CN107451981A - Picture noise level estimation method based on DCT and gradient covariance matrix - Google Patents

Picture noise level estimation method based on DCT and gradient covariance matrix Download PDF

Info

Publication number
CN107451981A
CN107451981A CN201710698252.XA CN201710698252A CN107451981A CN 107451981 A CN107451981 A CN 107451981A CN 201710698252 A CN201710698252 A CN 201710698252A CN 107451981 A CN107451981 A CN 107451981A
Authority
CN
China
Prior art keywords
mtd
mrow
msub
image
mtr
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.)
Pending
Application number
CN201710698252.XA
Other languages
Chinese (zh)
Inventor
寇婷婷
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.)
Sichuan Changhong Electric Co Ltd
Original Assignee
Sichuan Changhong Electric Co Ltd
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 Sichuan Changhong Electric Co Ltd filed Critical Sichuan Changhong Electric Co Ltd
Priority to CN201710698252.XA priority Critical patent/CN107451981A/en
Publication of CN107451981A publication Critical patent/CN107451981A/en
Pending legal-status Critical Current

Links

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
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/40—Analysis of texture
    • G06T7/41—Analysis of texture based on statistical description of texture
    • G06T7/42—Analysis of texture based on statistical description of texture using transform domain methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/40—Analysis of texture
    • G06T7/41—Analysis of texture based on statistical description of texture
    • G06T7/45—Analysis of texture based on statistical description of texture using co-occurrence matrix computation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a kind of picture noise level estimation method based on DCT and gradient covariance matrix, method and step is as follows, Step 1: obtain almost clean picture material and noise data first with the distribution character of frequency coefficient after dct transform, Step 2: then weigh the complexity of picture structure using the mark of image block gradient covariance matrix, Step 3: influence of the image detail information to estimated result accuracy is further removed by iteration, so as to obtain accurate estimated result.Compared with prior art, invention removes influence of the image detail of the high frequency to estimated result accuracy, the Stability and veracity of estimated result is greatly enhanced, particularly in the case of noise level is less.

Description

Picture noise level estimation method based on DCT and gradient covariance matrix
Technical field
The present invention relates to digital image processing field, more particularly to a kind of image based on DCT and gradient covariance matrix Noise level estimation method.
Background technology
With the rapid development of information technology, in daily life image have become people obtain information important channel it One, it is more and more directly perceived that the information content that image includes is compared with word.Image is during acquisition and transmission, inevitably It can be polluted by noise, noise can cause distorted signals, reduce the definition and sight of digital picture so that picture quality The requirement of people can not be met, need to remove noise from image to solve this problem.Generally, due to most of make an uproar Sound can be modeled and it is easily processed with Gaussian Profile, therefore additive Gaussian noise model is widely used in engineering practice In.Although the achievement in research on image denoising has had much at present, denoising effect is also more satisfactory, and most of images are gone Algorithm make an uproar all assuming that being carried out in the case of known to noise intensity, this is impossible in the application of reality, therefore I Need to obtain noise parameter from single image first, could continue to image carry out denoising.Except image denoising, very The all more or less parameter dependent on noise level of more image processing algorithms, including compression of images, segmentation, tampering detection, mesh Mark detection, identification etc..
The influence of noise on image is can not be ignored, and noise parameter estimation is the key of many image processing algorithms again Step, the accuracy of estimated result directly affect the performance of these algorithm subsequent treatments, and image noise estimation algorithm can at present It is divided into three classes, the noise Estimation Algorithm based on filtering, block-based noise Estimation Algorithm and the noise based on orthogonal transformation are estimated Calculating method.Noise Estimation Algorithm based on filtering, its thought are to allow noisy image to obtain filtering image by wave filter first, And think that filtered image is not by the clean image of noise pollution, therefore noisy image and the difference of filtering image are For noise, so as to estimate noise parameter.But meeting when being estimated based on the noise Estimation Algorithm of the filtering image more to texture Larger evaluated error be present, do not obtain relatively good solution method also so far.Block-based noise Estimation Algorithm, it Thought be to estimate noise parameter using some image blocks of relatively flat in image because the picture material ratio of flat site Relatively simple, isolated noise data is not influenceed by picture material substantially.This kind of algorithm is largely dependent upon image Type, fluctuation is larger, and the size and number of image block can also impact to estimated result, it is maximum the shortcomings that be for texture Image typically cannot get accurate estimated result.Noise Estimation Algorithm based on orthogonal transformation, conventional orthogonal transformation have discrete Cosine transform, wavelet transformation and principal component analysis etc., its thought is after carrying out certain orthogonal transformation to image, picture material and Noise can preferably separate in some parts, such as the orthogonal side that high-frequency information, the PCA after wavelet transformation are minimum after converting To etc., then using this partial data come estimating noise of input image parameter.Existing noise Estimation Algorithm belongs to the above three substantially Class, some algorithms have also obtained relatively good result, but the stability of algorithm and adaptivity do not solve also well.
The content of the invention
The purpose of the present invention, which is that, provides a kind of picture noise horizontal estimated side based on DCT and gradient covariance matrix Method, influence of the image detail of the high frequency to estimated result accuracy is eliminated, improve the standard of estimated result to a certain extent True property and stability, particularly in the case of noise level is less.
To achieve these goals, the technical solution adopted by the present invention is：It is a kind of based on DCT and gradient covariance matrix Picture noise level estimation method, method and step is as follows,
Step 1: obtain almost clean picture material and noise first with the distribution character of frequency coefficient after dct transform Data,
Step 2: the complexity of picture structure then is weighed using the mark of image block gradient covariance matrix,
Step 3: influence of the image detail information to estimated result accuracy is further removed by iteration, so as to obtain Accurate estimated result.
Preferably, in step 1, DCT processing method is,
(1) noise image is divided into the image block of the non-overlapping copies of M × M sizes.
(2) dct transform is done to each image block, retains upper left corner K × K low frequency component, remaining coefficient is set to zero.
(3) DCT inverse transformations are done to treated frequency coefficient, obtains approximate clean image and corresponding noise data.
Preferably, in step 2, dct transform is done to each image block, obtains coefficient of the correspondence image block in frequency domain Matrix, the matrix size are also M × M, and wherein first, upper left corner coefficient is DC coefficient, are the flat of pixel in the image block Average, in the frequency coefficient matrix, the upper left corner is low frequency component, contains the most information of image, and the lower right corner is high frequency Component, it is the detailed information of image, generally high fdrequency component is smaller or is zero, particularly to the weak texture area in image Domain, high frequency coefficient is almost nil, but noise data is concentrated in high fdrequency component, therefore we can be by the high frequency of noise image Component is set to zero, i.e., only retains image block upper left corner K × K coefficient, remaining coefficient is set to 0, according to test result, for smaller Image, M 8, K 3, for bigger image, M 16, K are that 5 can obtain more accurately estimated result.Then it is right The image block does idct transform, it is possible to obtains the almost clean image of the image block.Carry out the calculating of the dct transform of image block Formula is as follows,
Formula wherein F (u, v) is the DCT coefficient of correspondence position.
Preferably, in step 2, the mark of the gradient covariance matrix of each almost clean image block is calculated, and from small To big sequence, mark shows that more greatly image block texture information is more, as the weak texture region in standard screening image.
Preferably, in step 2, after obtaining the almost clean image of each image block, for weak texture region, The image that inverse transformation obtains is exactly clean image not affected by noise, therefore noisy image and the difference of clean image are exactly to scheme Noise as in, but for texture is than more rich image block, a part of image is lost in the image that inverse transformation obtains Texture information, the noise data so obtained is disturbed by image texture information, therefore we utilize image block gradient The mark of covariance matrix weighs whether image block is weak texture region, so as to which the noise data disturbed by texture information be arranged Remove.Because gradient reflects the change of pixel in image, for weak texture region, pixel is almost unchanged or change is smaller, So as to which Grad is smaller, therefore the mark (all characteristic value sums) of gradient covariance matrix can reflect that the texture of image block is strong Degree.For texture image block, although lost a part of detailed information, the change between pixel is still bigger, so The gradient covariance matrix of image block is calculated using following gradient operator horizontally and vertically, formula is as follows,
Wherein yiIt is pixel value M in the image block2× 1 one-dimensional vector expression, DhAnd DvIt is horizontal and vertical ladder respectively Operator is spent, T represents transposition operation.DhAnd DvIt is as follows：
Singular value decomposition is carried out to gradient covariance matrix, its all characteristic value is obtained, then calculates its mark to weigh The texture strength of image block is measured, mark shows that more greatly the image block texture information is more.
Preferably, if the image filtered out is weak texture region, the image that DCT inverse transformations obtain is exactly not by noise shadow The difference of loud clean image, noisy image and clean image is exactly the noise in image.
Preferably, if what is filtered out is the image block of texture-rich, then by iterating to calculate final noise variance, with This estimates final noise parameter estimated result.
Preferably, it is to select 20% minimum image block of mark to be used as just to iterate to calculate final noise variance mode The weak texture region to begin, the initial of estimated result is used as by the use of the noise variance that noise data corresponding to these regions calculates Value, by iterating to calculate final noise variance, each time in the upper weak texture region once chosen with 20% at Mark is threshold value, further screens weak texture region and carries out noise parameter estimation, the difference of front and rear estimated result twice is less than 0.01 Then stop iteration, the estimate is final noise parameter estimated result.
Compared with prior art, the advantage of the invention is that：Estimation is tied invention removes image detail of the high frequency The influence of fruit accuracy, the Stability and veracity of estimated result is greatly enhanced, it is particularly smaller in noise level In the case of.
Brief description of the drawings
Fig. 1 is this flow chart of the method for the present invention.
Embodiment
The invention will be further described below.
Embodiment 1：Referring to Fig. 1, a kind of picture noise level estimation method based on DCT and gradient covariance matrix,
Energy is concentrated mainly on low frequency component after make use of image DCT transform, and noise but concentrates on the characteristic of high fdrequency component； The texture strength of image block is weighed using the mark of gradient covariance matrix, the weak texture region in image is selected with this, most Estimate noise parameter level using the noise data of weak texture region afterwards.Comprise the following steps：
Step 1: noise image is divided into the image block of the non-overlapping copies of 8 × 8 sizes.
Step 2: doing dct transform to each image block, because image information is mainly distributed on low frequency, noise is but concentrated In high frequency, therefore retain the low frequency component in the upper left corner 3 × 3, remaining coefficient is set to zero.Then DCT inversions are done to DCT coefficient matrix Get approximate clean image in return.
DCT inverse transformations are done to treated frequency coefficient, obtain approximate clean image and corresponding noise data.For Weak texture region, the image that inverse transformation obtains are exactly clean image not affected by noise, therefore noisy image and clean image Difference be exactly noise in image, but for the image block of texture-rich, one is lost in the image that inverse transformation obtains The texture information of parts of images, the noise data so obtained are disturbed by image texture information.
Step 3: by the form that the graphical representation after idct transform is one-dimensional matrix, the gradient of the image block is then asked to assist Variance matrix, calculates the mark of the gradient covariance matrix of each almost clean image block, and sorts from small to large, the bigger table of mark The bright image block texture information is more.Judge whether picture material is flat using the mark of matrix, as standard screening image In weak texture region, filter out the influence that texture information is estimated noise.
Step 4: the mark that image block corresponds to gradient covariance matrix sorts from small to large, 20% image block is made before selection For initial weak texture region.For weak texture region, the image that inverse transformation obtains is exactly clean image not affected by noise, Therefore noisy image and the difference of clean image are exactly the noise in image, and variance is calculated as estimation by the use of these noise datas As a result initial value.Each time in the weak texture region that previous step chooses, using the mark of 20% opening position matrix as threshold value Weak texture region is further screened, parameter Estimation is carried out to the noise data that newly chooses, estimated result twice before and after calculating Difference, difference is less than 0.01 end iteration, and using the result as final estimated result.
Average value using 15 width experimental image estimated results is as shown in table 1, σnExpression is actually added into the noise mark of image It is accurate poor,Represent noise estimated result (average values of 15 width experimental image estimated results).Method two is D.L.Donoho et al. The wavelet field classics image noise estimation algorithm of proposition, method three utilize the average exhausted of smooth region for what S.Sari et al. was proposed To the picture noise method of estimation of deviation.
The picture noise estimated result of table 1
Conclusion：Noise criteria difference using the experimental data that the inventive method obtains and 15 width experimental images is closest, Accuracy is optimal.
In summary the present invention utilize dct transform after frequency coefficient distribution character obtain almost clean picture material and Noise data, the complexity of picture structure then is weighed using the mark of image block gradient covariance matrix, come as standard The noise data not influenceed by image information is screened, image detail information is further removed to estimated result accuracy by iteration Influence, so as to obtain accurate estimated result, particularly in the case of noise level is less.Needing picture noise horizontal The place of estimation can this patent method, this patent is applied to all calculations using image additive noise level as parameter of needs Method, such as denoising, compression etc..Above to provided by the present invention horizontal based on the picture noise of DCT and gradient covariance matrix Method of estimation has carried out exhaustive presentation, and specific case used herein is explained the principle and embodiment of the present invention State, the explanation of above example is only intended to help the method and its core concept for understanding the present invention；Meanwhile for this area Those skilled in the art, according to the thought of the present invention, there will be changes in specific embodiments and applications, to this hair Bright change and improvement will be possible, without the spirit and scope beyond accessory claim defined, in summary, sheet Description should not be construed as limiting the invention.

Claims (8)

  1. A kind of 1. picture noise level estimation method based on DCT and gradient covariance matrix, it is characterised in that：Method and step is such as Under,
    Step 1: almost clean picture material and noise data are obtained first with the distribution character of frequency coefficient after dct transform,
    Step 2: the complexity of picture structure then is weighed using the mark of image block gradient covariance matrix,
    Step 3: influence of the image detail information to estimated result accuracy is further removed by iteration, it is accurate so as to obtain Estimated result.
  2. 2. the picture noise level estimation method based on DCT and gradient covariance matrix, its feature exist according to claim 1 In：In step 1, DCT processing method is,
    (1) noise image is divided into the image block of the non-overlapping copies of M × M sizes.
    (2) dct transform is done to each image block, retains upper left corner K × K low frequency component, remaining coefficient is set to zero.
    (3) DCT inverse transformations are done to treated frequency coefficient, obtains approximate clean image and corresponding noise data.
  3. 3. the picture noise level estimation method based on DCT and gradient covariance matrix, its feature exist according to claim 1 In：In step 2,
    It is as follows to carry out the calculation formula of the dct transform of image block,
    <mrow> <mi>F</mi> <mrow> <mo>(</mo> <mi>u</mi> <mo>,</mo> <mi>v</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>c</mi> <mrow> <mo>(</mo> <mi>u</mi> <mo>)</mo> </mrow> <mi>c</mi> <mrow> <mo>(</mo> <mi>v</mi> <mo>)</mo> </mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>x</mi> <mo>=</mo> <mn>0</mn> </mrow> <mrow> <mi>M</mi> <mo>-</mo> <mn>1</mn> </mrow> </munderover> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>y</mi> <mo>=</mo> <mn>0</mn> </mrow> <mrow> <mi>M</mi> <mo>-</mo> <mn>1</mn> </mrow> </munderover> <mo>&amp;lsqb;</mo> <mi>I</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mi>cos</mi> <mo>(</mo> <mfrac> <mrow> <mo>(</mo> <mi>x</mi> <mo>+</mo> <mn>0.5</mn> <mo>)</mo> <mi>&amp;pi;</mi> <mi>u</mi> </mrow> <mi>M</mi> </mfrac> <mo>)</mo> <mi>cos</mi> <mo>(</mo> <mfrac> <mrow> <mo>(</mo> <mi>y</mi> <mo>+</mo> <mn>0.5</mn> <mo>)</mo> <mi>&amp;pi;</mi> <mi>v</mi> </mrow> <mi>M</mi> </mfrac> <mo>)</mo> <mo>&amp;rsqb;</mo> </mrow>
    <mrow> <mi>c</mi> <mrow> <mo>(</mo> <mi>u</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>c</mi> <mrow> <mo>(</mo> <mi>v</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <msqrt> <mfrac> <mn>1</mn> <mi>M</mi> </mfrac> </msqrt> </mtd> <mtd> <mrow> <mi>u</mi> <mo>,</mo> <mi>v</mi> <mo>=</mo> <mn>0</mn> </mrow> </mtd> </mtr> <mtr> <mtd> <msqrt> <mfrac> <mn>2</mn> <mi>M</mi> </mfrac> </msqrt> </mtd> <mtd> <mrow> <mi>u</mi> <mo>,</mo> <mi>v</mi> <mo>&amp;NotEqual;</mo> <mn>0</mn> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow>
    Formula wherein F (u, v) is the DCT coefficient of correspondence position.
  4. 4. the picture noise level estimation method based on DCT and gradient covariance matrix, its feature exist according to claim 1 In：In step 2, the mark of the gradient covariance matrix of each almost clean image block is calculated, and is sorted from small to large, mark is got over Show that image block texture information is more greatly, as the weak texture region in standard screening image.
  5. 5. the picture noise level estimation method based on DCT and gradient covariance matrix, its feature exist according to claim 1 In：In step 2,
    The gradient covariance matrix of image block is calculated using following gradient operator horizontally and vertically, formula is such as Under,
    <mrow> <msub> <mi>C</mi> <msub> <mi>y</mi> <mi>i</mi> </msub> </msub> <mo>=</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mrow> <msubsup> <mi>y</mi> <mi>i</mi> <mi>T</mi> </msubsup> <msubsup> <mi>D</mi> <mi>h</mi> <mi>T</mi> </msubsup> <msub> <mi>D</mi> <mi>h</mi> </msub> <msub> <mi>y</mi> <mi>i</mi> </msub> </mrow> </mtd> <mtd> <mrow> <msubsup> <mi>y</mi> <mi>i</mi> <mi>T</mi> </msubsup> <msubsup> <mi>D</mi> <mi>h</mi> <mi>T</mi> </msubsup> <msub> <mi>D</mi> <mi>v</mi> </msub> <msub> <mi>y</mi> <mi>i</mi> </msub> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msubsup> <mi>y</mi> <mi>i</mi> <mi>T</mi> </msubsup> <msubsup> <mi>D</mi> <mi>v</mi> <mi>T</mi> </msubsup> <msub> <mi>D</mi> <mi>h</mi> </msub> <msub> <mi>y</mi> <mi>i</mi> </msub> </mrow> </mtd> <mtd> <mrow> <msubsup> <mi>y</mi> <mi>i</mi> <mi>T</mi> </msubsup> <msubsup> <mi>D</mi> <mi>v</mi> <mi>T</mi> </msubsup> <msub> <mi>D</mi> <mi>v</mi> </msub> <msub> <mi>y</mi> <mi>i</mi> </msub> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow>
    Wherein yiIt is pixel value M in the image block2× 1 one-dimensional vector expression, DhAnd DvIt is that horizontal and vertical gradient is calculated respectively Son, T represent transposition operation.DhAnd DvIt is as follows：
    <mrow> <msub> <mi>D</mi> <mi>h</mi> </msub> <mo>=</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mn>0</mn> </mtd> <mtd> <mn>0</mn> </mtd> </mtr> <mtr> <mtd> <mrow> <mo>-</mo> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> </mrow> </mtd> <mtd> <mn>0</mn> </mtd> <mtd> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mn>0</mn> </mtd> <mtd> <mn>0</mn> </mtd> </mtr> </mtable> </mfenced> <msub> <mi>D</mi> <mi>v</mi> </msub> <mo>=</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mrow> <mo>-</mo> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> </mrow> </mtd> <mtd> <mn>0</mn> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mn>0</mn> </mtd> <mtd> <mn>0</mn> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> </mtd> <mtd> <mn>0</mn> </mtd> </mtr> </mtable> </mfenced> </mrow>
    Singular value decomposition is carried out to gradient covariance matrix, its all characteristic value is obtained, then calculates its mark to weigh figure As the texture strength of block, mark shows that more greatly the image block texture information is more.
  6. 6. the picture noise level estimation method based on DCT and gradient covariance matrix, its feature exist according to claim 4 In：If the image filtered out is weak texture region, the image that DCT inverse transformations obtain is exactly clean image not affected by noise, Noisy image and the difference of clean image are exactly the noise in image.
  7. 7. the picture noise level estimation method based on DCT and gradient covariance matrix, its feature exist according to claim 4 In：If what is filtered out is the image block of texture-rich, then by iterating to calculate final noise variance, estimates final make an uproar with this Sound parameter estimation result.
  8. 8. the picture noise level estimation method based on DCT and gradient covariance matrix, its feature exist according to claim 7 In：Iterating to calculate final noise variance mode is, selects 20% minimum image block of mark as weak texture region initially, Initial value by the use of the noise variance that noise data corresponding to these regions calculates as estimated result, by iterating to calculate most Whole noise variance, each time in the upper weak texture region once chosen using 20% at mark as threshold value, further sieve Weak texture region is selected to carry out noise parameter estimation, the difference of front and rear estimated result twice is less than 0.01 stopping iteration, the estimation It is worth for final noise parameter estimated result.
CN201710698252.XA 2017-08-15 2017-08-15 Picture noise level estimation method based on DCT and gradient covariance matrix Pending CN107451981A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710698252.XA CN107451981A (en) 2017-08-15 2017-08-15 Picture noise level estimation method based on DCT and gradient covariance matrix

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710698252.XA CN107451981A (en) 2017-08-15 2017-08-15 Picture noise level estimation method based on DCT and gradient covariance matrix

Publications (1)

Publication Number Publication Date
CN107451981A true CN107451981A (en) 2017-12-08

Family

ID=60492256

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710698252.XA Pending CN107451981A (en) 2017-08-15 2017-08-15 Picture noise level estimation method based on DCT and gradient covariance matrix

Country Status (1)

Country Link
CN (1) CN107451981A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107909586A (en) * 2017-12-11 2018-04-13 厦门美图之家科技有限公司 Picture noise computational methods and device
CN108492261A (en) * 2018-02-07 2018-09-04 厦门美图之家科技有限公司 A kind of image enchancing method and computing device
CN110503625A (en) * 2019-07-02 2019-11-26 杭州电子科技大学 A kind of cmos image signal dependent noise method for parameter estimation
CN112053295A (en) * 2020-08-21 2020-12-08 珠海市杰理科技股份有限公司 Image noise reduction method and device, computer equipment and storage medium
CN113269696A (en) * 2021-07-19 2021-08-17 贝壳技术有限公司 Method for denoising image, electronic device, and medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110085627A1 (en) * 2008-07-14 2011-04-14 Telefonaktiebolaget Lm Ericsson (Publ) Interference rejection combining for multi-user mimo telecommunications system
CN103561194A (en) * 2013-09-16 2014-02-05 湖南大学 Scanned image descreening method based on adaptive filtering
CN105279742A (en) * 2015-11-19 2016-01-27 中国人民解放军国防科学技术大学 Quick image denoising method on the basis of partition noise energy estimation
CN105678774A (en) * 2016-01-11 2016-06-15 浙江传媒学院 Image noise level estimation method based on principal component analysis
CN106658004A (en) * 2016-11-24 2017-05-10 浙江大学 Compression method and device based on characteristics of flat region of image

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110085627A1 (en) * 2008-07-14 2011-04-14 Telefonaktiebolaget Lm Ericsson (Publ) Interference rejection combining for multi-user mimo telecommunications system
CN103561194A (en) * 2013-09-16 2014-02-05 湖南大学 Scanned image descreening method based on adaptive filtering
CN105279742A (en) * 2015-11-19 2016-01-27 中国人民解放军国防科学技术大学 Quick image denoising method on the basis of partition noise energy estimation
CN105678774A (en) * 2016-01-11 2016-06-15 浙江传媒学院 Image noise level estimation method based on principal component analysis
CN106658004A (en) * 2016-11-24 2017-05-10 浙江大学 Compression method and device based on characteristics of flat region of image

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
XINHAO LIU等: "Noise level estimation using weak textured patches of a single noisy image", 《2012 19TH IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING》 *
寇婷婷: "基于高阶曲面拟合和统计推断的图像噪声强度估计", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *
王家文: "《Matlab7.6图形图像处理》", 31 March 2009, 国防工业出版社 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107909586A (en) * 2017-12-11 2018-04-13 厦门美图之家科技有限公司 Picture noise computational methods and device
CN107909586B (en) * 2017-12-11 2020-07-03 厦门美图之家科技有限公司 Image noise calculation method and device
CN108492261A (en) * 2018-02-07 2018-09-04 厦门美图之家科技有限公司 A kind of image enchancing method and computing device
CN110503625A (en) * 2019-07-02 2019-11-26 杭州电子科技大学 A kind of cmos image signal dependent noise method for parameter estimation
CN110503625B (en) * 2019-07-02 2021-08-17 杭州电子科技大学 CMOS image signal related noise parameter estimation method
CN112053295A (en) * 2020-08-21 2020-12-08 珠海市杰理科技股份有限公司 Image noise reduction method and device, computer equipment and storage medium
CN112053295B (en) * 2020-08-21 2024-04-05 珠海市杰理科技股份有限公司 Image noise reduction method, device, computer equipment and storage medium
CN113269696A (en) * 2021-07-19 2021-08-17 贝壳技术有限公司 Method for denoising image, electronic device, and medium

Similar Documents

Publication Publication Date Title
CN107451981A (en) Picture noise level estimation method based on DCT and gradient covariance matrix
Li et al. No-reference image blur assessment based on discrete orthogonal moments
Yu et al. Image denoising using trivariate shrinkage filter in the wavelet domain and joint bilateral filter in the spatial domain
Hyvärinen et al. Sparse code shrinkage: Denoising by nonlinear maximum likelihood estimation
Strela Denoising via block Wiener filtering in wavelet domain
CN103093434B (en) Non-local wiener filtering image denoising method based on singular value decomposition
CN100499741C (en) Method for filtering image noise using pattern information
CN106530244B (en) A kind of image enchancing method
CN103208097B (en) Filtering method is worked in coordination with in the principal component analysis of the multi-direction morphosis grouping of image
Yin et al. Image denoising with anisotropic bivariate shrinkage
CN101719268B (en) Generalized Gaussian model graph denoising method based on improved Directionlet region
WO2005055618A1 (en) Method and system for video quality measurements
Yin et al. Image denoising using trivariate prior model in nonsubsampled dual-tree complex contourlet transform domain and non-local means filter in spatial domain
CN104616249B (en) A kind of Wavelet Transformation of Image Denoising method based on curvature variation
CN103116873B (en) Image denoising method
Bhateja et al. Fast SSIM index for color images employing reduced-reference evaluation
CN108932699A (en) Three-dimensional matching reconciliation filtering image denoising method based on transform domain
Shih et al. A new iterated two-band diffusion equation: theory and its application
Sari et al. Noise estimation by utilizing mean deviation of smooth region in noisy image
CN111311508A (en) Noise reduction method for pavement crack image with noise
CN102496143A (en) Sparse K-SVD noise suppressing method based on chelesky decomposition and approximate singular value decomposition
Jumiawi et al. Image spectrum segmentation for lowpass and highpass filters
Zhang et al. Image denoising using hybrid singular value thresholding operators
CN111652810A (en) Image denoising method based on wavelet domain singular value differential model
CN108376396B (en) High-efficiency PM2.5 concentration prediction method based on image

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20171208