
| Title |
|---|
| OCT图像降噪混合滤波方法;李佳 等;《计算机工程与设计》;20111231;第32卷(第5期);全文 * |
| 一种医学超声图像降噪方法;王伟;《硅谷》;20091231;第37卷(第16期);第33页 * |
| 图像降噪的自适应高斯平滑滤波器;谢勤岚;《计算机工程与应用》;20091231;第45卷(第16期);第182-183页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| KR102106073B1 (en) | 2020-04-29 |
| CN103578081A (en) | 2014-02-12 |
| KR20140021487A (en) | 2014-02-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Aysal et al. | Rayleigh-maximum-likelihood filtering for speckle reduction of ultrasound images | |
| Chong et al. | Speckle reduction in optical coherence tomography images of human finger skin by wavelet modified BM3D filter | |
| Yu et al. | Ultrasound speckle reduction by a SUSAN-controlled anisotropic diffusion method | |
| CN103578081B (en) | For improving the image processing method and system of picture quality | |
| US9639915B2 (en) | Image processing method and apparatus | |
| Choi et al. | Speckle noise reduction in ultrasound images using SRAD and guided filter | |
| CN108042126A (en) | A kind of improved laser speckle contrasts blood flow imaging method | |
| KR102195047B1 (en) | Method and apparatus for enhancing quality of 3D image | |
| Gökdağ et al. | Image denoising using 2-D wavelet algorithm for Gaussian-corrupted confocal microscopy images | |
| Paul et al. | Speckle removal using diffusion potential for optical coherence tomography images | |
| KR100760251B1 (en) | Ultrasound image processing system and method | |
| Xu et al. | Wavelet domain compounding for speckle reduction in optical coherence tomography | |
| Liu et al. | Noise suppression in brain magnetic resonance imaging based on non-local means filter and fuzzy cluster | |
| Paul et al. | Filters used in x-ray chest images for initial stage tuberculosis detection | |
| CN103345730A (en) | Infrared image processing method based on lateral inhibition network | |
| Liang et al. | Nonlocal total variation based on symmetric Kullback-Leibler divergence for the ultrasound image despeckling | |
| Umamaheswari et al. | An enhanced approach for medical brain image enhancement | |
| Aarya et al. | Signal dependent rician noise denoising using nonlinear filter | |
| Hama et al. | Medical image enhancement based on an efficient approach for adaptive anisotropic diffusion | |
| Ahmed | Image enhancement and noise removal by using new spatial filters | |
| Gupta et al. | An efficient technique for speckle noise reduction in ultrasound images | |
| Rajalaxmi et al. | Entropy-based straight kernel filter for echocardiography image denoising | |
| Mustafa et al. | Wavelet domain bilateral filtering with subband mixing for magnetic resonance image enhancement | |
| Ramamoorthy et al. | An efficient method for speckle reduction in ultrasound liver images for e-health applications | |
| Rosa et al. | Speckle ultrasound image filtering: Performance analysis and comparison |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |