CN112581391B - Image denoising method based on PCNN - Google Patents
Image denoising method based on PCNN Download PDFInfo
- Publication number
- CN112581391B CN112581391B CN202011467189.7A CN202011467189A CN112581391B CN 112581391 B CN112581391 B CN 112581391B CN 202011467189 A CN202011467189 A CN 202011467189A CN 112581391 B CN112581391 B CN 112581391B
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- pcnn
- pixel point
- pixel points
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000000694 effects Effects 0.000 claims abstract description 25
- OIGNJSKKLXVSLS-VWUMJDOOSA-N prednisolone Chemical compound O=C1C=C[C@]2(C)[C@H]3[C@@H](O)C[C@](C)([C@@](CC4)(O)C(=O)CO)[C@@H]4[C@@H]3CCC2=C1 OIGNJSKKLXVSLS-VWUMJDOOSA-N 0.000 claims abstract description 9
- 230000000877 morphologic effect Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 abstract description 16
- 238000010304 firing Methods 0.000 abstract description 4
- 238000001914 filtration Methods 0.000 description 4
- 238000003384 imaging method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000002856 peripheral neuron Anatomy 0.000 description 1
- 230000001681 protective effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 210000000225 synapse Anatomy 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
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
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
-
- 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
- G06T2207/20084—Artificial neural networks [ANN]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a PCNN-based image denoising method, which comprises the following steps: initializing parameters of an input image to enable each pixel point to be in a flameout state; obtaining a threshold value theta of a PCNN model ij Connection strength beta ij Step length delta F and adjusting times N; carrying out iterative processing on the input image by utilizing a PCNN model to obtain the ignition state of each pixel point and the pixel points in the preset field; when the number of the pixel points in the ignition state in a certain pixel point and the pixel points in the preset field is larger than the preset number, reducing the pixel brightness value of the pixel point by the step length delta F; repeating the steps until the adjusting times N are equal to 0; according to the image denoising method disclosed by the invention, the pixels of the image are corrected by simulating the neuron firing value, so that the noise in the image can be effectively removed, the main detail information existing in the image is effectively reserved, the boundary detail information of the image can be reserved, and the denoising effect of the image is improved.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a PCNN-based image denoising method.
Background
In the process of image imaging, the difficulty that we cannot avoid is some factors that interfere with image quality, for example, the image may be contaminated by various noises in the process of generating or transmitting, the imaging device cannot be focused completely, the imaging object is in a shielding state and is not illuminated enough, the image quality may be seriously damaged by design defects existing in the device, and the like, and therefore, it is necessary to perform denoising processing on the image.
The current method for denoising an image includes the following three methods:
1. the mean filtering method is a linear filtering algorithm, which replaces the original pixel value with the average value of all pixels, but the algorithm is easy to cause image boundary blurring in the process of processing the image.
2. The median filtering method is a nonlinear smoothing technology, and the gray value of each pixel point is set as the median of the gray values of all the pixel points in a window of a certain field of the point; however, the algorithm is prone to some degree of boundary distortion during image processing.
3. The transform domain denoising algorithm transforms the filter coefficients using a specific function image and appropriate filtering processing, separates the image signal and the noise signal by inverse transformation into a spatial domain, and filters the noise signal, but the algorithm blurs the boundary of the image since the relative frequency of the image signal may also be removed.
That is, the boundary of the image is affected by the current image denoising method, and thus, the prior art needs to be improved and enhanced.
Disclosure of Invention
In view of the defects of the prior art, the invention aims to provide the image denoising method based on the PCNN, which can effectively remove noise in an image, retain boundary detail information of the image and improve the denoising effect of the image by simulating the neuron firing value to correct the pixels of the image.
In order to achieve the purpose, the invention adopts the following technical scheme:
a PCNN-based image denoising method comprises the following steps:
s100, initializing parameters of an input image to enable each pixel point to be in a flameout state;
s200, obtaining a threshold value theta of a PCNN model ij And the bond strength beta ij And a step size Δ F;
s300, obtaining the adjustment times N;
s400, carrying out iterative processing on the input image by using a PCNN model, and acquiring the ignition state of each pixel point and the pixel points in the preset field;
s500, when the number of the pixel points in the ignition state in a certain pixel point and the pixel points in the preset field is larger than the preset number, executing the step S600;
s600, reducing the pixel brightness value of the pixel point by a step length delta F;
s700, reducing the adjusting times N once, and judging whether the reduced N is equal to 0 or not;
s800, if N is not equal to 0, repeatedly executing the step S400 to the step S700; if N is equal to 0, the image is output to the PCNN model.
The image denoising method based on the PCNN further comprises the following steps:
and S900, carrying out secondary processing on the image output by the PCNN model by adopting digital morphological open operation.
In the image denoising method based on the PCNN, the preset number is 4.
In the PCNN-based image denoising method, the step S500 specifically includes the steps of:
when a certain pixel is in an ignition state and more than 3 pixels in the ignition state among the pixels in the preset field, or when a certain pixel is in a flameout state and more than 4 pixels in the ignition state among the pixels in the preset field, step S600 is executed.
In the image denoising method based on the PCNN, the preset field is 8 fields, and the preset field comprises 8 pixel points connected with the pixel points.
In the PCNN-based image denoising method, the step S400 specifically includes the steps of:
s410, carrying out iterative processing on the input image by utilizing a PCNN model;
s420, calculating the signal intensity of each pixel point and the signal intensity of the pixel points in the preset field;
s430, calculating the internal activity item of each pixel point and the internal activity items of the pixel points in the preset field;
s440, the internal activity item of each pixel point is compared with a threshold value theta ij Comparing, if the internal activity item is larger than the threshold value theta ij Then the pixel is in the ignition state.
In the image denoising method based on PCNN, the signal intensity is L ij The internal activity item is U ij ，
U ij ＝F ij (1+β ij L ij ) Wherein F is ij And inputting the connection of the pixel points.
Has the advantages that:
the invention provides a PCNN-based image denoising method, which can adjust the pixel brightness of a pixel point by acquiring the ignition state of each pixel point and comparing the number of the pixel points in the ignition state of a certain pixel point and a preset field thereof with a threshold value, namely, correcting the pixel by simulating the ignition value of a neuron, and effectively removing noise in a noisy image, so that main detail information and boundary detail information in the image are effectively reserved, namely, the image edge fuzzification caused by the traditional image denoising method is avoided, and the quality of the output denoised image is improved.
Drawings
FIG. 1 is a control flow chart of an image denoising method provided by the present invention;
FIG. 2 is a control flow diagram of one embodiment of step S500 provided by the present invention;
fig. 3 is a control flowchart of an embodiment of step S400 provided in the present invention.
Detailed Description
The invention provides a PCNN-based image denoising method, and in order to make the purpose, technical scheme and effect of the invention clearer and clearer, the invention is further described in detail below by referring to the attached drawings and embodiments.
In the description of the present invention, it should be understood that the terms "mounted," "connected," etc. should be construed broadly, and those skilled in the art can understand the specific meanings of the above terms in the present invention according to specific situations.
PCNN may be referred to as an advanced neural network model technique, which works in a mode close to the processing mode of mammalian neurons for information; the single neuron work of the PCNN model consists of three modules, namely a receiving input information domain, a nonlinear connection domain and a pulse generation domain; the receiving input information domain is composed of two channels and comprises an information feedback input channel F and a linear connection input channel L, wherein the L channel has the main task of receiving synapse input from local neighborhood neurons, and the F channel has the main task of receiving local interconnection information and needing to bear the task of an information input receiving port stimulated by the outside; the pulse generation domain comprises two parts of a pulse generator and a threshold change comparator.
Referring to fig. 1 to 3, the present invention provides a PCNN-based image denoising method, including the steps of:
s100, initializing parameters of an input image to enable each pixel point to be in a flameout state; when the pixel point is in flameout state, the pulse of the pixel point outputs Y ij ＝0。
S200, obtaining a threshold value theta of a PCNN model ij Connection strength beta ij And a step size Δ F;
s300, obtaining an adjustment number N, wherein the adjustment number N refers to the iteration number of the PCNN model, namely the processing number of the image; the threshold value theta ij Connection strength beta ij The step length Δ F and the number of times of adjustment N can be preset by the user.
S400, carrying out iterative processing on the input image by using the PCNN model, and acquiring the ignition state of each pixel point and the pixel points in the preset field.
In one embodiment, the neurons in the PCNN model are iterated according to equations (1-1) to (1-5):
F ij ＝S ij (1-1)
U ij (n)＝F ij (n)[1+βL ij (n)] (1-3)
wherein S is ij The gray value of the pixel point corresponding to the point (i, j) is excited by external input; f ij Is a neuronal input, U ij 、Y ij 、θ ij Within each neuronPartial activity terms, pulse outputs, and dynamic thresholds; if the dynamic threshold is smaller than the internal activity item, the neuron is activated to output a pulse, and then the system feedback rapidly increases the dynamic threshold to a preset threshold V E At the moment, the pulse generator starts to be closed, the adjacent neurons are inhibited, and the pulse output is stopped; the dynamic threshold value decays exponentially along with the change of time, the constraint on adjacent neurons is reduced, when the activity threshold value is smaller than the internal activity item, the neurons continue to be excited and output pulses immediately, and the steps are repeated.
S500, when the number of the pixel points in the ignition state in a certain pixel point and the pixel points in the preset field is larger than the preset number, executing the step S600;
s600, decreasing the step length delta F of the pixel brightness value of the pixel point, namely increasing the gray value of the pixel point;
s700, reducing the adjusting times N once, and judging whether the reduced N is equal to 0 or not;
s800, if N is not equal to 0, repeatedly executing the step S400 to the step S700; if N is equal to 0, the image is output to the PCNN model.
According to the image denoising method based on the PCNN, the ignition state of each pixel point is obtained, the sum of the pixel points in the ignition state in a certain pixel point and the pixel points in the preset field of the certain pixel point is calculated, the sum is compared with a threshold value, and the pixel brightness of the pixel point is adjusted according to the comparison result, so that the gray level of the pixel points belonging to noise information can be increased, and the main detail information and the boundary detail information of an image can be effectively presented; the pixels are corrected by simulating the firing values of the neurons, so that noise in a noisy image can be effectively removed, main detail information and boundary detail information in the image are effectively reserved, and the quality of the output denoised image is improved.
Further, referring to fig. 1, the PCNN-based image denoising method further includes the steps of:
and S900, carrying out secondary processing on the image output by the PCNN model by adopting digital morphological open operation, and further carrying out denoising processing on the image output by the PCNN model, thereby further improving the denoising effect of the output image.
Furthermore, the preset number is 4, and the preset number can be adjusted by a user according to the actual image size.
Further, referring to fig. 2, the step S500 specifically includes the steps of:
when a certain pixel point is in an ignition state and more than 3 pixel points in the ignition state among the pixel points in the preset field, or when a certain pixel point is in a flameout state and more than 4 pixel points in the ignition state among the pixel points in the preset field, executing the step S600; namely, when the number of the pixel points in the ignition state is judged, the two conditions that the pixel points are in the ignition state or the pixel points are not in the ignition state are included.
Further, the preset field is 8 fields, that is, the preset field includes 8 pixels connected to the pixel; in other embodiments, the preset domain may be 4 domains, that is, the preset domain includes 4 pixels connected to the pixel; the user can adjust the preset field according to the actual use condition, such as the size of the image to be processed.
Further, referring to fig. 3, in an embodiment, the step S400 specifically includes the steps of:
s410, carrying out iterative processing on the input image by utilizing a PCNN model, wherein neurons in the PCNN model are iterated according to formulas (1-1) to (1-5).
S420, calculating the signal intensity of each pixel point and the signal intensity of the pixel points in the preset field, wherein the signal intensity of each pixel point can be calculated according to the formula (1-2);
s430, calculating the internal activity item of each pixel point and the internal activity items of the pixel points in the preset field.
S440, the internal activity item of each pixel point is compared with a threshold value theta ij By comparison, if the internal activity item is greater than the threshold θ ij If yes, the pixel point is in an ignition state; when the internal activity item is less than or equal to the threshold value theta ij When the pixel point is in the off stateFire state, at this time, Y ij ＝0。
Further, the signal strength is L ij The internal activity item is U ij ，U ij ＝F ij (1+β ij L ij ) Wherein, F ij The method is characterized in that the method is a connection input of a pixel point, namely a neuron input, and pulse output of peripheral neurons is used as feedback to directly carry out weighted summation and then is used as a connection input of the neurons.
In summary, in the image denoising method based on the PCNN disclosed in the present application, the PCNN model corrects the pixel points of the image by simulating the firing value of the neuron, so that the noise in the image with noise can be effectively removed, and the gray value of the pixel points corresponding to the noise information is increased, so that the main detail information and the boundary detail information existing in the image can be effectively retained and highlighted; in addition, the image denoising method based on the PCNN disclosed by the application utilizes the digital morphology, and the image denoised by the PCNN model is subjected to secondary processing by adopting the open operation, so that the denoising effect of the image can be further improved, and the quality of the output image is improved.
It should be understood that equivalents and modifications of the technical solution and inventive concept thereof may occur to those skilled in the art, and all such modifications and alterations should fall within the protective scope of the present invention.
Claims (4)
1. A PCNN-based image denoising method is characterized by comprising the following steps:
s100, initializing parameters of an input image to enable each pixel point to be in a flameout state;
s200, obtaining a threshold theta of the PCNN model ij Connection strength beta ij And a step size Δ F;
s300, obtaining the adjustment times N;
s400, carrying out iterative processing on the input image by using a PCNN model, and acquiring the ignition state of each pixel point and the pixel points in the preset field;
s500, when the number of the pixel points in the ignition state in a certain pixel point and the pixel points in the preset field is larger than the preset number, executing the step S600, wherein the preset number is 4; specifically, when a certain pixel is in an ignition state and more than 3 pixels in the ignition state among the pixels in the preset field are present, or when a certain pixel is in a flameout state and more than 4 pixels in the ignition state among the pixels in the preset field are present, the step S600 is executed; the preset field is 8 fields, and the preset field comprises 8 pixel points connected with the pixel points;
s600, reducing the pixel brightness value of the pixel point by a step length delta F;
s700, reducing the adjusting times N once, and judging whether the reduced N is equal to 0 or not;
s800, if N is not equal to 0, repeatedly executing the step S400 to the step S700; if N is equal to 0, the image is output to the PCNN model.
2. The PCNN-based image denoising method of claim 1, further comprising the steps of:
and S900, carrying out secondary processing on the image output by the PCNN model by adopting digital morphological open operation.
3. The PCNN-based image denoising method according to claim 1, wherein the step S400 specifically comprises the steps of:
s410, carrying out iterative processing on the input image by utilizing a PCNN model;
s420, calculating the signal intensity of each pixel point and the signal intensity of the pixel points in a preset field;
s430, calculating the internal activity item of each pixel point and the internal activity items of the pixel points in the preset field;
s440, the internal activity item of each pixel point is compared with a threshold value theta ij By comparison, if the internal activity item is greater than the threshold θ ij Then the pixel is in the ignition state.
4. The PCNN-based image denoising method of claim 3, whereinThen, the signal intensity is L ij The internal activity item is U ij ，U ij ＝F ij (1+β ij L ij ) Wherein F is ij And inputting the connection of the pixel points.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011467189.7A CN112581391B (en) | 2020-12-14 | 2020-12-14 | Image denoising method based on PCNN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011467189.7A CN112581391B (en) | 2020-12-14 | 2020-12-14 | Image denoising method based on PCNN |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112581391A CN112581391A (en) | 2021-03-30 |
| CN112581391B true CN112581391B (en) | 2023-04-11 |
Family
ID=75134817
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011467189.7A Active CN112581391B (en) | 2020-12-14 | 2020-12-14 | Image denoising method based on PCNN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112581391B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106251315A (en) * | 2016-08-23 | 2016-12-21 | 南京邮电大学 | A kind of image de-noising method based on full variation |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102707322B (en) * | 2012-02-29 | 2014-11-12 | 北京无线电计量测试研究所 | Human body detection device based on millimeter wave imaging |
| CN102697503B (en) * | 2012-02-29 | 2015-03-18 | 北京无线电计量测试研究所 | Human detection method based on millimeter wave imaging |
| US20200184321A1 (en) * | 2018-12-11 | 2020-06-11 | Fotonation Limited | Multi-processor neural network processing apparatus |
| CN111583227B (en) * | 2020-05-08 | 2023-03-24 | 华侨大学 | Method, device, equipment and medium for automatically counting fluorescent cells |
-
2020
- 2020-12-14 CN CN202011467189.7A patent/CN112581391B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106251315A (en) * | 2016-08-23 | 2016-12-21 | 南京邮电大学 | A kind of image de-noising method based on full variation |
Non-Patent Citations (1)
| Title |
|---|
| 基于简化的PCNN在超声乳腺癌图像去噪方面的应用;陆玉婧等;《生物医学工程研究》;20130625(第02期);第80-83页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112581391A (en) | 2021-03-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110047055B (en) | Infrared image detail enhancement and denoising method | |
| Turkmen | Efficient impulse noise detection method with ANFIS for accurate image restoration | |
| CN108961172B (en) | Gamma correction-based image contrast enhancement method | |
| US20090074318A1 (en) | Noise-reduction method and apparatus | |
| CN117252773A (en) | Image enhancement method and system based on self-adaptive color correction and guided filtering | |
| CN112991199A (en) | Image high-low frequency decomposition noise removing method based on residual error dense network | |
| Tsutsui et al. | Halo artifacts reduction method for variational based realtime retinex image enhancement | |
| CN113902638B (en) | Video noise reduction method combining convolutional neural network and light self-adaptive median filtering | |
| CN109064413B (en) | Image contrast enhancement method and image acquisition medical equipment adopting same | |
| CN112581391B (en) | Image denoising method based on PCNN | |
| Jiang et al. | An effective adaptive median filter algorithm for removing salt & pepper noise in images | |
| CN115965552B (en) | Frequency-space-time domain joint denoising and recovering system for low signal-to-noise ratio image sequence | |
| CN116883259A (en) | Underwater image enhancement method based on denoising diffusion probability model | |
| Panchaxri et al. | Image denoising using adaptive NL means filtering with method noise thresholding | |
| CN115829870A (en) | Image denoising method based on variable scale filtering | |
| CN104966271A (en) | Image denoising method based on biological vision receptive field mechanism | |
| Subah et al. | Underwater image enhancement based on fusion technique via color correction and illumination adjustment | |
| Konieczka et al. | Method of adaptive pixel averaging for impulse noise reduction in digital images | |
| Pomanysochka et al. | Soft computing techniques for noise filtration in the image recognition processes | |
| Chowdhury et al. | Eradication of Salt and Pepper Noise from a Tumorous MRI image using SNPRB Filter | |
| RU2405200C2 (en) | Method and device for fast noise filtration in digital images | |
| Huang et al. | An improved retinex algorithm for underwater image enhancement based on HSV model | |
| Kumar et al. | Enhancement of image sharpness with bilateral and adaptive filter | |
| Albu | Linear prediction based image enhancement method | |
| CN112288643A (en) | Image denoising method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |