CN108805835A - Based on the SAR image bilateral filtering method for blocking statistical nature - Google Patents
Based on the SAR image bilateral filtering method for blocking statistical nature Download PDFInfo
- Publication number
- CN108805835A CN108805835A CN201810541504.2A CN201810541504A CN108805835A CN 108805835 A CN108805835 A CN 108805835A CN 201810541504 A CN201810541504 A CN 201810541504A CN 108805835 A CN108805835 A CN 108805835A
- Authority
- CN
- China
- Prior art keywords
- sliding window
- filtering
- bilateral filtering
- filter
- point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000001914 filtration Methods 0.000 title claims abstract description 121
- 230000002146 bilateral effect Effects 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims abstract description 49
- 230000000903 blocking effect Effects 0.000 title claims abstract description 23
- 230000001174 ascending effect Effects 0.000 claims abstract description 6
- 230000000694 effects Effects 0.000 claims description 11
- 238000013461 design Methods 0.000 claims description 8
- 230000002401 inhibitory effect Effects 0.000 claims description 8
- 101710198693 Invasin Proteins 0.000 claims description 4
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000003384 imaging method Methods 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 4
- 238000009499 grossing Methods 0.000 description 4
- 230000001629 suppression Effects 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 3
- 230000014759 maintenance of location Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000001427 coherent effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000004445 quantitative analysis Methods 0.000 description 2
- 230000001154 acute effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 210000001367 artery Anatomy 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000000205 computational method Methods 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000006735 deficit Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003018 immunoassay Methods 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 238000005191 phase separation Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 210000003462 vein Anatomy 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- 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
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
-
- 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
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a kind of based on the SAR image bilateral filtering method for blocking statistical nature, a certain size sliding window is arranged in this method, two-sided filter is used to carry out bilateral filtering to SAR image first, later by estimating the gray average and standard variance of all pixels in sliding window, and classify to the background type residing for current sliding window mouth central pixel point.Carry out ascending order arrangement by gray value size to the pixel in sliding window takes the sample of different depths to block in the sample in sliding window for different background types, and Speckle Filter is carried out using two-sided filter using the interlude sample after blocking.The filtering method can effectively keep the Edge texture of image while smooth speckle noise to greatest extent, have preferable engineering application value.
Description
Technical field
The present invention relates to SAR image speckle suppression technical fields more particularly to a kind of based on blocking statistical nature
SAR image bilateral filtering method.
Background technology
Synthetic aperture radar (Synthetic Aperture Radar, SAR) is a kind of high-resolution imaging radar,
Ability with round-the-clock and round-the-clock observation.By being irradiated to scenic spot coherent wave and to backscatter signal phase when SAR is imaged
Dry detection is to obtain the high-resolution of orientation.The whole signals returned from a surface units back scattering are in ground scatter
Backscatter signal is concerned with summation.Through confirmations such as Goodman, SAR image speckle noise can be modeled as a kind of multiplicative noise.
Earth's surface is usually made of the scattering surface of many random distributions, and this randomness is form of expression the reason of generating Image Speckle
Picture tone for pixel corresponding with simple target different units changes at random.The acute variation for showing as gradation of image, that is, exist
In the same a piece of uniform rough region of SAR image, some resolution elements are rendered as bright spot, and some is then rendered as dim spot, directly
The gray level resolution for affecting SAR images is connect, the detail section of SAR image is concealed.To give the interpretation of SAR image and determine
Quantization brings prodigious puzzlement, difficulty is brought to the information extraction of image, to hamper the application of image.Therefore, right
The research of speckle suppression is always one of the important topic of SAR image processing.
The elementary object of SAR image speckle noise filtering is to keep image side under the premise of inhibiting image speckle noise
The detailed information such as edge and texture.The inhibition of SAR image speckle noise can be realized by the multiple look processing technology in imaging process,
It can also be realized by the filtering of spatial domain after imaging.Since multiple look processing can increase the operation cost of imaging, and drop
The spatial resolution of low image.Therefore, inhibit speckle noise typically by the filtering of spatial domain is carried out to image.
In recent years, many traditional digital filtering techniques, such as mean value and medium filtering are used for SAR speckle noises and filter, but by
In the multiplying property feature and mean value of SAR speckle noises and the non adaptive of median filter, filter effect is unsatisfactory.
For this problem, many adaptive local statistics filter devices based on multiplying noise model are developed：Lee filtering,
Frost filtering, Kuan filtering, Gamma-MAP filtering and Sigma filtering etc..These method advantages are：Calculation amount is small, speed
Soon, spot cannot be taken into account to inhibit to keep with edge details.But they are by setting a certain size sliding window come to window
Interior data carry out phase separation immunoassay, their filter effect and the size of sliding window have substantial connection, when window is larger
When, filter transitions smooth leads to edge blurry so that some details of image impairment；It, can be fine when moving window is smaller
Protect edge information, but the smoothing capability of filter reduces at this time, and coherent speckle noise is less able.
Traditional SAR image Speckle Filter method has obscured the edge of SAR image while effectively inhibiting speckle noise,
And destroy the texture information of SAR image.Traditional two-sided filter not only cannot effectively inhibit strong speckle noise, also enhance
Strong speckle noise.
Invention content
The object of the invention is exactly to be provided a kind of based on the SAR figures for blocking statistical nature to make up the defect of prior art
As bilateral filtering method.
The present invention is achieved by the following technical solutions：
It is a kind of based on the SAR image bilateral filtering method for blocking statistical nature, it is characterised in that：Include the following steps：
Step (1)：The sliding window size of two-sided filter is set, spot is carried out to SAR image using bilateral filtering method
Noise filtering, effectively smooth clutter and weak speckle noise, while enhancing strong speckle noise.
Step (2)：Sliding window size L is set, allows the window to be slided in the SAR image after bilateral filtering, statistic window
The gray average μ and standard variance σ of all pixels in mouthful.
Step (3)：Ascending sort is carried out according to gray value to all pixels point in sliding window.It will be in sliding window
The relationship for the mean μ and standard variance σ that the gray value I of imago vegetarian refreshments is obtained with statistics compares and analyzes to judge sliding window
The type of clutter background residing for mouth central pixel point.The class of clutter background residing for sliding window central pixel point according to judgement
Type takes different degrees of sample to block, and carries out bilateral filtering using the sample after blocking, and realizes special based on statistics is blocked
The bilateral filtering of sign.
It is described based on the SAR image bilateral filtering method for blocking statistical nature, it is characterised in that：The step (1)
Specific method is：In two-sided filter, introduce a kind of filtering of weights, weights size simultaneously by geometric position phase recency and
Gray scale similarity determines that in this new filter, the two is combined with following formula at one piece：
In formula, h is output, and f is to input, the geometric position weights of pixel x and surrounding pixel ξ, s centered on c (ξ, x)
The gray scale similarity weights of pixel x and surrounding pixel ξ centered on (f (ξ), f (x)), k (x) are normalization coefficient, expression
Formula is：
Two-sided filter is a basic conception of noise remove, it needs to design two weight functions：Geometric position
Close function and gray level similarity function.The design of the two weight functions directly affects denoising effect.More commonly used is
Constant gaussian kernel function is linearly moved, the close function in geometric position and gray level similarity function are all gaussian kernel functions.Wherein geometry
The close gaussian kernel function in position is：
D (ξ, x) is the Euclidean distance of surrounding pixel point ξ and central pixel point x, σ in formuladFor the geometrical attenuation factor.
It controls the intensity of low-pass filtering, σdBigger, low-pass filtering intensity is stronger, and filter result is fuzzyyer.It usually chooses certain big
Wicket gray level similarity function is：
In formula, σrFor gray scale similarity invasin, to the filter effect got well, σdAnd σrIt chooses appropriate.
It is described based on the SAR image bilateral filtering method for blocking statistical nature, it is characterised in that：In the step (2)
Gray average μ and the computational methods of standard variance σ be：Assuming that all samples in sliding window are X={ x1,x2,…xN, lead to
Cross formula (5), (6) calculate the gray average μ and standard deviation sigma of entire sliding window all pixels；
It is described based on the SAR image bilateral filtering method for blocking statistical nature, it is characterised in that：The step (3)
In, the rule of bilateral filtering is：
(1)|I-μ|<2 σ, current point is common clutter point, step, roof edge point, at this time since noise intensity is not strong,
Therefore filtering more stresses edge holding.Therefore intermediate 7 × L × L/8 (rounding) a pixel is chosen to carry out bilateral filtering, is chosen
Bilateral filtering sample in point similar in removal and speckle noise gray value as possible, while can guarantee again and containing edge in sample
Point.Not only edge is maintained well, but also smooth to a certain extent clutter and noise.
(2)|I-μ|≥2σ&|I-μ|<3 σ, current point is strong spot or is continuous long roof edge point, is filtered at this time
Algorithm should pay attention to spot and inhibit to keep with edge simultaneously.Therefore it is double to carry out centre 2 × L × L/3 (rounding) a pixel can be chosen
Side filters, and both maintains roof edge, while greatly inhibiting strong speckle noise.
(3) | I- μ | >=3 σ, at this time current point be the strong spot in uniform clutter region or be short roof edge point, therefore
Filtering more stresses strong spot and inhibits.Intermediate 5 × L × L/6 (rounding) a pixel is chosen to carry out bilateral filtering, both maintains room
Ridged edge, while greatly inhibiting strong speckle noise.
It is an advantage of the invention that：
1, the present invention uses the speckle noise filtering that SAR image is realized based on the two-sided filter for blocking statistical nature,
Efficiently solve edge blurring problem present in traditional Speckle Filter method.
2, the present invention blocks the statistical sample in two-sided filter using adaptive method, and to the sample after blocking
This progress bilateral filtering, while keeping image border texture information, effectively smooth speckle noise.
3, the sample in the present invention block, bilateral filtering process it is accurately simple, computational efficiency is high, has higher engineering
Application value.
Description of the drawings
Fig. 1 is proposed by the present invention based on the SAR image bilateral filtering method flow chart for blocking statistical nature.
Fig. 2 is TerraSAR-X original images.
Fig. 3 is filter result of the Lee filtering to TerraSAR-X entire images.
Fig. 4 is filter result of the Frost filtering to TerraSAR-X entire images.
Fig. 5 is filter result of the Gamma-MAP filtering to TerraSAR-X entire images.
Fig. 6 is filter result of the Sigma filtering to TerraSAR-X entire images.
Fig. 7 is filter result of traditional bilateral filtering method to TerraSAR-X entire images.
Fig. 8 is filter result of the TS-BF filter methods proposed by the present invention to TerraSAR-X entire images.
Fig. 9 is each filtering algorithm to the filtered comparison diagrams of TerraSAR-X edge details region II.(a) it is that edge is thin
Region original image is saved, is (b) Lee filter results, is (c) Sigma filter results, is (d) Gamma-MAP filter results, (e)
For traditional bilateral filtering as a result, being (f) TS-BF filter results proposed by the present invention.
Figure 10 is each filtering algorithm to TerraSAR-X targets in ocean image filtering performance comparison figures.(a) it is
TerraSAR-X targets in ocean SAR images are (b) Lee filter results, are (c) Sigma filter results, are (d) Gamma-MAP
Filter result is (e) filter result (not iteration) of traditional bilateral filtering method, (h) is TS-BF filter methods proposed by the present invention
Filter result (not iteration), be (g) filter result after traditional bilateral filtering method iteration 5 times as a result, (h) being carried for the present invention
Filter result after the TS-BF filter methods iteration gone out 5 times.
Figure 11 is the design sketch for taking impulsive noise whole window samples progress bilateral filterings.(a) it is that signal adds arteries and veins
The pollution signal after noise is rushed, is (b) the joint weight function figure of 23 × 23 windows, is (c) to be carried out using whole window samples
Result after bilateral filtering.
Specific implementation mode
As shown in Figure 1, it is a kind of based on the SAR image bilateral filtering method for blocking statistical nature, include the following steps：
Step (1)：The sliding window size of two-sided filter is set, spot is carried out to SAR image using bilateral filtering method
Noise filtering, effectively smooth clutter and weak speckle noise, while enhancing strong speckle noise；
Step (2)：Sliding window size L is set, allows the window to be slided in the SAR image after bilateral filtering, statistic window
The gray average μ and standard variance σ of all pixels in mouthful；
Step (3)：Ascending sort is carried out according to gray value to all pixels point in sliding window, it will be in sliding window
The relationship for the mean μ and standard variance σ that the gray value I of imago vegetarian refreshments is obtained with statistics compares and analyzes to judge sliding window
The type of clutter background residing for mouth central pixel point；The class of clutter background residing for sliding window central pixel point according to judgement
Type takes different degrees of sample to block, and carries out bilateral filtering using the sample after blocking, and realizes special based on statistics is blocked
The bilateral filtering of sign.
Wherein in step (1), a kind of weights filtering is introduced in two-sided filter, weights size is simultaneously by geometry position
Phase recency and gray scale similarity are set to determine, in this new filter, the two is combined with following formula at one piece：
In formula, h is output, and f is to input, the geometric position weights of pixel x and surrounding pixel ξ, s centered on c (ξ, x)
The gray scale similarity weights of pixel x and surrounding pixel ξ centered on (f (ξ), f (x)), k (x) are normalization coefficient, expression
Formula is：
Two-sided filter is a basic conception of noise remove, it needs to design two weight functions：Geometric position
Close function and gray level similarity function.The design of the two weight functions directly affects denoising effect.More commonly used is
Constant gaussian kernel function is linearly moved, the close function in geometric position and gray level similarity function are all gaussian kernel functions；Wherein geometry
The close gaussian kernel function in position is：
D (ξ, x) is the Euclidean distance of surrounding pixel point ξ and central pixel point x, σ in formuladFor the geometrical attenuation factor.
The close gaussian kernel function in geometric position controls the intensity of low-pass filtering, σdBigger, low-pass filtering intensity is stronger, filter result
It is fuzzyyer.Usually choosing a certain size window gray level similarity function is：
In formula, σrFor gray scale similarity invasin, to the filter effect got well, σdAnd σrIt chooses appropriate.
If choosing whole window samples carries out bilateral filtering, it cannot play smooth effect well to non-Gaussian noise
Fruit, especially impulsive noise.Figure 11 is the design sketch for taking impulsive noise whole window samples progress bilateral filterings.(a) it is
Signal adds the pollution signal after impulsive noise, is (b) the joint weight function figure of 23 × 23 windows, is (c) using whole windows
Mouth sample carries out the result after bilateral filtering.It can be seen from the figure that not only cannot using whole window sample two-sided filters
Smooth impulsive noise well is also possible to Shangdi intensifier pulse noise to a certain degree on the contrary.If this is because in filter window
Including when impulsive noise, and the signal strength and space length of the impulsive noise and central pixel point were all very close at this time should
Impulsive noise occupies maximum ratio in filtered samples, causes impulsive noise that cannot filter well.
In step 2, sliding window size L is set, the window is allowed to be slided in the SAR image after bilateral filtering；Assuming that sliding
All samples in dynamic window are X={ x1,x2,…xN, the gray scale of entire sliding window all pixels is calculated by formula (5), (6)
Mean μ and standard deviation sigma；
In step 3, ascending sort is carried out according to gray scale size to all pixels point in sliding window first, is secondly existed
When carrying out bilateral filtering, it is contemplated that the spot restrainable algorithms of proposition are focused on keeping edge details while smooth speckle noise, because
This herein will analyze the type at edge；There are following two feelings for center pixel vertex type in current sliding window mouth
Condition：
(1) assume that the central pixel point of sliding window is marginal point, then can be divided into：
A) step edge, the standard variance σ counted in sliding window at this time can be very big.If central pixel point is step
Marginal point, then general satisfaction：|I-μ|<2σ.And when carrying out the selection of bilateral filtering sample, it is strong rejecting for Protect edge information
While speckle noise, it should be ensured that have the marginal point close to center pixel gray value in the interlude pixel samples of selection.
B) continuous long roof edge, the quasi- variances sigma counted in sliding window at this time can be inferior to the variance of step edge.
If central pixel point is roof edge point, then general satisfaction：σ≤|I-μ|<3σ.And when carrying out bilateral filtering, in order to very
Protect edge information well, when bilateral filtering sample is chosen, it should choose interlude pixel samples more as possible, it is to be ensured that choose
Sample in include roof edge point.
C) short roof edge, the quasi- variances sigma counted in sliding window at this time understands smaller, or even approaches without marginal point
Clutter region.If central pixel point is roof edge point, then general satisfaction：|I-μ|>3σ.And when carrying out bilateral filtering,
For Protect edge information well, when bilateral filtering sample is chosen, it should choose interlude pixel samples more as possible, pick
While except strong speckle noise, it is to be ensured that include roof edge point in the point of selection.
(2) assume that the central pixel point of current window is clutter point, then can be divided into：
A) Current central pixel is common intensity clutter point, then no matter it is in clutter region or fringe region, generally
All meet：|I-μ|<2 σ, therefore when bilateral filtering sample is chosen, selection interlude pixel more as possible rejects stronger spot
Spot noise sample carrys out the smooth clutter point.
B) Current central pixel is strong speckle noise point, it is likely to be in clutter region, it is also possible to be in marginal zone
Domain, therefore have：
|I-μ|<σ is in strong edge region or strong clutter region.
|I-μ|≥σ&|I-μ|<3 σ are in weak fringe region or common intensity clutter region
| I- μ | >=3 σ are in weak clutter region.
According to the residing background classification difference of current sliding window mouth center pixel select different length intermediate pixel section into
Row bilateral filtering.Filter rule is：
(1)|I-μ|<2 σ, current point is common clutter point, step, roof edge point, at this time since noise intensity is not strong,
Therefore filtering more stresses edge holding.Therefore intermediate 7 × L × L/8 (rounding) a pixel is chosen to carry out bilateral filtering, is chosen
Bilateral filtering sample in point similar in removal and speckle noise gray value as possible, while can guarantee again and containing edge in sample
Point.Not only edge is maintained well, but also smooth to a certain extent clutter and noise.
(2)|I-μ|≥2σ&|I-μ|<3 σ, current point is strong spot or is continuous long roof edge point, is filtered at this time
Algorithm should pay attention to spot and inhibit to keep with edge simultaneously.Therefore it is double to carry out centre 2 × L × L/3 (rounding) a pixel can be chosen
Side filters, and both maintains roof edge, while greatly inhibiting strong speckle noise.
(3) | I- μ | >=3 σ, at this time current point be the strong spot in uniform clutter region or be short roof edge point, therefore
Filtering more stresses strong spot and inhibits.Intermediate 5 × L × L/6 (rounding) a pixel is chosen to carry out bilateral filtering, both maintains room
Ridged edge, while greatly inhibiting strong speckle noise.
So far, it is basically completed based on the SAR image bilateral filtering method for blocking statistical nature.
The validity further illustrated the present invention below by way of TerraSAR-X imaging experiments.
TerraSAR-X Image Speckle Filter contrast experiments：
1. experimental setup：
Experimental data comes from TerraSAR-X satellite SAR images, as shown in Fig. 2, Fig. 6.It is proposed by the present invention in order to verify
There is good speckle suppression ability based on the SAR image bilateral filtering method (TS-BF) for blocking statistical nature, choose
Flat site I, III (the most left and most right image block that white box marks in Fig. 2) in Fig. 2 are tested, in order to verify this
The edge holding capacity of the TS-BF methods proposed is invented, Edge texture region II (the white box marks in Fig. 2 in Fig. 2 are chosen
The intermediate image block of note) it is tested.In experiment, filtered using Lee filtering (Lee), Kuan filtering (Kuan), Frost
(Frost), Gamma-MAP filter (Gamma-MAP), Sigma filtering (Sigma), bilateral filtering (BF) with it is proposed by the present invention
TS-BF methods carry out Speckle Filter performance comparison.
The sliding window size of each filtering method is set as 5 × 5.Traditional two-sided filter and TS-BF proposed by the present invention
Geometrical attenuation factor sigma in methoddWith gray scale similarity invasin σrIt is set as 5.
2. interpretation of result：
This experiment makes an uproar to the spot of SAR image flat site I, III to assess filtering algorithm using equivalent number (ENL)
Sound smoothing capability carries out quantitative analysis, is defined as：
Wherein For the SAR image after removal speckle noise.WithThe side of being respectively
Difference and mean value.Equivalent number is the parameter of most common evaluation Speckle reduction filter smoothing effect, and equivalent number is bigger,
Show that image-region is more smooth, the effect of Speckle noise removal is better.Especially to the clutter region without edge details and
Speech.
Using edge keep factor ESI to the edge holding capacity of SAR image fringe region II and whole picture SAR image into
Row quantitative analysis, is defined as：
M is former SAR image in formula, and M' is filtered SAR image.SAR image size is：m×n.ESI is bigger, shows
The edge retention of despeckle algorithm can be better, especially for multiple edge region.
Each filtering algorithm is to SAR image flat site I, III, filtering performance comparing result such as 1 institute of table of fringe region II
Show.TerraSAR-X original images are as shown in Fig. 2, Lee filtering, Frost filtering, Gamma-MAP filtering, Sigma filtering, tradition
Bilateral filtering and TS-BF proposed by the present invention are as shown in Figures 3 to 8 to the filter result of Fig. 2 whole picture SAR images.In order to compare
The edge details retention property of each algorithm individually shows the filter result of edge details region II in Fig. 2 in fig.9.
Figure 10 is the comparison diagram that each filtering method filters targets in ocean SAR image.According to table 1, Fig. 3 to Figure 10's as a result, can see
Go out：
(1) for the speckle suppression ability of smooth region, TS-BF filter methods proposed by the present invention have highest
Equivalent number, ENL are up to 125, and speckle noise smoothing capability is best；Fig. 9, Tu10Zhong, TS-BF filter methods iteration 5 times
Afterwards, speckle noise becomes very smooth, and the strong speckle noise in part is not in traditional bilateral filtering 5 post filtering results of algorithm iteration
Only do not inhibit, is enhanced instead.
(2) for the holding capacity of image edge detailss, the edge retention coefficient of TS-BF filter methods proposed by the present invention
ESI highests, and can be seen that traditional Lee, Kuan, Frost, Sigma, Gamma-MAP filter from the comparison in image 5, Fig. 6
The image border holding capacity of wave algorithm is poor, is obscured at object edge.Traditional bilateral filtering method is keeping the same of edge
When, strong speckle noise can be enhanced, as shown in Figure 10 (e), the strong speckle noise in part does not inhibit not only, is enhanced instead.
And TS-BF filter methods proposed by the present invention can effectively inhibit speckle noise while keeping SAR image edge details, such as scheme
Shown in 10 (f).After 5 iterative filterings, the strong speckle noise in part still cannot be filtered effectively in traditional bilateral filtering result
Remove, and TS-BF filter methods proposed by the present invention can whole smooth speckle noises, and keep object edge information.
(3) to sum up, TS-BF filtering methods proposed by the present invention, can be effective while smooth speckle noise to greatest extent
Keep the Edge texture of SAR image.
1 present invention of table is compared with the speckle noise filtering performance of other filtering methods
Claims (4)
1. a kind of based on the SAR image bilateral filtering method for blocking statistical nature, it is characterised in that：Include the following steps：
Step (1)：The sliding window size of two-sided filter is set, speckle noise is carried out to SAR image using bilateral filtering method
Filtering, effectively smooth clutter and weak speckle noise, while enhancing strong speckle noise；
Step (2)：Sliding window size L is set, allows the window to be slided in the SAR image after bilateral filtering, in statistical window
All pixels gray average μ and standard variance σ；
Step (3)：Ascending sort is carried out according to gray value to all pixels point in sliding window, according to imago in sliding window
The relationship for the mean μ and standard variance σ that the gray value I of vegetarian refreshments is obtained with statistics, classifies to sliding window, according to different
Type takes different degrees of sample to block, and carries out bilateral filtering using the sample after blocking, and realizes special based on statistics is blocked
The bilateral filtering of sign.
2. according to claim 1 based on the SAR image bilateral filtering method for blocking statistical nature, it is characterised in that：It is double
In the filter of side, a kind of weights filtering is introduced, weights size is determined by geometric position phase recency and gray scale similarity simultaneously,
In this new filter, the two is combined with following formula at one piece：
In formula, h is output, and f is to input, the geometric position weights of pixel x and surrounding pixel ξ centered on c (ξ, x), s (f (ξ),
F (x)) centered on pixel x and surrounding pixel ξ gray scale similarity weights, k (x) is normalization coefficient, and expression formula is：
Two-sided filter is a basic conception of noise remove, it needs to design two weight functions：Geometric position is close
The design of function and gray level similarity function, the two weight functions directly affects denoising effect, and more commonly used is linear
Constant gaussian kernel function is moved, the close function in geometric position and gray level similarity function are all gaussian kernel function, wherein geometric position
Close gaussian kernel function is：
D (ξ, x) is the Euclidean distance of surrounding pixel point ξ and central pixel point x, σ in formuladFor the geometrical attenuation factor, it is controlled
The intensity of low-pass filtering, σdBigger, low-pass filtering intensity is stronger, and filter result is fuzzyyer, usually chooses a certain size window
Gray level similarity function is：
In formula, σrFor gray scale similarity invasin, to the filter effect got well, σdAnd σrIt chooses appropriate.
3. according to described in claim 2 based on the SAR image bilateral filtering method for blocking statistical nature, it is characterised in that：
Sliding window size L is set, allows the window to be slided in the SAR image after bilateral filtering, it is assumed that all samples in sliding window
For X={ x1,x2,…xN, the gray average μ and standard deviation sigma of entire sliding window all pixels are calculated by formula (5), (6)；
4. according to claim 3 based on the SAR image bilateral filtering method for blocking statistical nature, it is characterised in that：It is right
All pixels point in sliding window carries out ascending sort according to gray scale size, secondly when carrying out bilateral filtering, by sliding window
The relationship for the gray average μ and standard deviation sigma that the gray value I of mouth central pixel point is obtained with statistics compares and analyzes to judge to work as
The classification of front slide window center pixel, and the intermediate pixel section of different length is selected to carry out bilateral filtering, filter rule is：
(1) | I- μ | 2 σ of <, current point is common clutter point, step, roof edge point, at this time since noise intensity is not strong,
Filtering more stresses edge holding, therefore chooses intermediate 7 × L × L/8 (rounding) a pixel to carry out bilateral filtering, selection it is bilateral
Removal and point similar in speckle noise gray value as possible in filtered samples, while can guarantee again and containing marginal point in sample, both very
Maintain edge well, while again smooth to a certain extent clutter and noise；
(2) | I- μ | >=2 σ & | I- μ | 3 σ of <, current point be strong spot or be continuous long roof edge point, at this time filtering algorithm
It should pay attention to spot simultaneously to inhibit to keep with edge, therefore centre 2 × L × L/3 (rounding) a pixel can be chosen to carry out bilateral filter
Wave both maintains roof edge, while greatly inhibiting strong speckle noise；
(3) | I- μ | >=3 σ, current point is the strong spot in uniform clutter region or is short roof edge point at this time, therefore is filtered
More stress strong spot to inhibit, chooses intermediate 5 × L × L/6 (rounding) a pixel to carry out bilateral filtering, both maintained ridge side
Edge, while greatly inhibiting strong speckle noise.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810541504.2A CN108805835B (en) | 2018-05-30 | 2018-05-30 | SAR image bilateral filtering method based on truncation statistical characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810541504.2A CN108805835B (en) | 2018-05-30 | 2018-05-30 | SAR image bilateral filtering method based on truncation statistical characteristics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108805835A true CN108805835A (en) | 2018-11-13 |
| CN108805835B CN108805835B (en) | 2021-08-10 |
Family
ID=64089419
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810541504.2A Active CN108805835B (en) | 2018-05-30 | 2018-05-30 | SAR image bilateral filtering method based on truncation statistical characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108805835B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111160260A (en) * | 2019-12-30 | 2020-05-15 | 湖北航天技术研究院总体设计所 | SAR image target detection method and system |
| CN111899195A (en) * | 2020-07-31 | 2020-11-06 | 深圳算子科技有限公司 | Rapid filtering method based on maximum value or minimum value of image |
| CN112529805A (en) * | 2020-12-14 | 2021-03-19 | 北京达佳互联信息技术有限公司 | Image restoration method and device, electronic equipment and storage medium |
| CN113724158A (en) * | 2021-08-13 | 2021-11-30 | 浙江大学 | Noise reduction method for dynamic contrast enhanced magnetic resonance imaging |
| CN114740476A (en) * | 2022-04-12 | 2022-07-12 | 合肥工业大学 | Video noise suppression method based on space-time trilateral filter |
| CN115661135A (en) * | 2022-12-09 | 2023-01-31 | 山东第一医科大学附属省立医院(山东省立医院) | Focus region segmentation method for cardio-cerebral angiography |
| CN115984148A (en) * | 2023-03-21 | 2023-04-18 | 吉林华瑞基因科技有限公司 | Denoising enhancement method for high-throughput gene sequencing data |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104134071A (en) * | 2014-06-20 | 2014-11-05 | 北京工业大学 | Deformable part model object detection method based on color description |
| CN105574829A (en) * | 2016-01-13 | 2016-05-11 | 合肥工业大学 | Adaptive bilateral filtering algorithm for polarized SAR image |
| US20170046822A1 (en) * | 2013-09-19 | 2017-02-16 | At&T Intellectual Property I, L.P. | Method and apparatus for image filtering |
-
2018
- 2018-05-30 CN CN201810541504.2A patent/CN108805835B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170046822A1 (en) * | 2013-09-19 | 2017-02-16 | At&T Intellectual Property I, L.P. | Method and apparatus for image filtering |
| CN104134071A (en) * | 2014-06-20 | 2014-11-05 | 北京工业大学 | Deformable part model object detection method based on color description |
| CN105574829A (en) * | 2016-01-13 | 2016-05-11 | 合肥工业大学 | Adaptive bilateral filtering algorithm for polarized SAR image |
Non-Patent Citations (2)
| Title |
|---|
| KHAOULA LASSOUED等: "Cooperative Localization with Reliable Confidence Domains Between Vehicles Sharing GNSS Pseudoranges Errors with No Base Station", 《IEEE INTELLIGENT TRANSPORTATION SYSTEMS MAGAZINE》 * |
| 孙增国 等: "基于区域分类、自适应滑动窗和结构检测的合成孔径雷达图像联合降斑算法", 《物理学报》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111160260A (en) * | 2019-12-30 | 2020-05-15 | 湖北航天技术研究院总体设计所 | SAR image target detection method and system |
| CN111160260B (en) * | 2019-12-30 | 2023-07-04 | 湖北航天技术研究院总体设计所 | SAR image target detection method and system |
| CN111899195A (en) * | 2020-07-31 | 2020-11-06 | 深圳算子科技有限公司 | Rapid filtering method based on maximum value or minimum value of image |
| CN112529805A (en) * | 2020-12-14 | 2021-03-19 | 北京达佳互联信息技术有限公司 | Image restoration method and device, electronic equipment and storage medium |
| CN112529805B (en) * | 2020-12-14 | 2024-04-30 | 北京达佳互联信息技术有限公司 | Image restoration method, device, electronic equipment and storage medium |
| CN113724158A (en) * | 2021-08-13 | 2021-11-30 | 浙江大学 | Noise reduction method for dynamic contrast enhanced magnetic resonance imaging |
| CN113724158B (en) * | 2021-08-13 | 2024-01-02 | 浙江大学 | Noise reduction method for dynamic contrast enhanced magnetic resonance imaging |
| CN114740476A (en) * | 2022-04-12 | 2022-07-12 | 合肥工业大学 | Video noise suppression method based on space-time trilateral filter |
| CN114740476B (en) * | 2022-04-12 | 2024-04-26 | 合肥工业大学 | Video noise suppression method based on space-time trilateral filter |
| CN115661135A (en) * | 2022-12-09 | 2023-01-31 | 山东第一医科大学附属省立医院(山东省立医院) | Focus region segmentation method for cardio-cerebral angiography |
| CN115984148A (en) * | 2023-03-21 | 2023-04-18 | 吉林华瑞基因科技有限公司 | Denoising enhancement method for high-throughput gene sequencing data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108805835B (en) | 2021-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108805835A (en) | Based on the SAR image bilateral filtering method for blocking statistical nature | |
| Bianchi et al. | Segmentation-based MAP despeckling of SAR images in the undecimated wavelet domain | |
| Jarabo-Amores et al. | Spatial-range mean-shift filtering and segmentation applied to SAR images | |
| Ahmed et al. | Speckle noise reduction in SAR images using adaptive morphological filter | |
| CN101482969B (en) | SAR image speckle filtering method based on identical particle computation | |
| Mridula et al. | Speckle noise reduction in images using Wiener filtering and adaptive Wavelet thresholding | |
| Gragnaniello et al. | SAR image despeckling by soft classification | |
| Das et al. | A comparative study of different noise filtering techniques in digital images | |
| Khosravi et al. | An introduction to ENVI tools for synthetic aperture radar (SAR) image despeckling and quantitative comparison of denoising filters | |
| Shedbalkar et al. | A comparative analysis of filters for noise reduction and smoothening of brain MRI images | |
| Kofidis et al. | Nonlinear adaptive filters for speckle suppression in ultrasonic images | |
| Yin et al. | A method for automatic target recognition using shadow contour of SAR image | |
| Mastriani et al. | Neural shrinkage for wavelet-based SAR despeckling | |
| Spreeuwers | Neural network edge detector | |
| Parikh et al. | Analysis of denoising techniques for speckle noise removal in synthetic aperture radar images | |
| Kapoor et al. | A brief review: Speckle reducing filtering for ultrasound images | |
| Li et al. | Ultrasound speckle reduction based on image segmentation and diffused region growing | |
| Nageswari et al. | A comparative study of despeckling filters for enhancement of medical images | |
| Ginolhac et al. | Morphological and statistical approaches to improve detection in the presence of reverberation | |
| Zhang et al. | Contrast study of side scan sonar image enhancement methods | |
| AKINTOYE et al. | COMPOSITE MEDIAN WIENER FILTER BASED TECHNIQUE FOR IMAGE ENHANCEMENT. | |
| Li et al. | Ultrasound image enhancement using dynamic filtering | |
| Chandra et al. | Noise Reduction Using Cellular Automata | |
| Naik et al. | Underwater Acoustic Image Processing for Detection of Marine Debris | |
| Kushwaha | An Effective Adaptive Fuzzy Filter for SAR Image Noise Reduction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |