CN108492245A - Low light images based on wavelet decomposition and bilateral filtering are to fusion method - Google Patents
Low light images based on wavelet decomposition and bilateral filtering are to fusion method Download PDFInfo
- Publication number
- CN108492245A CN108492245A CN201810119668.6A CN201810119668A CN108492245A CN 108492245 A CN108492245 A CN 108492245A CN 201810119668 A CN201810119668 A CN 201810119668A CN 108492245 A CN108492245 A CN 108492245A
- Authority
- CN
- China
- Prior art keywords
- band
- sub
- images
- short exposed
- collection
- 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
- 238000001914 filtration Methods 0.000 title claims abstract description 43
- 230000002146 bilateral effect Effects 0.000 title claims abstract description 30
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 27
- 238000007500 overflow downdraw method Methods 0.000 title claims abstract description 13
- 230000004927 fusion Effects 0.000 claims abstract description 31
- 238000000034 method Methods 0.000 claims abstract description 9
- 230000015572 biosynthetic process Effects 0.000 claims abstract 2
- 238000003786 synthesis reaction Methods 0.000 claims abstract 2
- 238000001514 detection method Methods 0.000 claims description 21
- 230000009467 reduction Effects 0.000 claims description 14
- 238000011946 reduction process Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 9
- 238000009499 grossing Methods 0.000 claims description 8
- 230000014509 gene expression Effects 0.000 claims description 4
- 230000000694 effects Effects 0.000 abstract description 2
- 238000012986 modification Methods 0.000 abstract 1
- 230000004048 modification Effects 0.000 abstract 1
- 238000004364 calculation method Methods 0.000 description 7
- 230000004075 alteration Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000002156 mixing Methods 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 206010016322 Feeling abnormal Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 230000000574 ganglionic effect Effects 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/14—Transformations for image registration, e.g. adjusting or mapping for alignment of images
- G06T3/147—Transformations for image registration, e.g. adjusting or mapping for alignment of images using affine transformations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Facsimile Image Signal Circuits (AREA)
Abstract
The invention discloses a kind of low light images based on wavelet decomposition and bilateral filtering to fusion method.Shooting obtains two images of short exposure and long exposure under low light conditions, to obtain a high quality graphic, the present invention chooses short exposed images and is used as with reference to figure, using the brightness of color histogram Matching and modification and color, is registrated and is aligned to short exposed images with duration exposure image.Two images after registration carry out wavelet decomposition respectively, bilateral filtering is carried out to the low frequency sub-band of short exposed images, high frequency time band carries out hard -threshold filtering and achievees the effect that reduce noise, the fusion weight map per level band is then calculated according to sub-band strength difference, according to fusion weight map into the reconstruct of places band is brought to each level band of short exposed images and each level of long exposure image, the sub-band of final small echo synthesis reconstruct obtains result images.The edge sharpness that the method for the present invention maintains short exposed images remains brightness and the color of long exposure image again, and effectively inhibits noise.
Description
Technical field
The invention belongs to digital image processing fields, are related to a kind of low light images based on wavelet decomposition and bilateral filtering
To fusion method.
Background technology
When taking pictures under low light conditions, due to insufficient light, the image for shooting gained includes often much noise, and
Color is dim, and contrast is relatively low, we can obtain the lower image of rich in color and noise by prolonging exposure time, but by
Longer in the time for exposure, the movement of taken the photograph object often causes the On Local Fuzzy of image.Example images are as shown in Figure 1.We are uncommon
It hopes and a preferable image of quality is obtained by suitably fusion so that keep the edge of short exposed images sharp in result images
Degree, and possess the color of long exposure image and brightness and reduce noise.The high dynamic of traditional multi-frame exposure image goes ghost to merge
Technology generally requires multiple continuous shooting images, and does not consider Denoising Problems.When input picture only has long short exposure two to open image, this
Traditional methods are susceptible to ghost and noise leftover problem due to information deficiency.
Invention content
The purpose of the present invention is by being combined wavelet decomposition with multiple dimensioned bilateral filtering, inhibiting the same of noise and ghost
When reconstruction image so that the edge sharpness of short exposed images is kept in result images, but possess the color of long exposure image with
Brightness simultaneously reduces noise.
To achieve the above objectives, the present invention uses following technical scheme:It is a kind of low based on wavelet decomposition and bilateral filtering
Luminosity image is to fusion method, and this approach includes the following steps:
1. pair input picture carries out wavelet decomposition and multiple dimensioned bilateral filtering, include the following steps:
1-1 carries out Histogram Matching to short exposed images first and generates image Ih, referring next to short after Histogram Matching
Exposure image carries out registration alignment to long exposure image and generates image Ir;
1-2 is to step 1-1 treated two image Ih、IrWavelet decomposition is carried out respectively, generates corresponding short exposed images
Sub-band collectionWith long exposure image sub-band collection
The short exposed images sub-band collection that 1-3 estimating steps 1-2 is obtainedNoise variance, formula is:
Wherein, σLFor the noise estimate variance of L level bands, median () expressions take median operation, HHLIndicate small wavelength-division
Solve L layers of high frequency detail layer, C1For constant parameter, C1Size determine estimation noise variance size, C1It obtains more greatly
Denoising image is more smooth, and general value range is 2 to 4.
1-4 is to reduce the interference that noise detects follow-up ghost in short exposed images, the short exposure figure obtained to step 1-2
As sub-band collectionLow frequency sub-band carry out bilateral filtering, formula is:
Wherein,Indicate the low frequency sub-band of L layers of short exposed images sub-band collection,After indicating bilateral filtering
Low frequency sub-band, C are normalized function, σdFor filter window parameter, σLFor the noise estimate variance of L level bands, N indicates definition
Domain, x, y, k, l indicate location of pixels coordinate.
The short exposed images sub-band collection that step 1-2 is obtainedHigh frequency time band collection carry out hard -threshold filtering, formula is:
Wherein,Indicate i-th of high frequency time band of the short exposed images that step 1-2 is obtained,After indicating noise reduction filtering
I-th of high frequency time band of short exposed images, σLFor the noise estimate variance of L level bands.
2. the long exposure image sub-band collection obtained using step 1-2 and the short exposed images after step 1-4 denoisings
Band collection obtains corresponding sub-band difference, and then obtains fusion weight map corresponding with sub-band, includes the following steps:
2-1 calculates the long exposure image sub-band collection that step 1-2 is obtained and the short exposed images after step 1-4 denoisings
The absolute value D of corresponding sub-band difference with collectioni, calculation formula is:
Wherein,For the long exposure image sub-band collection that step 1-2 is obtained,For the short exposure after step 1-4 noise reduction process
Image sub-band collection, DiFor the absolute difference of i-th of sub-band of short exposed images after noise reduction and i-th of sub-band of long exposure image.
2-2 utilizes mediant estimation ghost detection threshold value T of each level with differencei, formula is as follows:
Ti=C2*median(Di) (14)
Wherein, TiFor the ghost detection threshold value of i-th of sub-band, DiIt is exposed with long for i-th of sub-band of short exposed images after noise reduction
The absolute difference of i-th of sub-band of image, C2For constant parameter, C2Size determine ghost detection level, C2It detects more greatly
Ghost region is smaller, and general value range is 2 to 5.
2-3 obtains fusion weight map using the ghost detection threshold value of absolute value of each level with difference and estimation, for reality
Existing natural fusion, we build smooth weighting function using Gaussian function characteristic, and using Steerable filter to merging weight map
Smoothing processing is done, calculation formula is as follows:
Wherein, WiFor the fusion weight map of i-th of sub-band, TiFor the ghost detection threshold value of estimation, DiFor short exposure after noise reduction
The absolute difference of i-th of sub-band of image and i-th of sub-band of long exposure image, G () indicate Steerable filter smooth operation.
3. rebuilding sub-band collection obtains result images, include the following steps:
Short exposed images after long exposure image sub-band collection that 3-1 is obtained using step 1-2, step 1-4 denoisings
New sub-band collection is built with the fusion weight map after collection and step 2-3 smoothing processings, calculation formula is:
Wherein, FiFor the new sub-band collection of structure,For the long exposure image sub-band collection that step 1-2 is obtained,For step
Short exposed images sub-band collection after 1-4 noise reduction process, WiFor the fusion weight map of i-th of sub-band.
The sub-band collection that 3-2 is synthesized using small echo, obtains final image.
Beneficial effects of the present invention:For photographed scene be low luminosity the case where, under low light conditions shooting obtain it is short
Two images of exposure and long exposure, short exposed images are since under-exposed whole partially dark and there are much noise, long exposure images
Since there are On Local Fuzzies for taken the photograph bulk motion.Short exposed images and long exposure image of the present invention using shooting gained, pass through
Wavelet decomposition image inhibits noise with bilateral filtering and hard -threshold filtering, while estimating ghost using the absolute value of sub-band difference
Shadow detection threshold value, and then smooth fusion weight map is built using Gaussian function characteristic, then rebuild sub-band so that final result
Not only the sharpened edge of short exposed images had been maintained in image, but also has kept color and the brightness of long exposure image, while effectively having been inhibited
Noise.The present invention improves image image quality and visual effect.
Description of the drawings
Fig. 1 is the example images shot under low light conditions, and (a) is short exposed images, is (b) long exposure image.
Fig. 2 is the flow diagram of the method for the present invention.
Fig. 3 is that color histogram matches example images, and (a) is short exposed images, is (b) long exposure image, is (c) overall situation
Histogram Matching result.
Fig. 4 is the power that small echo sub-band collection layer corresponding fusion weight map W, (a)-(d) are respectively subband LL, LH, HL, HH
Multigraph.
Fig. 5 is experimental result picture, and (a) is short exposed images, is (b) final result of the method for the present invention, is (c) long exposure
Image.
Specific implementation mode
Below in conjunction with the drawings and specific embodiments, the invention will be further described.
The present invention is directed to the case where taking pictures under low light conditions, and at this time due to insufficient light, the image for shooting gained is frequent
Including much noise, and color is dim, contrast is relatively low, and it is relatively low to obtain rich in color and noise by prolonging exposure time
Image, but since the time for exposure is longer, the movement of taken the photograph object often causes image local fuzzy.Example images such as Fig. 1 institutes
Show.It is desirable that by suitably merging to obtain a preferable image of quality so that keep short exposed images in result images
Edge sharpness, and possess the color of long exposure image and reduce noise.The flow of the present invention is as shown in Fig. 2, includes mainly
Wavelet decomposition, utilizes several steps such as sub-band mathematic interpolation weight map and sub-band reconstruction at multiple dimensioned bilateral filtering.
Step 1. carries out wavelet decomposition and multiple dimensioned bilateral filtering to input picture
1-1 carries out Histogram Matching to short exposed images first and generates image Ih, Histogram Matching is in tri- channels R, G, B
It carries out respectively, example images are as shown in Figure 3.To long exposure image, we take Fast Corner Detection to coordinate RANSAC methods, with
I after Histogram MatchinghFor reference, to long exposure image IlAffine transformation is carried out, is obtained and short exposed images IsThe long of registration exposes
Light image Ir;
1-2 is to the input picture I after Histogram Matching and registration processh、IrWavelet decomposition is carried out respectively, generates short exposure
Sub-band collectionSub-band collection is exposed with long
1-3 is due to image setIncluding noise, to the sub-band of its low-pass filteringBilateral filtering is carried out, to its high frequency
Sub-bandHard -threshold filtering is carried out, we define the sub-band after noise reduction and areWe are first depending on tradition side
Method estimates that every layer of noise level is as follows:
Wherein, σLFor the noise estimate variance of L level bands, median () expressions take median operation, HHLIndicate small wavelength-division
Solve L layers of high frequency detail layer, C1For constant parameter, C1Size determine estimation noise variance size, C1It obtains more greatly
Denoising image is more smooth, and general value range is 2 to 4.
To reduce the interference that noise detects follow-up ghost in short exposed images, and more preferably preserve image detail and side
Edge, we refer to multiple dimensioned bilateral filtering, to low frequency sub-bandCarry out bilateral filtering:
Wherein,Indicate the low frequency sub-band of L layers of short exposed images sub-band collection,After indicating bilateral filtering
Low frequency sub-band, C are normalized function, and N indicates that domain, x, y, k, l indicate location of pixels coordinate, σdFor filter window parameter, σL
For the noise estimate variance of L level bands, median () expressions take median operation, HHLIndicate that the high frequency of L layers of wavelet decomposition is thin
Ganglionic layer, C1For constant parameter, C1Size determine estimation noise variance size, C1The denoising image obtained more greatly is more smooth,
General value range is 2 to 4.
For high frequency time bandWe carry out hard -threshold filtering, and the calculation formula that this method utilizes is:
Wherein,Indicate i-th of high frequency time band of short exposed images,Indicate the short exposed images after noise reduction filtering
I-th of high frequency time band, σLFor the noise estimate variance of L level bands.
Step 2. utilizes the step 1-2 long exposure image sub-band collection obtained and the short exposure figure after step 1-4 denoisings
As sub-band collection, corresponding sub-band difference is obtained, and then obtain fusion weight map corresponding with sub-band, included the following steps:
2-1 calculates the long exposure image sub-band collection that step 1-2 is obtained and the short exposed images after step 1-4 denoisings
The absolute value D of corresponding sub-band difference with collectioni, calculation formula is:
Wherein,For the long exposure image sub-band collection that step 1-2 is obtained,For the short exposure after step 1-4 noise reduction process
Image sub-band collection, DiFor the absolute difference of i-th of sub-band of short exposed images after noise reduction and i-th of sub-band of long exposure image.
Although the short exposed images after 2-2 noise reductions are close with long exposure image brightness but still there is fine difference, noise reduction process
Although bilateral filtering in reduces influence of the noise to ghost region detection in short exposed images with hard -threshold filtering, also holds
Minor detail is easily caused by the residual of the faint noise of transitions smooth or part.Based on this, we are considering to utilize corresponding sub-band poor
When being worth absolute value structure fusion weight, larger difference can be determined that be caused by ghost, and smaller difference may be then by micro-
Weak texture, color error ratio or residual noise deviation causes, and still needs to be determined as non-ghost area.Therefore we need a difference threshold
Value divides ghost area and non-ghost area, it is assumed that ghost area area proportion in entire image is smaller, we can choose image
Difference mediant estimation ghost detection threshold value Ti, formula is as follows:
Ti=C2*median(Di) (22)
Wherein, TiFor the ghost detection threshold value of i-th of sub-band, DiIt is exposed with long for i-th of sub-band of short exposed images after noise reduction
The absolute difference of i-th of sub-band of image, C2For constant parameter, C2Size determine ghost detection level, C2It detects more greatly
Ghost region is smaller, and general value range is 2 to 5.
2-3 obtains fusion weight map using the ghost detection threshold value of absolute value of each level with difference and estimation, for reality
Existing natural fusion, we build smooth weighting function using Gaussian function characteristic, and using Steerable filter to merging weight map
Smoothing processing is done, calculation formula is as follows:
Wherein, WiFor the fusion weight map of i-th of sub-band, TiFor the ghost detection threshold value of estimation, DiFor short exposure after noise reduction
The absolute difference of i-th of sub-band of image and i-th of sub-band of long exposure image, G () indicate Steerable filter smooth operation.From upper
Formula can be seen that when two images difference is smaller, and fusion weight is close to 1, when two images difference is detected much larger than ghost
When threshold value, fusion weight tends to 0.
3. rebuilding sub-band collection obtains result images, include the following steps:
Short exposed images after long exposure image sub-band collection that 3-1 is obtained using step 1-2, step 1-4 denoisings
New sub-band collection is built with the fusion weight map after collection and step 2-3 smoothing processings, calculation formula is:
Wherein, FiFor the new sub-band collection of structure,For the long exposure image sub-band collection that step 1-2 is obtained,For step
Short exposed images sub-band collection after 1-4 noise reduction process, WiFor the fusion weight map of i-th of sub-band.As can be seen from the above equation, when two
When width image difference is smaller, for fusion weight close to 1, result images information more comes from long exposure image, is merged in ghost area
Weight tends to 0, and result images information more comes from the short exposed images after denoising.
The sub-band collection that 3-2 is synthesized using small echo, obtains final image.
Involved wavelet basis is ' db8 ' wavelet basis that noise reduction process is commonly used in this experiment wavelet decomposition, and fusion is calculated
Fado multi-scale wavelet Decomposition order is set as 2, the constant parameter C in blending algorithm1With C2Be 3, the window parameter of Steerable filter with
The window parameter σ of quick bilateral filteringdIdentical value is all 7.When image resolution ratio increases, can suitably be adjusted with reference to empirical value
It is whole, to obtain the blending image with good visual effect.
Part of detecting image of the present invention obtains experiment effect, and example images are as shown in figure 5, it will be seen that result figure
As remaining the edge sharpness of short exposed images in ghost region, such as the facial area, the second row of spadger in the first row image
The eyes of Ms in image;It is close in the whole chroma-luminance information of the non-ghost area results image such as background and long exposure image, such as
Floor area in Fig. 5 (c).Although color histogram matching has adjusted brightness and the color of short exposed images, image in Fig. 3
In retain much noise and partial color deviation, as floor area noise is apparent and with the presence of aberration in Fig. 3 (c), but we
Floor area noise and color error ratio in result images significantly reduce, closer to long exposure image information;And our result figures
As whole without apparent noise, such as the clothes of little girl in the third line image.Experimental result image integral transition of the present invention is not naturally,
Situations such as there are apparent aberration, area joints, achieves the desired results.
Claims (8)
1. a kind of low light images based on wavelet decomposition and bilateral filtering are to fusion method, which is characterized in that this method includes
Following steps:
(1) wavelet decomposition and multiple dimensioned bilateral filtering are carried out to input picture, specifically:
(1.1) color histogram matching is carried out to short exposed images with reference first to long exposure image, referring next to Histogram Matching
Short exposed images afterwards carry out registration alignment to long exposure image;
(1.2) to step (1.1), treated that two images carry out wavelet decomposition respectively, generates corresponding short exposed images sub-band
Collection and long exposure image sub-band collection;
(1.3) the high frequency time band that the short exposed images sub-band that step (1.2) obtains is concentrated is utilized to estimate the noise side of each level band
Difference;
(1.4) to step (1.2) obtain short exposed images sub-band collection low frequency sub-band carry out bilateral filtering, high frequency time band collection into
Row hard -threshold filters, and reaches and inhibits noise purpose;
(2) the long exposure image sub-band collection that step (1.2) obtains and the short exposed images after step (1.4) noise reduction process are utilized
Band collection obtains the difference of corresponding sub-band, and then estimates ghost detection threshold value, obtains fusion weight map corresponding with sub-band, specifically
It is:
(2.1) the long exposure image sub-band collection and the short exposed images after step (1.4) noise reduction process that step (1.2) obtains are calculated
The absolute value of the corresponding sub-band difference of sub-band collection;
(2.2) mediant estimation ghost detection threshold value of each level with absolute difference is utilized;
(2.3) fusion weight map is obtained using the ghost detection threshold value of absolute value of each level with difference and estimation, and utilizes and leads
Smoothing processing is done to fusion weight map to filtering;
(3) it rebuilds sub-band collection and obtains result images, specifically:
(3.1) long exposure image sub-band collection, the short exposed images after step (1.4) denoising that step (1.2) obtains are utilized
Fusion weight map after sub-band collection and step (2.3) smoothing processing builds new sub-band collection;
(3.2) the sub-band collection for utilizing small echo synthesis step (3.1) structure, obtains final result image.
2. the low light images according to claim 1 based on wavelet decomposition and bilateral filtering are to fusion method, feature
It is, in the step (1.3), the high frequency time band of the short exposed images sub-band collection obtained using step (1.2) estimates each level
The formula of the noise variance of band is:
Wherein, σLFor the noise estimate variance of L level bands, median () expressions take median operation, HHLIndicate wavelet decomposition L
The high frequency detail layer of layer, C1For constant parameter, C1Size determine estimation noise variance size, C1The denoising obtained more greatly
Image is more smooth, and general value range is 2 to 4.
3. the low light images according to claim 1 based on wavelet decomposition and bilateral filtering are to fusion method, feature
It is, in the step (1.4), to the low frequency sub-band progress bilateral filtering for the short exposed images sub-band collection that step (1.2) obtains
Formula be:
Wherein,Indicate the low frequency sub-band of L layers of short exposed images sub-band collection,Indicate the low frequency after bilateral filtering
Sub-band, C are normalized function, and N indicates that domain, x, y, k, l indicate location of pixels coordinate, σdFor filter window parameter, σLIt is
The noise estimate variance of L level bands.
4. the low light images according to claim 1 based on wavelet decomposition and bilateral filtering are to fusion method, feature
It is, in the step (1.4), to the high frequency time band collection progress hard -threshold for the short exposed images sub-band collection that step (1.2) obtains
The formula of filtering is:
Wherein,Indicate i-th of high frequency time band of the short exposed images that step (1.2) obtains,Indicate short after noise reduction filtering
I-th of high frequency time band of exposure image, σLFor the noise estimate variance of L level bands.
5. the low light images according to claim 1 based on wavelet decomposition and bilateral filtering are to fusion method, feature
It is, in the step (2.1), after the long exposure image sub-band collection and step (1.4) noise reduction process that are obtained using step (1.2)
Short exposed images sub-band collection, the formula that the absolute value of corresponding sub-band difference is calculated is:
Wherein,For the long exposure image sub-band collection that step (1.2) obtains,For the short exposure after step (1.4) noise reduction process
Image sub-band collection, DiFor the absolute difference of i-th of sub-band of short exposed images after noise reduction and i-th of sub-band of long exposure image.
6. the low light images according to claim 1 based on wavelet decomposition and bilateral filtering are to fusion method, feature
It is, in the step (2.2), the formula using mediant estimation ghost detection threshold value of each level with difference is:
Ti=C2*median(Di) (6)
Wherein, TiFor the ghost detection threshold value of i-th of sub-band, DiFor i-th of sub-band of short exposed images after noise reduction and long exposure image
The absolute difference of i-th of sub-band, C2For constant parameter, C2Size determine ghost detection level, C2The ghost detected more greatly
Region is smaller, and general value range is 2 to 5.
7. the low light images according to claim 1 based on wavelet decomposition and bilateral filtering are to fusion method, feature
It is, in the step (2.3), fusion weight is obtained using the ghost detection threshold value of absolute value of each level with difference and estimation
Figure, and the formula of smoothing processing is done to fusion weight map using Steerable filter and is:
Wherein, WiFor the fusion weight map of i-th of sub-band, TiFor the ghost detection threshold value of estimation, DiFor short exposed images after noise reduction
The absolute difference of i-th sub-band and i-th of sub-band of long exposure image, G () indicate Steerable filter smooth operation.
8. the low light images according to claim 1 based on wavelet decomposition and bilateral filtering are to fusion method, feature
It is, in the step (3.1), after the long exposure image sub-band collection that is obtained using step (1.2), step (1.4) denoising
Short exposed images sub-band collection and step (2.3) smoothing processing after fusion weight map build the formula of new sub-band collection and be:
Wherein, FiFor the new sub-band collection of structure,For the long exposure image sub-band collection that step (1.2) obtains,For step
(1.4) the short exposed images sub-band collection after noise reduction process, WiFor the fusion weight map of i-th of sub-band.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810119668.6A CN108492245B (en) | 2018-02-06 | 2018-02-06 | Low-luminosity image pair fusion method based on wavelet decomposition and bilateral filtering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810119668.6A CN108492245B (en) | 2018-02-06 | 2018-02-06 | Low-luminosity image pair fusion method based on wavelet decomposition and bilateral filtering |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108492245A true CN108492245A (en) | 2018-09-04 |
CN108492245B CN108492245B (en) | 2020-06-30 |
Family
ID=63344607
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810119668.6A Expired - Fee Related CN108492245B (en) | 2018-02-06 | 2018-02-06 | Low-luminosity image pair fusion method based on wavelet decomposition and bilateral filtering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108492245B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110322409A (en) * | 2019-06-14 | 2019-10-11 | 浙江大学 | A kind of modified wavelet image fusion method based on label figure |
CN110428391A (en) * | 2019-08-02 | 2019-11-08 | 上海兆芯集成电路有限公司 | To remove the image interfusion method and device of ghost illusion |
CN111028152A (en) * | 2019-12-02 | 2020-04-17 | 哈尔滨工程大学 | Super-resolution reconstruction method of sonar image based on terrain matching |
CN111325694A (en) * | 2020-02-25 | 2020-06-23 | 深圳市景阳科技股份有限公司 | Image noise removing method and device |
CN112689099A (en) * | 2020-12-11 | 2021-04-20 | 北京邮电大学 | Double-image-free high-dynamic-range imaging method and device for double-lens camera |
CN116095517A (en) * | 2022-08-31 | 2023-05-09 | 荣耀终端有限公司 | Blurring method and blurring device |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104240203A (en) * | 2014-09-09 | 2014-12-24 | 浙江工业大学 | Medical ultrasound image denoising method based on wavelet transform and quick bilateral filtering |
CN105279746A (en) * | 2014-05-30 | 2016-01-27 | 西安电子科技大学 | Multi-exposure image integration method based on bilateral filtering |
CN106127718A (en) * | 2016-06-17 | 2016-11-16 | 中国人民解放军国防科学技术大学 | A kind of many exposure images fusion method based on wavelet transformation |
CN107220956A (en) * | 2017-04-18 | 2017-09-29 | 天津大学 | A kind of HDR image fusion method of the LDR image based on several with different exposures |
-
2018
- 2018-02-06 CN CN201810119668.6A patent/CN108492245B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105279746A (en) * | 2014-05-30 | 2016-01-27 | 西安电子科技大学 | Multi-exposure image integration method based on bilateral filtering |
CN104240203A (en) * | 2014-09-09 | 2014-12-24 | 浙江工业大学 | Medical ultrasound image denoising method based on wavelet transform and quick bilateral filtering |
CN106127718A (en) * | 2016-06-17 | 2016-11-16 | 中国人民解放军国防科学技术大学 | A kind of many exposure images fusion method based on wavelet transformation |
CN107220956A (en) * | 2017-04-18 | 2017-09-29 | 天津大学 | A kind of HDR image fusion method of the LDR image based on several with different exposures |
Non-Patent Citations (1)
Title |
---|
MARCELO BERTALMÍO,STACEY LEVINE: "Variational Approach for the Fusion of Exposure Bracketed Pairs", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110322409A (en) * | 2019-06-14 | 2019-10-11 | 浙江大学 | A kind of modified wavelet image fusion method based on label figure |
CN110322409B (en) * | 2019-06-14 | 2021-08-31 | 浙江大学 | Improved wavelet transform image fusion method based on labeled graph |
CN110428391A (en) * | 2019-08-02 | 2019-11-08 | 上海兆芯集成电路有限公司 | To remove the image interfusion method and device of ghost illusion |
CN110428391B (en) * | 2019-08-02 | 2022-05-03 | 格兰菲智能科技有限公司 | Image fusion method and device for removing ghost artifacts |
CN111028152A (en) * | 2019-12-02 | 2020-04-17 | 哈尔滨工程大学 | Super-resolution reconstruction method of sonar image based on terrain matching |
CN111028152B (en) * | 2019-12-02 | 2023-05-05 | 哈尔滨工程大学 | Super-resolution reconstruction method of sonar image based on terrain matching |
CN111325694A (en) * | 2020-02-25 | 2020-06-23 | 深圳市景阳科技股份有限公司 | Image noise removing method and device |
CN111325694B (en) * | 2020-02-25 | 2024-02-13 | 深圳市景阳科技股份有限公司 | Image noise removing method and device |
CN112689099A (en) * | 2020-12-11 | 2021-04-20 | 北京邮电大学 | Double-image-free high-dynamic-range imaging method and device for double-lens camera |
CN116095517A (en) * | 2022-08-31 | 2023-05-09 | 荣耀终端有限公司 | Blurring method and blurring device |
CN116095517B (en) * | 2022-08-31 | 2024-04-09 | 荣耀终端有限公司 | Blurring method, terminal device and readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN108492245B (en) | 2020-06-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108492245A (en) | Low light images based on wavelet decomposition and bilateral filtering are to fusion method | |
Zhang et al. | Kindling the darkness: A practical low-light image enhancer | |
Jian et al. | Multi-scale image fusion through rolling guidance filter | |
Yun et al. | Image enhancement using a fusion framework of histogram equalization and Laplacian pyramid | |
Lee et al. | A space-variant luminance map based color image enhancement | |
Li et al. | Image enhancement based on Retinex and lightness decomposition | |
Talebi et al. | Fast multilayer Laplacian enhancement | |
US8965141B2 (en) | Image filtering based on structural information | |
CN103606132A (en) | Multiframe digital image denoising method based on space domain and time domain combination filtering | |
CN110246088B (en) | Image brightness noise reduction method based on wavelet transformation and image noise reduction system thereof | |
CN112734650A (en) | Virtual multi-exposure fusion based uneven illumination image enhancement method | |
CN106664368A (en) | Image processing apparatus, image processing method, recording medium, and program | |
Mangiat et al. | Spatially adaptive filtering for registration artifact removal in HDR video | |
Zhang et al. | Self-supervised low light image enhancement and denoising | |
US9111339B1 (en) | System and method for reducing noise from an image | |
CN111260580A (en) | Image denoising method based on image pyramid, computer device and computer readable storage medium | |
Bao et al. | An edge-preserving filtering framework for visibility restoration | |
Dileep et al. | A comparison between different colour image contrast enhancement algorithms | |
Liu et al. | Signal dependent noise removal from a single image | |
Wei et al. | DA-DRN: A degradation-aware deep Retinex network for low-light image enhancement | |
CN110351453A (en) | A kind of computer video data processing method | |
Han et al. | Automatic illumination and color compensation using mean shift and sigma filter | |
Weibel et al. | Contrast-enhancing seam detection and blending using graph cuts | |
Yan et al. | Noise removal of MRI data with edge enhancing | |
CN110517210B (en) | Multi-exposure welding area image fusion method based on Haar wavelet gradient reconstruction |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200630 |
|
CF01 | Termination of patent right due to non-payment of annual fee |