CN104318525B - Space guiding filtering based image detail enhancement method - Google Patents
Space guiding filtering based image detail enhancement method Download PDFInfo
- Publication number
- CN104318525B CN104318525B CN201410552788.7A CN201410552788A CN104318525B CN 104318525 B CN104318525 B CN 104318525B CN 201410552788 A CN201410552788 A CN 201410552788A CN 104318525 B CN104318525 B CN 104318525B
- Authority
- CN
- China
- Prior art keywords
- image
- filtering
- space
- source images
- formula
- 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
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a space guiding filtering based image detail enhancement method. The space guiding filtering based image detail enhancement method is characterized by comprising the following steps of firstly, extracting an edge feature response diagram for a source image through an image edge detection algorithm and normalizing; secondly, respectively establishing binaryzation space indicating diagrams for different gray level intervals, performing gauss convolution on every space indicating diagram, obtaining a space filtering diagram and calculating a weight value of every space filtering diagram; thirdly, calculating an accumulation diagram and performing guiding image filtering on the accumulation diagram to obtain a space guiding diagram; finally, solving a foundation image and a residual image and establishing a space guiding filtering based image detail enhancement model to perform image detail enhancement on the source image. The space guiding filtering based image detail enhancement method can effectively improve the enhancement effect of image details.
Description
Technical field
The present invention relates to image detail enhancement method, more specifically a kind of visual effect for improving image, plus
The interpretation of detail portion bit image and the image detail enhancement method of discernment in strong image.
Background technology
21 century is the information age, and internet develops rapidly, and the portable intelligent mobile device such as mobile phone, iPad has spread all over
People's daily life.Almost all of portable intelligent mobile device all possesses image collecting function.The user in the whole world utilizes hand
Machine or iPad have taken substantial amounts of picture daily, and share on network.But, under the pressure of natural environment residing during photograph taking
Factor or the restriction of capture apparatus, the visual effect of a lot of network pictures is not notable.Then, in order to lift the vision of picture
Effect, a large amount of digital image processing methods are proposed by researcher.In substantial amounts of digital image processing method, image detail strengthens
Method receives the concern of a large number of researchers of academia and industrial quarters in recent years.
In image, minutia usually contains important information, especially in terms of medical image.Yet with noise
With the impact of the factors such as contrast, the visuality of the minutia of these images substantially reduces, and is unfavorable for image detail feature
Effectively utilizes.Image detail strengthens as a kind of image processing method, can not only project the minutia information in image, also
Can weaken or eliminate interference signal.
Currently, multiple based on the image filtering method of edge feature driving be used for image detail strengthen, such as locally draw
This filtering method of pula and navigational figure filtering method.The main target of this kind of image filtering method is not introduce " artifact "
On the premise of strengthen image detail content.But, this kind of filtering method is during image filtering to all regions of image
Pixel adopts unified filtering strength, does not account for the contact between filtering strength and zones of different picture material, result in
Following defect：
(1) when strengthening to image detail, " artifact " needs extraly to be introduced, and reduces the applicable model of method
Enclose；
(2) these filtering methods will be unable to adopt different filtering strengths to different image-regions, reduces filtering effect
Really.
During image detail strengthens, user typically increases to regions comprising very important visual information some in image
By force, and to other normal areas then do not process, keep constant.For example, " blue sky ", " tree " and " mountain peak " is contained for a width
Natural land photo, most of users may want to carry out details enhancing to the region of " mountain peak " or " tree " in image, and right
Common " blue sky " region in daily life keeps constant.If the image-region of different semantemes can be identified exactly, and then
Different filtering strengths are adopted to zones of different, the enhanced effect of image detail will greatly promote.But, due to Low Level Vision
Existing " semantic gap " problem between feature and high semantic content, leads to identify exactly image-region from semantically
Become highly difficult.Therefore, the enhancing effect of current image detail enhancement method all limited it is impossible to reach the requirement of user.
Content of the invention
The present invention is for avoiding the weak point existing for above-mentioned prior art, provide a kind of based on space guiding filtering
Image detail enhancement method, to the enhancing effect to image detail for effective lifting.
The present invention adopts scheme with the following method for solution problem：
A kind of feature of the image detail enhancement method based on space guiding filtering of the present invention is to carry out as follows：
Step 1, for the source images I for m × n for the resolution ratio, I ∈ Rm×n, using source described in Image Edge-Detection operator extraction
The skirt response figure of image I, and described skirt response figure is normalized divided by 255, obtain normalization skirt response matrixM is the length of source images I；N is the width of source images I；
Step 2, by described normalization skirt response matrixTonal range interval [0,1] be evenly dividing into k sub-district
Between Ωi, using formula (1) respectively to described k subinterval ΩiIn each subinterval set up corresponding space instruction figure M (i),
M(i)∈Rm×n, 1≤i≤k；
In formula (1)：(x, y) represents described normalization skirt response matrixThe coordinate of middle element, and corresponding to described
The position of pixel in source images I；1≤x≤m, 1≤y≤n；
Represent described normalization skirt response matrixThe value of middle xth row y column element is located at
I-th subinterval ΩiIn；M(x,y)I () represents the value of xth row y column element in i-th space instruction figure M (i)；
Step 3, using Gaussian convolution successively each space instruction figure is filtered process, obtain k space filtering
Figure IGauss(i), IGauss(i)∈Rm×n；
Step 4, count successively described normalization skirt response matrixMiddle all elements fall first in each subinterval
The number of element, and the weight of each space filtering figure is calculated using formula (2)；
In formula (2), hiRepresent described normalization skirt response matrixIn element fall in i-th subinterval ΩiMiddle unit
The number of element；W (i) represents to fall the weight in i-th space filtering figure；
Step 5, formula (3) calculating of utilizing add up and scheme Sa, Sa∈Rm×n；
Step 6, using described source images I as guiding figure, using navigational figure filtering method to described cumulative figure SaCarry out
Navigational figure filtering process, obtains space guiding figure S, S ∈ Rm×n；
Step 7, using described source images I as guiding figure, using navigational figure filtering method, described source images I itself is entered
Row navigational figure filtering process, obtains base image Ib, Ib∈Rm×n；
Step 8, strengthen model using the image detail shown in formula (4), described source images I is carried out at image detail enhancing
Reason, obtains details and strengthens image Io, Io∈Rm×n；
Io=Ib+S0·S⊙Ir(4)
In formula (4)：S0For filtering strength；S0For scalar；⊙ is Hadamard product signs, represents that two matrixes are corresponding and is multiplied；
IrRepresent residual image, and have Ir=I-Ib.
Compared with the prior art, the present invention has the beneficial effect that：
1st, the present invention constructs a space guiding figure, can be overcome with different semantic regions in approximate evaluation image
The image-region None- identified problem that " semantic gap " is brought.When image detail strengthens, serve a space guiding and make
With.
2nd, the present invention proposes an image detail enhancement method based on space guiding filtering, by space guiding figure with draw
Lead image filtering to combine, form space navigational figure filtering method.In the navigational figure filtering of space, to different figures
As content area is treated with a certain discrimination, using different filtering strengths, overcome original navigational figure filter method due to using unification
The defect brought of filtering strength, when image detail strengthens, so that enhanced image seems more natural, vision is imitated
Fruit becomes apparent from.
3rd, the image detail enhancement method of the present invention simply uses simple low-level visual feature, therefore computation complexity
Low, speed when image is processed, it is possible to obtain good Consumer's Experience.
Brief description
Fig. 1 needs to carry out the enhanced source images of image detail for the present invention；
The space guiding figure that Fig. 2 is set up according to source images by the present invention；
Fig. 3 is using uniform filtering intensity, source images to be carried out with the enhanced image of image detail in prior art；
Fig. 4 carries out image detail using the image detail enhancement method based on space guiding filtering to source images for the present invention
Enhanced image.
Specific embodiment
In the present embodiment, a kind of to be mainly used in visual effect based on the image detail enhancement method of space guiding filtering not good
Picture carry out image detail enhancing, lift image vision conspicuousness.The method can make software APP, be arranged on mobile phone etc. and move
On moved end or on PC end.The feature of the method be propose a kind of space guiding figure, and with original navigational figure filtering method
Combine, form space navigational figure filtering method, for strengthening to the detail content of image.
It is as follows that the inventive method carries out detailed process when image detail strengthens：
Step 1, for the source images I for m × n for the resolution ratio, I ∈ Rm×n, using Image Edge-Detection operator extraction source images
The skirt response figure of I, and skirt response figure is normalized divided by 255, obtain normalization skirt response matrix M is the length of source images I；N is the width of source images I；
For the ease of introducing, in step 1, source images illustrate taking gray level image as a example.If details to be carried out
Enhanced image is coloured image, then the image array of three Color Channels of red, green, blue in coloured image is carried out carefully respectively
Section enhancing is processed, and finally enhanced for the details of three Color Channels image array is merged into a complete coloured image,
It is coloured image and carry out the enhanced image of details.Fig. 1 is the width source images that the present invention uses in test, mainly
The image on one mountain peak, but the visual effect of entirety is not fine, and the detailed information on mountain peak is not abundant.
Image Edge-Detection operator in step 1 can be Sobel edge edge detective operators or Laplace rim detection is calculated
Son, is all more classical edge detection operator, and the function all having correlation in matlab software platform can directly invoke.Image
Fringe region typically all comprise image detail information.
Step 2, will normalize skirt response matrixTonal range interval [0,1] be evenly dividing into k subinterval
Ωi, using formula (1) respectively to k subinterval ΩiIn each subinterval set up corresponding space instruction figure M (i), M (i) ∈
Rm×n, 1≤i≤k；
In formula (1)：(x, y) represents normalization skirt response matrixThe coordinate of middle element, and corresponding in source images I
The position of pixel；1≤x≤m, 1≤y≤n；
Represent normalization skirt response matrixThe value of middle xth row y column element is located at i-th
Subinterval ΩiIn；M(x,y)I () represents the value of xth row y column element in i-th space instruction figure M (i)；
In step 2, space instruction figure is a binary picture, and step 2 is actually in each gray scale layer respectively to normalization
Skirt response matrixCarry out binaryzation.In fact, in piece image, the semantic region of identical is typically at same ash
Degree layer, so space instruction figure reflects the spatial information of image detail content.
In testing, k takes 16 to the inventive method.
Step 3, using Gaussian convolution successively each space instruction figure is filtered process, obtain k space filtering
Figure IGauss(i), IGauss(i)∈Rm×n；
In fact, in step 3 Gaussian convolution method be carried out with continuous three Boxfilter computings approximate it is therefore an objective to
Reduce computational complexity on the premise of the basic holding degree of accuracy.
Step 4, successively statistics normalization skirt response matrixMiddle all elements fall element in each subinterval
Number, and the weight of each space filtering figure is calculated using formula (2)；
In formula (2), hiRepresent normalization skirt response matrixIn element fall in i-th subinterval ΩiMiddle element
Number；W (i) represents the weight of i-th space filtering figure；
The material particular information of image is typically at strong edge region in image, but strong edge is in whole skirt response
Middle proportion is less.The weak edge of image often occupies larger specific gravity in whole skirt response, but weak edge typically by
The formation of noise.In view of this phenomenon, less for proportion strong edge is just distributed high weight, the larger weak edge of proportion by formula (2)
Distribute low weight.Step 4 is actually to normalization skirt response matrixCarry out statistics with histogram, the result of statistics turns
Weights for space filtering figure in step 3.
A lot of image detail enhancement methods before all " are made no exception " for different gray scale layers, using unified filtering
Intensity, this leads to image detail enhancing effect to be had a greatly reduced quality.The inventive method has carried out " treating with a certain discrimination " in step 4.
Step 5, formula (3) calculating of utilizing add up and scheme Sa, Sa∈Rm×n；
Step 3 is carried out the space filtering figure that Gaussian convolution obtains and is multiplied by corresponding weights being added up by formula (3), is tired out
Plus figure.
Step 6, using source images I as guiding figure, using navigational figure filtering method to cumulative figure SaGuide image
Filtering process, obtains space guiding figure S, S ∈ Rm×n；
In step 6 navigational figure filtering method be by Microsoft Research, Asia's vision calculating group doctor He Kaiming 2010
Propose in the European Computer visual conference in year.When source images are guiding figure, navigational figure filtering is exactly a holding figure
Filtering operation as edge.In this step, using navigational figure filtering, figure cumulative in step 5 is filtered, obtains final
Space guiding figure.Guiding figure in space directly reflects the spatial information of image difference semantic content, increases for image detail below
By force, play one " guiding " effect.Fig. 2 is exactly the space guiding figure that the source images according to Fig. 1 are set up.
Step 7, using source images I as guiding figure, figure itself is guided to source images I using navigational figure filtering method
As filtering process, obtain base image Ib, Ib∈Rm×n；
Base image I in this stepbReflect the base image content of low-frequency range.
Step 8, strengthen model using the image detail shown in formula (4), source images I carried out with image detail enhancing process,
Obtain details and strengthen image Io, Io∈Rm×n；
Io=Ib+S0·S⊙Ir(4)
In formula (4)：S0For filtering strength, S0For scalar, ⊙ is Hadamard product signs, represents that two matrixes are corresponding and is multiplied；
IrRepresent residual image, and have Ir=I-Ib.The inventive method test when, filtering strength S0Take 3.Residual image IrReflect height
The image detail content of frequency range.
What formula (4) reflected is that the image detail based on space guiding filtering proposed by the present invention strengthens model.This model is real
It is the improvement of the image detail model shown in formula (5) on border：
Io=Ib+S0·Ir(5)
It is apparent that it is the unified filtering strength adopting that the image detail of formula (5) strengthens model, and formula (4)
Image detail strengthens model and compares formula (5)) have more a space guiding figure S, and guiding figure S in space proposed by the present invention exactly examines
Consider the relation between picture material and filtering strength, take different filtering strengths for different images content area.
Image detail exactly according to formula (5) for the Fig. 3 strengthens model and the source images shown in Fig. 1 is carried out after details enhancing
Image it can be seen that the image detail in Fig. 3 not only " mountain peak " and " trees " region is strengthened, and the figure in " sky " region
As also being strengthened, this subregion is should not to be strengthened, and after enhancing, image integrally seems very lofty, here it is making
With uniform filtering intensity brought the drawbacks of.
Fig. 4 is to strengthen model to the source figure shown in Fig. 1 using the image detail based on space guiding filtering shown in formula (4)
As carrying out enhanced image.Compare Fig. 3, Fig. 4 is strengthened on mountain peak and trees region, and " sky " region is not entered
Row strengthens, and this is just intended to the image enhaucament result obtaining.
More than, only preferably a kind of embodiment of the present invention, any those familiar with the art is at this
In the technical scope of bright exposure, equivalent or relevant parameter change in addition for technology according to the present invention scheme and its inventive concept
Become, all should be included within the scope of the present invention.
Claims (1)
1. a kind of image detail enhancement method based on space guiding filtering, is characterized in that carrying out as follows：
Step 1, for the source images I for m × n for the resolution ratio, I ∈ Rm×n, using source images described in Image Edge-Detection operator extraction
The skirt response figure of I, and described skirt response figure is normalized divided by 255, obtain normalization skirt response matrixM is the length of source images I；N is the width of source images I；
Step 2, by described normalization skirt response matrixTonal range interval [0,1] be evenly dividing into k subinterval
Ωi, using formula (1) respectively to described k subinterval ΩiIn each subinterval set up corresponding space instruction figure M (i), M
(i)∈Rm×n, 1≤i≤k；
In formula (1)：(x, y) represents described normalization skirt response matrixThe coordinate of middle element, and correspond to described source images
The position of pixel in I；1≤x≤m, 1≤y≤n；
Represent described normalization skirt response matrixThe value of middle xth row y column element is located at i-th
Subinterval ΩiIn；M(x,y)I () represents the value of xth row y column element in i-th space instruction figure M (i)；
Step 3, using Gaussian convolution successively each space instruction figure is filtered process, obtain k space filtering figure
IGauss(i), IGauss(i)∈Rm×n；
Step 4, count successively described normalization skirt response matrixMiddle all elements fall in each subinterval element
Number, and the weight of each space filtering figure is calculated using formula (2)；
In formula (2), hiRepresent described normalization skirt response matrixIn element fall in i-th subinterval ΩiMiddle element
Number；W (i) represents the weight of i-th space filtering figure；
Step 5, formula (3) calculating of utilizing add up and scheme Sa, Sa∈Rm×n；
Step 6, using described source images I as guiding figure, using navigational figure filtering method to described cumulative figure SaGuide figure
As filtering process, obtain space guiding figure S, S ∈ Rm×n；
Step 7, using described source images I as guiding figure, using navigational figure filtering method, described source images I itself is drawn
Lead image filtering to process, obtain base image Ib, Ib∈Rm×n；
Step 8, strengthen model using the image detail shown in formula (4), image detail enhancing process carried out to described source images I,
Obtain details and strengthen image Io, Io∈Rm×n；
Io=Ib+S0·S⊙Ir(4)
In formula (4)：S0For filtering strength；S0For scalar；⊙ is Hadamard product signs, represents that two matrixes are corresponding and is multiplied；IrTable
Show residual image, and have Ir=I-Ib.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410552788.7A CN104318525B (en) | 2014-10-17 | 2014-10-17 | Space guiding filtering based image detail enhancement method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410552788.7A CN104318525B (en) | 2014-10-17 | 2014-10-17 | Space guiding filtering based image detail enhancement method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104318525A CN104318525A (en) | 2015-01-28 |
| CN104318525B true CN104318525B (en) | 2017-02-15 |
Family
ID=52373751
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410552788.7A Active CN104318525B (en) | 2014-10-17 | 2014-10-17 | Space guiding filtering based image detail enhancement method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104318525B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9710892B2 (en) * | 2015-10-21 | 2017-07-18 | Chunghwa Picture Tubes, Ltd. | Image enhancement method and image processing apparatus thereof |
| CN112163994B (en) * | 2020-09-01 | 2022-07-01 | 重庆邮电大学 | Multi-scale medical image fusion method based on convolutional neural network |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020917A (en) * | 2012-12-29 | 2013-04-03 | 中南大学 | Method for restoring ancient Chinese calligraphy and painting images on basis of conspicuousness detection |
| EP2608549A2 (en) * | 2011-12-23 | 2013-06-26 | MediaTek Inc. | Method and apparatus for adjusting depth-related information map according to quality measurement result of the depth-related information map |
| KR20130084643A (en) * | 2012-01-17 | 2013-07-25 | 삼성전자주식회사 | Image processing apparatus and method |
| CN103337061A (en) * | 2013-07-18 | 2013-10-02 | 厦门大学 | Rain and snow removing method for image based on multiple guided filtering |
| CN103440630A (en) * | 2013-09-02 | 2013-12-11 | 南京理工大学 | Large-dynamic-range infrared image display and detail enhancement method based on guiding filter |
| CN103955899A (en) * | 2014-05-02 | 2014-07-30 | 南方医科大学 | Dynamic PET image denoising method based on combined image guiding |
| CN104050637A (en) * | 2014-06-05 | 2014-09-17 | 华侨大学 | Quick image defogging method based on two times of guide filtration |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8682087B2 (en) * | 2011-12-19 | 2014-03-25 | Cisco Technology, Inc. | System and method for depth-guided image filtering in a video conference environment |
-
2014
- 2014-10-17 CN CN201410552788.7A patent/CN104318525B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2608549A2 (en) * | 2011-12-23 | 2013-06-26 | MediaTek Inc. | Method and apparatus for adjusting depth-related information map according to quality measurement result of the depth-related information map |
| KR20130084643A (en) * | 2012-01-17 | 2013-07-25 | 삼성전자주식회사 | Image processing apparatus and method |
| CN103020917A (en) * | 2012-12-29 | 2013-04-03 | 中南大学 | Method for restoring ancient Chinese calligraphy and painting images on basis of conspicuousness detection |
| CN103337061A (en) * | 2013-07-18 | 2013-10-02 | 厦门大学 | Rain and snow removing method for image based on multiple guided filtering |
| CN103440630A (en) * | 2013-09-02 | 2013-12-11 | 南京理工大学 | Large-dynamic-range infrared image display and detail enhancement method based on guiding filter |
| CN103955899A (en) * | 2014-05-02 | 2014-07-30 | 南方医科大学 | Dynamic PET image denoising method based on combined image guiding |
| CN104050637A (en) * | 2014-06-05 | 2014-09-17 | 华侨大学 | Quick image defogging method based on two times of guide filtration |
Non-Patent Citations (1)
| Title |
|---|
| "Spatially guided local Laplacian filter for nature image detail enhancement";Shijie Hao et al;《Multimedia Tools and Applications》;20140507;第75卷(第3期);第1529-1542页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104318525A (en) | 2015-01-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2020151489A1 (en) | Living body detection method based on facial recognition, and electronic device and storage medium | |
| CN104103033B (en) | View synthesis method | |
| CN106650770A (en) | Mura defect detection method based on sample learning and human visual characteristics | |
| CN106556940A (en) | A kind of background suppression method in TFT LCD screens automatic optics inspection | |
| CN103136766B (en) | A kind of object conspicuousness detection method based on color contrast and color distribution | |
| CN111915704A (en) | Apple hierarchical identification method based on deep learning | |
| CN106407917A (en) | Dynamic scale distribution-based retinal vessel extraction method and system | |
| CN108830294A (en) | A kind of augmentation method of image data | |
| CN101551853A (en) | Human ear detection method under complex static color background | |
| CN105809173B (en) | A kind of image RSTN invariable attribute feature extraction and recognition methods based on bionical object visual transform | |
| CN106920211A (en) | U.S. face processing method, device and terminal device | |
| CN109858539A (en) | A kind of ROI region extracting method based on deep learning image, semantic parted pattern | |
| CN110288561A (en) | Refractory brick surface scratch recognition methods based on frequency filtering enhancing | |
| CN105389541B (en) | The recognition methods of fingerprint image and device | |
| CN107330360A (en) | A kind of pedestrian's clothing colour recognition, pedestrian retrieval method and device | |
| CN106203326B (en) | A kind of image processing method, device and mobile terminal | |
| CN104392441B (en) | High anti-noise fabric spray rating detecting appraisal method based on image procossing | |
| CN104143102B (en) | Online image processing method | |
| CN107103619A (en) | A kind of processing method of hair grain direction, apparatus and system | |
| CN111860369A (en) | Fraud identification method and device and storage medium | |
| CN110111347A (en) | Logos extracting method, device and storage medium | |
| CN107590785A (en) | A kind of Brillouin spectrum image-recognizing method based on sobel operators | |
| CN104318525B (en) | Space guiding filtering based image detail enhancement method | |
| CN106599891A (en) | Remote sensing image region-of-interest rapid extraction method based on scale phase spectrum saliency | |
| CN109726721A (en) | A kind of Chinese chess recognition methods based on convolutional neural networks model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |