CN101944230A - Multi-scale-based natural image non-local mean noise reduction method - Google Patents

Multi-scale-based natural image non-local mean noise reduction method Download PDF

Info

Publication number
CN101944230A
CN101944230A CN 201010268546 CN201010268546A CN101944230A CN 101944230 A CN101944230 A CN 101944230A CN 201010268546 CN201010268546 CN 201010268546 CN 201010268546 A CN201010268546 A CN 201010268546A CN 101944230 A CN101944230 A CN 101944230A
Authority
CN
China
Prior art keywords
image
high frequency
meant
low
frequency imaging
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
Application number
CN 201010268546
Other languages
Chinese (zh)
Other versions
CN101944230B (en
Inventor
钟桦
焦李成
王灿
王爽
侯彪
王桂婷
马文萍
尚荣华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xidian University
Original Assignee
Xidian University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xidian University filed Critical Xidian University
Priority to CN2010102685467A priority Critical patent/CN101944230B/en
Publication of CN101944230A publication Critical patent/CN101944230A/en
Application granted granted Critical
Publication of CN101944230B publication Critical patent/CN101944230B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention discloses a multi-scale-based natural image non-local mean noise reduction method, which mainly solves the problem that image textures and details are difficult to maintain in the conventional natural image noise reduction results. The method comprises the following steps of: (1) performing wavelet transform on input noise-containing natural images to decompose the noise-containing natural images into low-frequency images and high-frequency images; (2) correcting coefficients of the high-frequency images in different scales by using a BayesShrink method to obtain corrected high-frequency images; (3) reconstructing the low-frequency images and the corrected high-frequency images to obtain new low-frequency images; (4) correcting the new reconstructed low-frequency images by using a non-local mean method to obtain corrected low-frequency images; and (5) reconstructing the corrected low-frequency images and corrected high-frequency images to obtain noise-reduced images. Compared with other classic noise reduction methods, the method can better inhibit noise, also maintains the edges and texture details of the natural images, and can be used for noise reduction processing of the natural images.

Description

Based on multiple dimensioned natural image non-local mean denoising method
Technical field
The invention belongs to technical field of image processing, specifically a kind of denoising method can be used for the denoising to natural image.
Background technology
Universal day by day along with computing machine and digital image-forming equipment, Digital Image Processing more and more is subject to people's attention.Yet because the restriction of imaging device and image-forming condition, digital picture is being gathered, conversion, and the pollution that unavoidably is subjected to noise in the transportation, so image denoising is subjected to paying attention to widely as one of basic fundamental of image processing field.The noise of many reality can be similar to thinks white Gaussian noise, and the white Gaussian noise of removing in the image becomes important direction in the image denoising field.
Traditional denoising method roughly can be divided into two classes, and a class is based on the method in spatial domain, and a class is based on the method for transform domain.Compare classic methods in the denoising method of spatial domain and comprise gaussian filtering, medium filtering, bilateral filtering etc.Their common feature is exactly to utilize the continuity of local window interior pixel gray-scale value to come current pixel is carried out the gray scale adjustment.These methods have mostly been blured image when removing noise detailed information, edge of image for example, texture etc.
Because the contained information of natural image, particularly texture image has certain redundancy, people such as Buades have proposed a kind of denoising method of non-local mean.This method is that the certain window of size is got at the center with the current pixel, seeks the window that has analog structure with it in entire image, is that weights are adjusted the gray-scale value of current pixel with the similarity between the window.Because this method is in the good performance in denoising field, caused numerous scholars' extensive concern rapidly since proposing, but still there is following problem in it: 1: algorithm complexity is bigger; 2: it is not good enough that weights calculate accuracy; 3: still there be to a certain degree fuzzy in edge of image and details.
Based on the denoising method comparative maturity of multi-scale geometric analysis is exactly the various denoising methods of wavelet field, the key issue of small echo denoising is to the image wavelet coefficient Research of Statistical Model, this class basic idea is the prior probability model of statistical model as wavelet coefficient, utilize this prior imformation then, the wavelet coefficient to original image under Bayesian frame is estimated.Yet wavelet transformation still has following deficiency: 1: the sparse property of coefficient is relatively poor during higher-dimension; 2: the picture breakdown rear is limited to information, in order to overcome the deficiency of wavelet transformation, has occurred Ridgelet in recent years, Curvelet, Contourlet, Brushlet, the new tool that a series of images such as Bandelet decompose.But because the method for multi-scale geometric analysis is just done the adjustment of atrophy threshold value to the wavelet coefficient of high frequency imaging, low-frequency image is not processed,, and usually can produce Gibbs phenomenon so its final denoising effect is not very satisfactory.
Summary of the invention
The objective of the invention is to overcome the deficiency of above-mentioned prior art, proposed based on multiple dimensioned natural image non-local mean denoising method, low frequency and high-frequency information to image adopt diverse ways to handle, to realize taking into account to edge and smooth region in the natural image denoising, weaken Gibbs phenomenon greatly, improve image denoising effect.
For achieving the above object, the present invention includes following steps:
(1) the noisy natural image of input is done wavelet transformation, it is decomposed into low-frequency image and high frequency imaging;
(2) with the BayesShrink method coefficient of the high frequency imaging of different scale is revised, is obtained revised high frequency imaging:
V ij ′ = sign ( V ij ) ( | V ij | - T B ) | V ij | > T B 0 | V ij | ≤ T B
Wherein, V IjBe meant the high frequency imaging wavelet coefficient of j direction of i layer,
Figure BSA00000251464700022
Be meant the atrophy threshold value of j direction of i layer, Be meant the noise variance of j direction high frequency imaging of i layer, σ xBe meant the standard deviation of high frequency imaging signal, sign is meant sign function,
(3) the revised high frequency imaging that obtains of low-frequency image that step (1) is obtained and step (2) is reconstructed, and obtains new low-frequency image;
(4) with the non-local mean method reconstruct is obtained new low-frequency image and revises, obtain revised low-frequency image:
V ′ ( x ) = Σ y ∈ J V ( y ) w ( x , y ) ,
Wherein,
Figure BSA00000251464700026
Expression reconstruct obtains the similarity of new frequency image, L 1(x), L 2(y) represent that respectively reconstruct obtains in the new low-frequency image with x, y is that the size at center is the image block of n * n, and J is the set of search box pixel point, in experiment, the value of n is followed successively by 3 * 3, and 5 * 5,7 * 7, the value of J is followed successively by 7 * 7,11 * 11,21 * 21, h is level and smooth controlled variable, h=0.2 σ, V (y) are meant that reconstruct obtains the coefficient value in the new low-frequency image Search Area;
(5) revised low-frequency image that step (4) is obtained and step (2) obtain revised high frequency imaging and are reconstructed, and obtain the image after the denoising.
The present invention has the following advantages compared with prior art:
1. the present invention revises high frequency imaging with the BayesShrink method, and then farthest weakens Gibbs phenomenon owing to the non-local mean method low-frequency image is revised.
2. the present invention revises high frequency imaging with the BayesShrink method, and then can better suppress noise owing to the non-local mean method low-frequency image is revised, and keeps and recover the edge and the grain details of natural image simultaneously.
Description of drawings
Fig. 1 is a process flow diagram of the present invention;
Fig. 2 is the test pattern that the present invention uses;
Fig. 3 is the noisy image that the present invention uses;
Fig. 4 carries out figure as a result after the denoising with existing BayesShrink wavelet threshold method to Fig. 3;
Fig. 5 is the figure as a result that Fig. 3 is carried out denoising with existing two-sided filter method;
Fig. 6 is with the existing two-sided filter method is carried out denoising to Fig. 3 the figure as a result that differentiate more;
Fig. 7 is the figure as a result that Fig. 3 is carried out denoising with existing non local filtered method;
Fig. 8 is the figure as a result that Fig. 3 is carried out denoising with the inventive method.
Embodiment
With reference to accompanying drawing 1, the present invention is based on multiple dimensioned natural image non-local mean denoising method, comprise the steps:
Step 1 is done wavelet transformation to the noisy natural image of input, and it is decomposed into low-frequency image and high frequency imaging.
Digital picture is because the restriction of imaging device and image-forming condition inevitably will be subjected to the pollution of noise, and the noise of many reality can be thought Gauss's additive white noise, and noisy natural image model is:
v=u+n
Wherein, v is the gray-scale value of noisy image, u is the gray-scale value of clean image, n is Gauss's additive white noise, the present invention selects orthogonal wavelet that image is carried out 4 layers of decomposition in experiment, and the wavelet basis of use is db8, but is not limited to this decomposition method, for example also can use stationary wavelet transfer pair image to decompose, noisy natural image is decomposed into low-frequency image and high frequency imaging two parts.
Step 2 is revised the coefficient of the high frequency imaging of different scale with the BayesShrink method, obtains revised high frequency imaging.
People such as Chang are separate and obey under the hypothesis of generalized Gaussian distribution at wavelet coefficient, utilize bayes method to obtain a kind of estimation of actual signal, and then obtain BayesShrink atrophy threshold method, and its step is as follows:
2.1) the robust median method that utilizes Donoho to propose, estimate that the noise criteria of each high frequency imaging is poor:
σ ij = median ( abs ( V ij ( : ) ) ) 0.6745
Wherein, V IjBe meant the high frequency imaging wavelet coefficient of j direction of i layer, V Ij(:) is meant and takes out V IjIn whole wavelet coefficients of containing, abs is meant and takes absolute value that median is meant and gets intermediate value;
2.2) estimate the atrophy threshold value of each high frequency imaging:
T B ( σ x ) = σ ij 2 σ x
Wherein,
Figure BSA00000251464700043
It is meant the high-frequency signal variance of j direction of i layer,
Figure BSA00000251464700044
N is meant the number of the high frequency imaging wavelet coefficient of j direction of i layer;
2.3) use the BayesShrink method with the correction of high frequency imaging coefficient, obtain revised high frequency imaging:
V ij ′ = sign ( V ij ) ( | V ij | - T B ) | V ij | > T B 0 | V ij | ≤ T B .
Step 3, the revised high frequency imaging that low-frequency image that step (1) is obtained and step (2) obtain is reconstructed, and obtains new low-frequency image.
In the experiment, select orthogonal wavelet that image is reconstructed, the wavelet basis of use is db8, but is not limited to this reconstructing method, for example also can use stationary wavelet that image is reconstructed, as long as keep consistent with the picture breakdown instrument of use in the step 1.
Step 4 obtains new low-frequency image with the non-local mean method to reconstruct and revises, and obtains revised low-frequency image:
V ′ ( x ) = Σ y ∈ J V ( y ) w ( x , y ) ,
Wherein,
Figure BSA00000251464700047
Expression reconstruct obtains the similarity of new frequency image, L 1(x), L 2(y) represent that respectively reconstruct obtains in the new low-frequency image with x, y is that the size at center is the image block of n * n, and J is the set of search box pixel point, in experiment, the value of n is followed successively by 3 * 3, and 5 * 5,7 * 7, the value of J is followed successively by 7 * 7, and 11 * 11,21 * 21, h is level and smooth controlled variable, h=0.2 σ, σ are meant that the contained noise criteria of the noisy natural image of input is poor, and V (y) is meant that reconstruct obtains the coefficient value in the new low-frequency image Search Area.
Step 5, revised low-frequency image that step (4) is obtained and step (2) obtain revised high frequency imaging and are reconstructed, and obtain the image after the denoising.
In the experiment, select orthogonal wavelet that image is reconstructed, the wavelet basis of use is db8, but is not limited to this reconstructing method, for example also can use stationary wavelet that image is reconstructed, as long as keep consistent with the picture breakdown instrument of use in the step 1.
Effect of the present invention can further confirm by following experiment:
One. experiment condition and content
Experiment condition: test employed input picture shown in Fig. 2 and 3, wherein, Fig. 2 (a) is test pattern lena, Fig. 2 (b) is test pattern barbara, Fig. 2 (c) is test pattern peppers, Fig. 2 (d) is test pattern house, and Fig. 3 is that Fig. 2 (a) is added the noise criteria difference is 20 noisy lena image.
Experiment content: under above-mentioned experiment condition, use existing BayesShrink wavelet threshold filtering method respectively, the two-sided filter method is differentiated two-sided filter method and non local filtered method and the inventive method more Fig. 3 is experimentized.
Two. experimental result
With BayesShrink wavelet threshold method Fig. 3 is done the denoising result that obtains under the condition of five layers of decomposition as shown in Figure 4, as can be seen from Figure 4, the noise inhibiting ability of the method is limited, and edge and details exist fuzzy.
With the denoising result of two-sided filter method as shown in Figure 5, σ wherein d=1.8, σ r=2 * σ, the size of searching window is 11 * 11, as can be seen from Figure 5, the noise inhibiting ability of the method is better than BayesShrink wavelet threshold method, but it exists edge and the fuzzy problem of details equally;
With many resolutions two-sided filter method Fig. 3 is done two-layer wavelet decomposition, with the BayesShrink method high frequency imaging is handled, two-sided filter is handled low-frequency image, the denoising result that obtains as shown in Figure 6, σ wherein d=1.8, σ r=1.0 * σ, the size of searching window is 11 * 11, as can be seen from Figure 6, the method noise inhibiting ability is better than top two kinds of methods, but has edge and the not good enough problem of details conservation degree;
With the denoising result of non local filtered method as shown in Figure 7, wherein searching the window size is 21 * 21, similar window size is 7 * 7, smoothing parameter h=15 σ, σ is that the contained noise criteria of image is poor, as can be seen from Figure 7: the method noise inhibiting ability is relatively good, but can not well keep edge of image and texture information.
With the denoising result of the inventive method as shown in Figure 8, as can be seen from Figure 8: all methods of being mentioned above its denoising effect is better than, homogeneous region is also more level and smooth, and the brightness of image keeps effect better, and edge of image, details has also obtained good maintenance;
It is 10 that test pattern among Fig. 2 is added the noise criteria difference respectively, Gauss's additive white noise of 20,30 is with the evaluation index of PSNR as denoising effect, above-mentioned four kinds of existing denoising methods and method of the present invention are compared, and the denoising effect PSNR value of the whole bag of tricks is listed in the table 1.
The various denoising result contrasts of table 1
Figure BSA00000251464700061
Result in the table 1 is the result after average 10 times, as can be seen from Table 1, the denoising effect of the inventive method is than BayesShrink wavelet threshold, two-sided filter, two-sided filter and the non-local mean algorithms of differentiating all improve a lot on the PSNR value more, particularly also have denoising effect preferably for the relatively abundanter barbara image of texture information.
Above experimental result shows that the present invention is better than existing other denoising method on overall performance, can eliminate Gibbs phenomenon, keeps details such as the edge of natural image and texture better smooth noise the time.

Claims (2)

1. one kind based on multiple dimensioned natural image non-local mean denoising method, comprises the steps:
(1) the noisy natural image of input is done wavelet transformation, it is decomposed into low-frequency image and high frequency imaging;
(2) with the BayesShrink method coefficient of the high frequency imaging of different scale is revised, is obtained revised high frequency imaging:
V ij ′ = sign ( V ij ) ( | V ij | - T B ) | V ij | > T B 0 | V ij | ≤ T B
Wherein, V IjBe meant the high frequency imaging wavelet coefficient of j direction of i layer,
Figure FSA00000251464600012
Be meant the atrophy threshold value of j direction of i layer,
Figure FSA00000251464600013
Be meant the noise variance of j direction high frequency imaging of i layer, σ xBe meant the standard deviation of high frequency imaging signal, sign is meant sign function,
Figure FSA00000251464600014
(3) the revised high frequency imaging that obtains of low-frequency image that step (1) is obtained and step (2) is reconstructed, and obtains new low-frequency image;
(4) with the non-local mean method reconstruct is obtained new low-frequency image and revises, obtain revised low-frequency image:
V ′ ( x ) = Σ y ∈ J V ( y ) w ( x , y ) ,
Wherein,
Figure FSA00000251464600016
Expression reconstruct obtains the similarity of new frequency image, L 1(x), L 2(y) represent that respectively reconstruct obtains in the new low-frequency image with x, y is that the size at center is the image block of n * n, and J is the set of search box pixel point, in experiment, the value of n is followed successively by 3 * 3, and 5 * 5,7 * 7, the value of J is followed successively by 7 * 7,11 * 11,21 * 21, h is level and smooth controlled variable, h=0.2 σ, V (y) are meant that reconstruct obtains the coefficient value in the new low-frequency image Search Area;
(5) revised low-frequency image that step (4) is obtained and step (2) obtain revised high frequency imaging and are reconstructed, and obtain the image after the denoising.
2. the non local denoising method based on multiple dimensioned natural image according to claim 1 is characterized in that high frequency imaging is revised coefficient with the BayesShrink method described in the step (2), carries out as follows:
2a) the robust median method that utilizes Donoho to propose, estimate that the noise criteria of each high frequency imaging is poor:
σ ij = median ( abs ( V ij ( : ) ) ) 0.6745
Wherein, V IjBe meant the high frequency imaging wavelet coefficient of j direction of i layer, V Ij(:) is meant and takes out V IjIn whole wavelet coefficients of containing, abs is meant and takes absolute value that median is meant and gets intermediate value;
2b) estimate the atrophy threshold value of each high frequency imaging:
T B ( σ x ) = σ ij 2 σ x
Wherein, It is meant the not noisy high frequency imaging standard deviation of j direction of i layer,
Figure FSA00000251464600024
Be meant noisy high frequency imaging variance, n is meant the number of the high frequency imaging wavelet coefficient of j direction of i layer;
2c) use the BayesShrink method, obtain revised high frequency imaging the correction of high frequency imaging coefficient:
V ij ′ = sign ( V ij ) ( | V ij | - T B ) | V ij | > T B 0 | V ij | ≤ T B
Symbol is identical with above-mentioned explanation in the formula.
CN2010102685467A 2010-08-31 2010-08-31 Multi-scale-based natural image non-local mean noise reduction method Expired - Fee Related CN101944230B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2010102685467A CN101944230B (en) 2010-08-31 2010-08-31 Multi-scale-based natural image non-local mean noise reduction method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2010102685467A CN101944230B (en) 2010-08-31 2010-08-31 Multi-scale-based natural image non-local mean noise reduction method

Publications (2)

Publication Number Publication Date
CN101944230A true CN101944230A (en) 2011-01-12
CN101944230B CN101944230B (en) 2012-03-28

Family

ID=43436210

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2010102685467A Expired - Fee Related CN101944230B (en) 2010-08-31 2010-08-31 Multi-scale-based natural image non-local mean noise reduction method

Country Status (1)

Country Link
CN (1) CN101944230B (en)

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102156963A (en) * 2011-01-20 2011-08-17 中山大学 Denoising method for image with mixed noises
CN102298773A (en) * 2011-09-19 2011-12-28 西安电子科技大学 Shape-adaptive non-local mean denoising method
CN102622597A (en) * 2011-01-29 2012-08-01 中国第一汽车集团公司 Self-adaptive orthogonal median hybrid filtering method
CN103345726A (en) * 2013-06-14 2013-10-09 华为技术有限公司 Image de-noising processing method, device and terminal
CN103839235A (en) * 2014-02-24 2014-06-04 西安电子科技大学 Method for denoising global Bandelet transformation domain based on non-local directional correction
CN104504659A (en) * 2014-12-19 2015-04-08 成都品果科技有限公司 Quick ISO (international standardization organization) denoising method and system based on lifting wavelet transform
WO2015067186A1 (en) * 2013-11-08 2015-05-14 华为终端有限公司 Method and terminal used for image noise reduction
CN105738948A (en) * 2016-02-24 2016-07-06 重庆地质矿产研究院 Micro-seismic data denoising method based on wavelet transformation
CN103345726B (en) * 2013-06-14 2016-11-30 华为技术有限公司 Image denoising processing method, device and terminal
CN104156928B (en) * 2014-08-25 2017-04-26 深圳先进技术研究院 Ultrasonoscopy speckle noise filtering method based on Bayesian model
WO2017114473A1 (en) * 2015-12-31 2017-07-06 Shanghai United Imaging Healthcare Co., Ltd. Methods and systems for image processing
CN108629740A (en) * 2017-03-24 2018-10-09 展讯通信(上海)有限公司 A kind of processing method and processing device of image denoising
CN108898155A (en) * 2018-05-18 2018-11-27 浙江工业大学 A kind of SAR image target recognition method of Threshold Denoising combination convolutional neural networks

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050036707A1 (en) * 2000-11-09 2005-02-17 Canon Kabushiki Kaisha Image processing apparatus and its method, program and storage medium
CN1723691A (en) * 2003-09-29 2006-01-18 三星电子株式会社 Noise-reduction method and equipment
US20070036459A1 (en) * 2005-08-12 2007-02-15 Fuji Photo Film Co., Ltd. Digital signal processing apparatus
CN101661611A (en) * 2009-09-25 2010-03-03 西安电子科技大学 Realization method based on bayesian non-local mean filter
CN101719267A (en) * 2009-11-09 2010-06-02 中兴通讯股份有限公司 Method and system for denoising noise image

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050036707A1 (en) * 2000-11-09 2005-02-17 Canon Kabushiki Kaisha Image processing apparatus and its method, program and storage medium
CN1723691A (en) * 2003-09-29 2006-01-18 三星电子株式会社 Noise-reduction method and equipment
US20070036459A1 (en) * 2005-08-12 2007-02-15 Fuji Photo Film Co., Ltd. Digital signal processing apparatus
CN101661611A (en) * 2009-09-25 2010-03-03 西安电子科技大学 Realization method based on bayesian non-local mean filter
CN101719267A (en) * 2009-11-09 2010-06-02 中兴通讯股份有限公司 Method and system for denoising noise image

Cited By (29)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102156963A (en) * 2011-01-20 2011-08-17 中山大学 Denoising method for image with mixed noises
CN102622597B (en) * 2011-01-29 2016-05-04 中国第一汽车集团公司 Adaptive quadrature intermediate value mixed filtering method
CN102622597A (en) * 2011-01-29 2012-08-01 中国第一汽车集团公司 Self-adaptive orthogonal median hybrid filtering method
CN102298773A (en) * 2011-09-19 2011-12-28 西安电子科技大学 Shape-adaptive non-local mean denoising method
CN102298773B (en) * 2011-09-19 2012-12-26 西安电子科技大学 Shape-adaptive non-local mean denoising method
CN103345726A (en) * 2013-06-14 2013-10-09 华为技术有限公司 Image de-noising processing method, device and terminal
CN103345726B (en) * 2013-06-14 2016-11-30 华为技术有限公司 Image denoising processing method, device and terminal
CN104639800A (en) * 2013-11-08 2015-05-20 华为终端有限公司 Image denoising method and terminal
CN104639800B (en) * 2013-11-08 2017-11-24 华为终端(东莞)有限公司 A kind of method and terminal for image noise reduction
WO2015067186A1 (en) * 2013-11-08 2015-05-14 华为终端有限公司 Method and terminal used for image noise reduction
US9904986B2 (en) 2013-11-08 2018-02-27 Huawei Device (Dongguan) Co., Ltd. Image denoising method and terminal
CN103839235B (en) * 2014-02-24 2017-02-08 西安电子科技大学 Method for denoising global Bandelet transformation domain based on non-local directional correction
CN103839235A (en) * 2014-02-24 2014-06-04 西安电子科技大学 Method for denoising global Bandelet transformation domain based on non-local directional correction
CN104156928B (en) * 2014-08-25 2017-04-26 深圳先进技术研究院 Ultrasonoscopy speckle noise filtering method based on Bayesian model
CN104504659A (en) * 2014-12-19 2015-04-08 成都品果科技有限公司 Quick ISO (international standardization organization) denoising method and system based on lifting wavelet transform
CN104504659B (en) * 2014-12-19 2017-07-11 成都品果科技有限公司 A kind of quick ISO denoising methods and system based on lifting wavelet transform
WO2017114473A1 (en) * 2015-12-31 2017-07-06 Shanghai United Imaging Healthcare Co., Ltd. Methods and systems for image processing
GB2548767A (en) * 2015-12-31 2017-09-27 Shanghai United Imaging Healthcare Co Ltd Methods and systems for image processing
GB2548767B (en) * 2015-12-31 2018-06-13 Shanghai United Imaging Healthcare Co Ltd Methods and systems for image processing
CN108780571A (en) * 2015-12-31 2018-11-09 上海联影医疗科技有限公司 A kind of image processing method and system
US10290108B2 (en) 2015-12-31 2019-05-14 Shanghai United Imaging Healthcare Co., Ltd. Methods and systems for image processing
US11049254B2 (en) 2015-12-31 2021-06-29 Shanghai United Imaging Healthcare Co., Ltd. Methods and systems for image processing
CN108780571B (en) * 2015-12-31 2022-05-31 上海联影医疗科技股份有限公司 Image processing method and system
US11880978B2 (en) 2015-12-31 2024-01-23 Shanghai United Imaging Healthcare Co., Ltd. Methods and systems for image processing
CN105738948A (en) * 2016-02-24 2016-07-06 重庆地质矿产研究院 Micro-seismic data denoising method based on wavelet transformation
CN105738948B (en) * 2016-02-24 2018-03-23 重庆地质矿产研究院 Micro-seismic data denoising method based on wavelet transformation
CN108629740A (en) * 2017-03-24 2018-10-09 展讯通信(上海)有限公司 A kind of processing method and processing device of image denoising
CN108629740B (en) * 2017-03-24 2020-06-02 展讯通信(上海)有限公司 Image denoising processing method and device
CN108898155A (en) * 2018-05-18 2018-11-27 浙江工业大学 A kind of SAR image target recognition method of Threshold Denoising combination convolutional neural networks

Also Published As

Publication number Publication date
CN101944230B (en) 2012-03-28

Similar Documents

Publication Publication Date Title
CN101950414B (en) Non-local mean de-noising method for natural image
CN101944230B (en) Multi-scale-based natural image non-local mean noise reduction method
Fathi et al. Efficient image denoising method based on a new adaptive wavelet packet thresholding function
Hyvarinen et al. Image feature extraction by sparse coding and independent component analysis
Wang et al. A new wavelet-based image denoising using undecimated discrete wavelet transform and least squares support vector machine
CN103700072A (en) Image denoising method based on self-adaptive wavelet threshold and two-sided filter
CN101980284A (en) Two-scale sparse representation-based color image noise reduction method
CN103473740A (en) Sparse representation and low-rank double restraints-based nonlocal denoising method
CN108550121A (en) A kind of sediment sonar image processing method based on medium filtering and wavelet transformation
CN103093441A (en) Image denoising method based on non-local means and double variant model of transform domain
CN103020918A (en) Shape-adaptive neighborhood mean value based non-local mean value denoising method
Xizhi The application of wavelet transform in digital image processing
CN100417191C (en) Method of reducing noise for combined images
Dai et al. Pmpa: A patch-based multiscale products algorithm for image denoising
Wang et al. A wavelet-based image denoising using least squares support vector machine
CN102222327A (en) Image denoising method based on Treelet transformation and minimum mean-square error estimation
CN101957984B (en) Image de-noising method based on parametric estimation of non-local shrinkage factor
CN108428221A (en) A kind of neighborhood bivariate shrinkage function denoising method based on shearlet transformation
Guerrero-Colón et al. Image denoising using mixtures of Gaussian scale mixtures
Gleich et al. Despeckling of TerraSAR-X data using second-generation wavelets
CN102314675B (en) Wavelet high-frequency-based Bayesian denoising method
Dong Adaptive image denoising using wavelet thresholding
Sun et al. Color image denoising based on guided filter and adaptive wavelet threshold
kumar Dass et al. Improvising MSN and PSNR for finger-print image noised by GAUSSIAN and SALT & PEPPER
CN103854258A (en) Image denoising method based on Contourlet transformation self-adaptation direction threshold value

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20120328

Termination date: 20170831