CN105913427A - Machine learning-based noise image saliency detecting method - Google Patents
Machine learning-based noise image saliency detecting method Download PDFInfo
- Publication number
- CN105913427A CN105913427A CN201610222900.XA CN201610222900A CN105913427A CN 105913427 A CN105913427 A CN 105913427A CN 201610222900 A CN201610222900 A CN 201610222900A CN 105913427 A CN105913427 A CN 105913427A
- Authority
- CN
- China
- Prior art keywords
- noise
- image
- amplitude
- denoising
- value
- 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
- 238000010801 machine learning Methods 0.000 title claims abstract description 37
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000001514 detection method Methods 0.000 claims abstract description 39
- 238000002790 cross-validation Methods 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 18
- 230000002146 bilateral effect Effects 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 5
- 238000003708 edge detection Methods 0.000 claims description 3
- 230000006835 compression Effects 0.000 description 5
- 238000007906 compression Methods 0.000 description 5
- 230000004438 eyesight Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 230000000717 retained effect Effects 0.000 description 2
- 210000000697 sensory organ Anatomy 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000003935 attention Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000004424 eye movement Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000010332 selective attention Effects 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 230000008786 sensory perception of smell Effects 0.000 description 1
- 230000014860 sensory perception of taste Effects 0.000 description 1
- 230000016776 visual perception Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
-
- 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
- G06T2207/20081—Training; Learning
Abstract
The invention relates to a machine learning-based noise image saliency detecting method which comprises the following steps: 1, a plurality kinds of denoising parameters are adopted for a noise image of each amplitude, and an optimal denoising parameter for each amplitude is obtained; 2, each noise image is subjected to characteristic extracting operation via a noise assessing algorithm, noise value characteristics are obtained, and a noise value characteristic set is formed; 3, the noise value characteristic set is used as a machine learning algorithm characteristic set, and a noise amplitude prediction model is obtained via a machine learning algorithm and a quinquesection cross validation method; 4, a noise image corresponding to the noise amplitude prediction model is adopted for prediction, and predicted noise amplitude value is obtained; 5, predicted noise amplitude value of each noise image and a corresponding optimal denoising parameter are used for denoising operation, and a denoised image set can be obtained; 6, images in the denoised image set is subjected to saliency detecting operation via a saliency detection method, and a final salient image can be obtained. According to the machine learning-based noise image saliency detecting method, noise image detecting performance can be improved.
Description
Technical field
The present invention relates to image and Video processing and technical field of computer vision, particularly a kind of based on machine learning
Noise image conspicuousness detection method.
Background technology
Human sensory mainly includes vision, sense of smell, the sense of taste, the sense of hearing and sense of touch.The mankind rely on sense organ and accept extraneous transmission
Information.Visual perception account for critically important status in the sense organ of the mankind.The vision system of the mankind can at short notice by
Notice pays close attention to the most of paramount importance part, the part that namely human eye is interested.Along with multimedia era
Arrive, the universal and propagation of digitized image cybertimes of various digital products, all produce every day and communicate substantial amounts of figure
As resource.Although the view data of magnanimity enriches life, but also brings many challenges.
How can efficiently and accurately process these image resources is a problem the most crucial.Researcher is found that
After the Selective Attention Mechanism of human visual system, it is intended to allow computer simulation human visual system, thus propose conspicuousness
Detection method.Conspicuousness detection has been applied to compression of images and coding, image retrieval, image segmentation, target identification and content
Perceptual image scaling etc..As in compression of images and coding, first detect marking area, then marking area is retained more
Details, the most both have compressed image, more more important details can have been retained again.
Vision significance detection has been obtained for reasonable research, but most of conspicuousness detection model is for nothing
Distorted image proposes, and experimental data is undistorted image set.Minority article notes has arrived distorted image to conspicuousness
The impact of detection.Zhang et al. finds that noise, fuzzy and compression change image low-level feature, it is proposed that based on image low layer
The bottom-up conspicuousness detection model of feature.Meanwhile, Zhang et al. finds that image quality distortion can cause Saliency maps
Change, and between the change of Saliency maps and the image quality measure of subjectivity, there is certain contact.Gide and Karam exists
Have evaluated 5 kinds of conspicuousness detection models on the eye movement data collection of image quality measure, evaluated type of distortion comprise fuzzy, make an uproar
Sound and JPEG compression distortion.The low-level feature such as brightness of image and contrast is extracted by Mittal et al., and special based on these
Levy the salient region using machine learning framework prediction JPEG distorted image.Kim and Milanfar proposes for noise image
Conspicuousness detection model of based on non parametric regression framework.
Image in real life is mostly with distortion, as caused by the peripheral hardware such as camera sensor, image processor
Distortion, the photographing device shake distortion that causes of shake and compression of images distortion etc..Making an uproar to improve conspicuousness detection method
Application on acoustic image, the present invention proposes a kind of noise image conspicuousness detection method based on machine learning.
Summary of the invention
It is an object of the invention to provide a kind of noise image conspicuousness detection method based on machine learning, the method can
To improve conspicuousness detection method detection performance in noise image.
For achieving the above object, the technical scheme is that a kind of noise image conspicuousness based on machine learning is examined
Survey method, comprises the following steps:
Step S1: the noise image of each amplitude is respectively adopted multiple denoising parameter and carries out denoising, it is thus achieved that each amplitude
Corresponding optimal denoising parameter;
Step S2: use noise evaluation algorithm to carry out feature extraction every amplitude and noise acoustic image, it is thus achieved that the noise of every amplitude and noise acoustic image
Value tag, forms noise figure feature set with thisP;
Step S3: by noise figure feature setPAs the feature set of machine learning algorithm, and by machine learning algorithm and five deciles
Cross validation method, it is thus achieved that the noise amplitude forecast model of noise image;
Step S4: use noise amplitude forecast model that corresponding noise image is predicted, it is thus achieved that often amplitude and noise acoustic image is pre-
Survey noise amplitude value;
Step S5: the optimal denoising parameter using the prediction noise amplitude value of every amplitude and noise acoustic image corresponding with this amplitude carries out denoising
Process, it is thus achieved that denoising image set;
Step S6: use conspicuousness detection method to detect the image in denoising image set, it is thus achieved that final Saliency maps.
Further, in described step S1, the noise image of each amplitude is respectively adopted multiple denoising parameter
Make an uproar process, it is thus achieved that each amplitude optimal denoising parameter accordingly, specifically include following steps:
Step S11: usenPlant Gassian low-pass filter denoising parameter and the noise image of each amplitude carried out denoising, it is thus achieved that
Each amplitude containsnPlant image collection S after the denoising of denoising parameter;
Step S12: use conspicuousness detection method VA to calculate Saliency maps image collection S after denoising, it is thus achieved that image after denoising
Saliency maps set T;
Step S13: the Saliency maps set T of image after denoising is estimated, for each width by in-service evaluation index PR-AUC
The denoising parameter that degree uses when finding out average PR-AUC peak, obtains the optimal denoising parameter of each amplitude.
Further, in described step S2, noise evaluation algorithm is used to carry out feature extraction every amplitude and noise acoustic image, it is thus achieved that
The noise figure feature of every amplitude and noise acoustic image, forms noise figure feature set with thisP, specifically include following steps:
Step S21: noise image is carried out gray processing process, obtains gray level imageI;
Step S22: use bilateral filtering to process gray level imageI, obtain bilateral filtering result figure;
Step S23: calculate gray level imageIWith bilateral filtering result figureDifference, obtain error imageD;
Step S24: to gray level imageICanny edge detection method is used to obtain edge imageE, to edge imageEUse and expand
Operator expands fringe region, obtains the edge image expanded;
Step S25: calculate noise size assessed value imageM, computing formula is:
, wherein
Wherein,D v Represent pixel in gray level imagevValue,tRepresent pixel,Represent the edge image expanded,NRepresent and expand
Big edge imageMiddle pixeltThe set that value is 0;
Step S26: by noise size assessed value imageMBe evenly dividing be 3 × 3 net region, respectively calculate noise size comment
Valuation imageMFull figure and each net region noise size assessed value, computing formula is:
Wherein,M r Represent corresponding region,r=1,2 ..., 10 represent full figure and 9 net regions respectively,M r,v Represent
RegionrMiddle pixelvValue;It is calculated noise figure feature setP={P 1, P 2, …, P 10}。
Further, in described step S3, by noise figure feature setPAs the feature set of machine learning algorithm, and pass through
Machine learning algorithm and five decile cross validation methods, it is thus achieved that the noise amplitude forecast model of noise image, specifically include following
Step:
Step S31: by noise figure feature setPMiddle characteristic valueP 1, P 2, …, P 10From small to large ord after arrangement, as machine
Feature set F of device learning algorithm, and be divided into random for feature set F five: F1, F2, F3, F4 and F5;
Step S32: using F2, F3, F4 and F5 as the training dataset of machine learning, by its correspondence in image quality measure data
The image fault amplitude in storehouse obtains noise amplitude forecast model M1 as the training label of machine learning, study;
Step S33: repeat step S32, obtains F1, F3, F4 and F5 respectively and predicts mould as noise amplitude during training dataset
Type M2, F1, F2, F4 and F5 are as noise amplitude forecast model M3 during training dataset, and F1, F2, F3 and F5 are as training number
According to noise amplitude forecast model M4 during collection, F1, F2, F3 and F4 are as noise amplitude forecast model M5 during training dataset.
Further, in described step S4, use noise amplitude forecast model that corresponding noise image is predicted, obtain
The prediction noise amplitude value of every amplitude and noise acoustic image, specifically include following steps:
Step S41: use noise amplitude forecast model M1 that the image set that feature set F1 is corresponding is predicted, obtain noise amplitude
Value prediction sets V1;
Step S42: repeat step S41 method, be respectively adopted noise amplitude forecast model M2, M3, M4, M5 to feature set F2,
The image set prediction that F3, F4 are corresponding with F5, obtains noise amplitude value prediction sets V2, V3, V4, V5;
Step S43: integrated noise range value prediction sets V={V1, V2, V3, V4, V5}, obtains the noise width of complete image set
Angle value prediction sets V.
Further, in described step S5, the prediction noise amplitude value using every amplitude and noise acoustic image is corresponding with this amplitude
Optimal denoising parameter carries out denoising, it is thus achieved that denoising image set, specifically includes following steps:
Step S51: for every amplitude and noise acoustic image, find, from noise amplitude value prediction sets V, the noise that this noise image is corresponding
Range value;
Step S52: according to noise amplitude value, uses corresponding optimal denoising parameter that noise image is used Gassian low-pass filter
Process, it is thus achieved that denoising image set FI.
Compared to prior art, the invention has the beneficial effects as follows: first with the noise of machine learning prediction noise image
Amplitude, then uses the optimal denoising parameter being suitable for this amplitude to carry out denoising, finally uses conspicuousness detection method to calculate
The Saliency maps of image after denoising, owing to the present invention is in view of the noise image impact on conspicuousness detection method, therefore, it is possible to
Effective raising conspicuousness detection method detection performance on noise image, can be applicable to image and Video processing, computer
The numerous areas such as vision.
Accompanying drawing explanation
Fig. 1 is the FB(flow block) of the inventive method.
Fig. 2 be sample picture in step S2 of one embodiment of the invention (for more preferable display effect, (d) in Fig. 2,
G the pixel value of () and (h) is mapped to [0,1]).
Fig. 3 is the flowchart of the holistic approach of one embodiment of the invention.
Fig. 4 is one embodiment of the invention Central Plains noise image and the final effect sample picture through step S5 and S6.
Detailed description of the invention
Below in conjunction with the accompanying drawings and specific embodiment the present invention is described in further detail.
The present invention provides a kind of noise image conspicuousness detection method based on machine learning, as shown in figures 1 and 3, bag
Include following steps:
Step S1: the noise image of each amplitude is respectively adopted multiple denoising parameter and carries out denoising, it is thus achieved that each amplitude
Corresponding optimal denoising parameter.In the present embodiment, step S1 specifically includes following steps:
Step S11: (template size is respectively { 3 × 3,5 × 5,7 × 7}, standard to use 9 kinds of Gassian low-pass filter denoising parameters
Difference be respectively 0.5,0.7,0.9}) noise image of each amplitude is carried out denoising, it is thus achieved that each amplitude is gone containing 9 kinds
Image collection S after the denoising of parameter of making an uproar;
Step S12: image collection S after denoising is used conspicuousness detection method VA(Saliency detection via
Absorbing markov chain) calculate Saliency maps, it is thus achieved that the Saliency maps set T of image after denoising;
Step S13: in-service evaluation index PR-AUC(the area under precision-recall curve) to denoising after
The Saliency maps set T of image is estimated, the denoising parameter used when finding out average PR-AUC peak for each amplitude,
Obtain the optimal denoising parameter of each amplitude.
Step S2: use noise evaluation algorithm to carry out feature extraction every amplitude and noise acoustic image, it is thus achieved that every amplitude and noise acoustic image
Noise figure feature, forms noise figure feature set with thisP.In the present embodiment, as in figure 2 it is shown, step S2 specifically includes following step
Rapid:
Step S21: noise image is carried out gray processing process, obtains gray level imageI(such as Fig. 2 (b));
Step S22: use bilateral filtering to process gray level imageI, obtain bilateral filtering result figure(such as Fig. 2 (c));
Step S23: calculate gray level imageIWith bilateral filtering result figureDifference, obtain error imageD(such as Fig. 2 (d));
Step S24: to gray level imageICanny edge detection method is used to obtain edge imageE(such as Fig. 2 (e)), to edge graph
PictureEUse Expanded Operators to expand fringe region, obtain the edge image expanded(such as Fig. 2 (f));
Step S25: calculate noise size assessed value imageM, computing formula is:
, wherein
Wherein,D v Represent pixel in gray level imagevValue,tRepresent pixel,Represent the edge image expanded,NRepresent and expand
Big edge imageMiddle pixeltThe set that value is 0;
Step S26: by noise size assessed value imageM(such as Fig. 2 (g)) be evenly dividing be 3 × 3 net region (such as Fig. 2
(h)), calculate noise size assessed value image respectivelyMFull figure and each net region noise size assessed value, computing formula is:
Wherein,M r Represent corresponding region,r=1,2 ..., 10 represent full figure and 9 net regions respectively,M r,v Represent
RegionrMiddle pixelvValue;It is calculated noise figure feature setP={P 1, P 2, …, P 10}。
Step S3: by noise figure feature setPAs the feature set of machine learning algorithm, and by machine learning algorithm and five
Decile cross validation method, it is thus achieved that the noise amplitude forecast model of noise image.In the present embodiment, step S3 specifically include with
Lower step:
Step S31: by noise figure feature setPMiddle characteristic valueP 1, P 2, …, P 10From small to large ord after arrangement, as machine
Feature set F of device learning algorithm, and be divided into random for feature set F five: F1, F2, F3, F4 and F5;
Step S32: using F2, F3, F4 and F5 as the training dataset of machine learning, by its correspondence in image quality measure data
The image fault amplitude of storehouse TID2013 obtains noise amplitude forecast model M1 as the training label of machine learning, study;
Step S33: repeat step S32, obtains F1, F3, F4 and F5 respectively and predicts mould as noise amplitude during training dataset
Type M2, F1, F2, F4 and F5 are as noise amplitude forecast model M3 during training dataset, and F1, F2, F3 and F5 are as training number
According to noise amplitude forecast model M4 during collection, F1, F2, F3 and F4 are as noise amplitude forecast model M5 during training dataset.
Step S4: use noise amplitude forecast model that corresponding noise image is predicted, it is thus achieved that every amplitude and noise acoustic image
Prediction noise amplitude value.In the present embodiment, step S4 specifically includes following steps:
Step S41: use noise amplitude forecast model M1 that the image set that feature set F1 is corresponding is predicted, obtain noise amplitude
Value prediction sets V1;
Step S42: repeat step S41 method, be respectively adopted noise amplitude forecast model M2, M3, M4, M5 to feature set F2,
The image set prediction that F3, F4 are corresponding with F5, obtains noise amplitude value prediction sets V2, V3, V4, V5;
Step S43: integrated noise range value prediction sets V={V1, V2, V3, V4, V5}, obtains the noise width of complete image set
Angle value prediction sets V.
Step S4: use noise amplitude forecast model that corresponding noise image is predicted, it is thus achieved that every amplitude and noise acoustic image
Prediction noise amplitude value.Specifically include following steps:
Step S41: use noise amplitude forecast model M1 that the image set that feature set F1 is corresponding is predicted, obtain noise amplitude
Value prediction sets V1;
Step S42: repeat step S41 method, be respectively adopted noise amplitude forecast model M2, M3, M4, M5 to feature set F2,
The image set prediction that F3, F4 are corresponding with F5, obtains noise amplitude value prediction sets V2, V3, V4, V5;
Step S43: integrated noise range value prediction sets V={V1, V2, V3, V4, V5}, obtains the noise width of complete image set
Angle value prediction sets V.
Step S5: the optimal denoising parameter using the prediction noise amplitude value of every amplitude and noise acoustic image corresponding with this amplitude is carried out
Denoising, it is thus achieved that denoising image set.In the present embodiment, as shown in Figure 4, step S5 specifically includes following steps:
Step S51: for every amplitude and noise acoustic image, find, from noise amplitude value prediction sets V, the noise that this noise image is corresponding
Range value;
Step S52: according to noise amplitude value, uses corresponding optimal denoising parameter that noise image is used Gassian low-pass filter
Process, it is thus achieved that denoising image set FI.
Step S6: use conspicuousness detection method VA to detect the image in denoising image set FI, it is thus achieved that final
Saliency maps.
The noise image conspicuousness detection method based on machine learning that the present invention provides, it is contemplated that noise image is to significantly
Property detection method impact, excavate the noise width in noise size assessed value feature and image quality measure database TID2013
The association of degree, design obtains machine learning noise prediction model, and combines denoising method and for the parameter of this amplitude setting to figure
As carrying out denoising, the Saliency maps of image after finally employing conspicuousness detection method VA calculates denoising.The method can have
The raising conspicuousness detection method of effect detection performance on noise image, can be applicable to image and Video processing, computer regards
The fields such as feel.
Being above presently preferred embodiments of the present invention, all changes made according to technical solution of the present invention, produced function is made
With during without departing from the scope of technical solution of the present invention, belong to protection scope of the present invention.
Claims (6)
1. a noise image conspicuousness detection method based on machine learning, it is characterised in that comprise the following steps:
Step S1: the noise image of each amplitude is respectively adopted multiple denoising parameter and carries out denoising, it is thus achieved that each amplitude
Corresponding optimal denoising parameter;
Step S2: use noise evaluation algorithm to carry out feature extraction every amplitude and noise acoustic image, it is thus achieved that the noise of every amplitude and noise acoustic image
Value tag, forms noise figure feature set with thisP;
Step S3: by noise figure feature setPAs the feature set of machine learning algorithm, and by machine learning algorithm and five deciles
Cross validation method, it is thus achieved that the noise amplitude forecast model of noise image;
Step S4: use noise amplitude forecast model that corresponding noise image is predicted, it is thus achieved that often amplitude and noise acoustic image is pre-
Survey noise amplitude value;
Step S5: the optimal denoising parameter using the prediction noise amplitude value of every amplitude and noise acoustic image corresponding with this amplitude carries out denoising
Process, it is thus achieved that denoising image set;
Step S6: use conspicuousness detection method to detect the image in denoising image set, it is thus achieved that final Saliency maps.
A kind of noise image conspicuousness detection method based on machine learning the most according to claim 1, it is characterised in that:
In described step S1, the noise image of each amplitude is respectively adopted multiple denoising parameter and carries out denoising, it is thus achieved that each width
The corresponding optimal denoising parameter of degree, specifically includes following steps:
Step S11: usenPlant Gassian low-pass filter denoising parameter and the noise image of each amplitude carried out denoising, it is thus achieved that
Each amplitude containsnPlant image collection S after the denoising of denoising parameter;
Step S12: use conspicuousness detection method VA to calculate Saliency maps image collection S after denoising, it is thus achieved that image after denoising
Saliency maps set T;
Step S13: the Saliency maps set T of image after denoising is estimated, for each width by in-service evaluation index PR-AUC
The denoising parameter that degree uses when finding out average PR-AUC peak, obtains the optimal denoising parameter of each amplitude.
A kind of noise image conspicuousness detection method based on machine learning the most according to claim 1, it is characterised in that:
In described step S2, noise evaluation algorithm is used to carry out feature extraction every amplitude and noise acoustic image, it is thus achieved that making an uproar of every amplitude and noise acoustic image
Sound value tag, forms noise figure feature set with thisP, specifically include following steps:
Step S21: noise image is carried out gray processing process, obtains gray level imageI;
Step S22: use bilateral filtering to process gray level imageI, obtain bilateral filtering result figure;
Step S23: calculate gray level imageIWith bilateral filtering result figureDifference, obtain error imageD;
Step S24: to gray level imageICanny edge detection method is used to obtain edge imageE, to edge imageEUse and expand
Operator expands fringe region, obtains the edge image expanded;
Step S25: calculate noise size assessed value imageM, computing formula is:
, wherein
Wherein,D v Represent pixel in gray level imagevValue,tRepresent pixel,Represent the edge image expanded,NRepresent and expand
Edge imageMiddle pixeltThe set that value is 0;
Step S26: by noise size assessed value imageMBe evenly dividing be 3 × 3 net region, respectively calculate noise size comment
Valuation imageMFull figure and each net region noise size assessed value, computing formula is:
Wherein,M r Represent corresponding region,r=1,2 ..., 10 represent full figure and 9 net regions respectively,M r,v Represent
RegionrMiddle pixelvValue;It is calculated noise figure feature setP={P 1, P 2, …, P 10}。
A kind of noise image conspicuousness detection method based on machine learning the most according to claim 3, it is characterised in that:
In described step S3, by noise figure feature setPAs the feature set of machine learning algorithm, and by machine learning algorithm and five etc.
Point cross validation method, it is thus achieved that the noise amplitude forecast model of noise image, specifically includes following steps:
Step S31: by noise figure feature setPMiddle characteristic valueP 1, P 2, …, P 10From small to large ord after arrangement, as machine
Feature set F of learning algorithm, and be divided into random for feature set F five: F1, F2, F3, F4 and F5;
Step S32: using F2, F3, F4 and F5 as the training dataset of machine learning, by its correspondence in image quality measure data
The image fault amplitude in storehouse obtains noise amplitude forecast model M1 as the training label of machine learning, study;
Step S33: repeat step S32, obtains F1, F3, F4 and F5 respectively and predicts mould as noise amplitude during training dataset
Type M2, F1, F2, F4 and F5 are as noise amplitude forecast model M3 during training dataset, and F1, F2, F3 and F5 are as training number
According to noise amplitude forecast model M4 during collection, F1, F2, F3 and F4 are as noise amplitude forecast model M5 during training dataset.
A kind of noise image conspicuousness detection method based on machine learning the most according to claim 4, it is characterised in that:
In described step S4, use noise amplitude forecast model that corresponding noise image is predicted, it is thus achieved that every amplitude and noise acoustic image
Prediction noise amplitude value, specifically includes following steps:
Step S41: use noise amplitude forecast model M1 that the image set that feature set F1 is corresponding is predicted, obtain noise amplitude
Value prediction sets V1;
Step S42: repeat step S41 method, be respectively adopted noise amplitude forecast model M2, M3, M4, M5 to feature set F2,
The image set prediction that F3, F4 are corresponding with F5, obtains noise amplitude value prediction sets V2, V3, V4, V5;
Step S43: integrated noise range value prediction sets V={V1, V2, V3, V4, V5}, obtains the noise width of complete image set
Angle value prediction sets V.
A kind of noise image conspicuousness detection method based on machine learning the most according to claim 5, it is characterised in that
In described step S5, the optimal denoising parameter using the prediction noise amplitude value of every amplitude and noise acoustic image corresponding with this amplitude is gone
Make an uproar process, it is thus achieved that denoising image set, specifically include following steps:
Step S51: for every amplitude and noise acoustic image, find, from noise amplitude value prediction sets V, the noise that this noise image is corresponding
Range value;
Step S52: according to noise amplitude value, uses corresponding optimal denoising parameter that noise image is used Gassian low-pass filter
Process, it is thus achieved that denoising image set FI.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610222900.XA CN105913427B (en) | 2016-04-12 | 2016-04-12 | Machine learning-based noise image saliency detecting method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610222900.XA CN105913427B (en) | 2016-04-12 | 2016-04-12 | Machine learning-based noise image saliency detecting method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105913427A true CN105913427A (en) | 2016-08-31 |
CN105913427B CN105913427B (en) | 2017-05-10 |
Family
ID=56744939
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610222900.XA Expired - Fee Related CN105913427B (en) | 2016-04-12 | 2016-04-12 | Machine learning-based noise image saliency detecting method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105913427B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108416756A (en) * | 2018-03-26 | 2018-08-17 | 福州大学 | A kind of region perceptual image denoising method based on machine learning |
CN111788476A (en) * | 2018-02-26 | 2020-10-16 | 株式会社高迎科技 | Component mounting state inspection method, printed circuit board inspection apparatus, and computer-readable recording medium |
CN111798658A (en) * | 2019-11-08 | 2020-10-20 | 方勤 | Traffic lane passing efficiency detection platform |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140193066A1 (en) * | 2012-12-10 | 2014-07-10 | Brain Corporation | Contrast enhancement spiking neuron network sensory processing apparatus and methods |
CN104463870A (en) * | 2014-12-05 | 2015-03-25 | 中国科学院大学 | Image salient region detection method |
CN104616259A (en) * | 2015-02-04 | 2015-05-13 | 西安理工大学 | Non-local mean image de-noising method with noise intensity self-adaptation function |
CN105376506A (en) * | 2014-08-27 | 2016-03-02 | 江南大学 | Design of image pattern noise relevance predictor |
-
2016
- 2016-04-12 CN CN201610222900.XA patent/CN105913427B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140193066A1 (en) * | 2012-12-10 | 2014-07-10 | Brain Corporation | Contrast enhancement spiking neuron network sensory processing apparatus and methods |
CN105376506A (en) * | 2014-08-27 | 2016-03-02 | 江南大学 | Design of image pattern noise relevance predictor |
CN104463870A (en) * | 2014-12-05 | 2015-03-25 | 中国科学院大学 | Image salient region detection method |
CN104616259A (en) * | 2015-02-04 | 2015-05-13 | 西安理工大学 | Non-local mean image de-noising method with noise intensity self-adaptation function |
Non-Patent Citations (2)
Title |
---|
张巧荣 等: "利用多尺度频域分析的图像显著区域检测", 《哈尔滨工程大学学报》 * |
钱晓亮 等: "视觉显著性检测:一种融合长期和短期特征的信息论算法", 《电子与信息学报》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111788476A (en) * | 2018-02-26 | 2020-10-16 | 株式会社高迎科技 | Component mounting state inspection method, printed circuit board inspection apparatus, and computer-readable recording medium |
CN111788476B (en) * | 2018-02-26 | 2023-07-07 | 株式会社高迎科技 | Method for inspecting component mounting state, printed circuit board inspection device, and computer-readable recording medium |
CN108416756A (en) * | 2018-03-26 | 2018-08-17 | 福州大学 | A kind of region perceptual image denoising method based on machine learning |
CN108416756B (en) * | 2018-03-26 | 2021-11-02 | 福州大学 | Regional perception image denoising method based on machine learning |
CN111798658A (en) * | 2019-11-08 | 2020-10-20 | 方勤 | Traffic lane passing efficiency detection platform |
Also Published As
Publication number | Publication date |
---|---|
CN105913427B (en) | 2017-05-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11113842B2 (en) | Method and apparatus with gaze estimation | |
US10860879B2 (en) | Deep convolutional neural networks for crack detection from image data | |
EP3333768A1 (en) | Method and apparatus for detecting target | |
US9767363B2 (en) | System and method for automatic detection of spherical video content | |
Islam et al. | Solid waste bin detection and classification using Dynamic Time Warping and MLP classifier | |
CN106874826A (en) | Face key point-tracking method and device | |
WO2014044158A1 (en) | Identification method and device for target object in image | |
CN101179713A (en) | Method of detecting single moving target under complex background | |
CN101930549B (en) | Second generation curvelet transform-based static human detection method | |
CN110659659A (en) | Method and system for intelligently identifying and early warning pests | |
CN107301376B (en) | Pedestrian detection method based on deep learning multi-layer stimulation | |
CN104751153B (en) | A kind of method and device of identification scene word | |
KR101191516B1 (en) | Enhanced image identification | |
CN108108731A (en) | Method for text detection and device based on generated data | |
KR20120019425A (en) | Image processing device, method, and program | |
CN105913427B (en) | Machine learning-based noise image saliency detecting method | |
US20210327041A1 (en) | Image based novelty detection of material samples | |
CN114821102A (en) | Intensive citrus quantity detection method, equipment, storage medium and device | |
CN114445670A (en) | Training method, device and equipment of image processing model and storage medium | |
CN110782413A (en) | Image processing method, device, equipment and storage medium | |
WO2023123924A1 (en) | Target recognition method and apparatus, and electronic device and storage medium | |
CN114462469B (en) | Training method of target detection model, target detection method and related device | |
CN116797590A (en) | Mura defect detection method and system based on machine vision | |
Lin et al. | An antagonistic training algorithm for TFT-LCD module mura defect detection | |
CN104715257B (en) | A kind of image median filter detection method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170510 |
|
CF01 | Termination of patent right due to non-payment of annual fee |