CN109934801A - A kind of image Focus field emission array implementation method based on piecemeal Hadamard transform - Google Patents
A kind of image Focus field emission array implementation method based on piecemeal Hadamard transform Download PDFInfo
- Publication number
- CN109934801A CN109934801A CN201910101960.XA CN201910101960A CN109934801A CN 109934801 A CN109934801 A CN 109934801A CN 201910101960 A CN201910101960 A CN 201910101960A CN 109934801 A CN109934801 A CN 109934801A
- Authority
- CN
- China
- Prior art keywords
- image
- hadamard transform
- field emission
- subgraph
- emission array
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The image Focus field emission array implementation method based on piecemeal Hadamard transform that the invention discloses a kind of, belongs to imaging and passive imaging technical field.This method carries out noise reduction process to image first, then carries out piecemeal as desired to the image after noise reduction and calculates the Hadamard transform of each width subgraph, obtains the frequency coefficient of corresponding subgraph.Choose the low frequency coefficient of each width subgraph again to construct the characteristic information of image.The variance of all constituted set of subgraph characteristic information is finally sought, and using the variance yields as the Focus field emission array of entire image.The present invention is by the way of image block and Hadamard transform extracts image detail information, have the advantages that principle is simple, computation complexity is low, implementation through the above steps simultaneously, the interference of noise on image detailed information is reduced, the especially noise robustness under low contrast image-forming condition is strong.This method is suitable for the passive imaging system of camera, convenient for promoting the use of.
Description
Technical field
The invention belongs to imaging and passive imaging technical fields, and in particular to a kind of image focusing survey based on piecemeal Hadamard transform
Spend implementation method.
Background technique
Photographing device in daily life, such as the monitoring of slr camera, the mobile phone with camera function and crossing are grabbed
Shooting system etc. can obtain clearly image.However the acquisition of clear image is to rely on the automatic focusing performance of photographing device
It realizes.Currently, photographing device in the market mainly uses the Techniques of Automatic Focusing of imaging and passive imaging.Its core is one use of design
In the Focus field emission array of evaluation image definition, clearest image and preservation are selected by Focus field emission array.Therefore, a performance
Excellent Focus field emission array implementation method directly influences the quality of captured image.
Be using relatively broad image Focus field emission array method at present constructed based on image detail information, such as based on
The Focus field emission array of Edge extraction.Typical method has image single order Gaussian derivative method, Second Derivative Methods, single order local derviation
Counting method, gradient summation method and Laplce's summation method etc..The essence of such method is construction one having a size of 3 × 3
Or 5 × 5 convolution masks, convolution algorithm is carried out with the template and entire image.Convolution algorithm the result is that extracting image
Marginal information, then using take absolute value or square summation in the form of construct the Focus field emission array of entire image.Such construct
There are following two major defects for the method for Focus field emission array.It is that convolution algorithm complexity is high first, needs the institute to entire image
There is pixel to carry out traversing operation, there is presently no the fast algorithm on more mature fast algorithm, especially hardware device,
So that the focusing real-time index of such Focus field emission array method is poor.The followed by noise of image and marginal information belongs to high frequency
Information can enhance noise information after convolution algorithm, that is to say, that such Focus field emission array eventually leads to mistake vulnerable to influence of noise
Focusing accidentally.
In addition a kind of Focus field emission array method is the method based on image transformation, i.e., the height of image is extracted at transform domain (frequency domain)
Frequency information constructs Focus field emission array.The ratio of typical method has a thin multi-scale wavelet coefficient and high frequency and low-frequency wavelet coefficients,
Focus field emission array based on discrete cosine transform, the Focus field emission array based on Fourier transformation and Short-Time Fractional Fourier Transform
Focus field emission array.These methods based on transformation have the characteristics that one it is common, extract high-frequency information after exactly converting to image,
With this as image Focus field emission array value.Such methods are consistent in the way of thinking based on the method for edge extracting with front
, emphasize high-frequency information.Only the former is Focus field emission array to be constructed using the method for convolution in airspace, and the latter is in frequency
Domain constructs Focus field emission array by the way of transformation.Based on image transformation method be also easy it is affected by noise, and some become
The computation complexity changed is bigger, such as wavelet transformation and Fourier Transform of Fractional Order, unmature hardware fast algorithm.
Above-mentioned two classes method is to construct Focus field emission array with the global information of entire image there are one common feature.Such as
The background of fruit image is relatively uniform or smoother, easily affected by noise at this time, so that corresponding Focus field emission array can not be anti-
Mirror the sharpness information of image.Such as with the camera function of mobile phone under the weaker indoor or night scenes of illumination condition
It takes pictures, we can have found that the automatic focusing function of (mobile phone) camera is not handy, and captured image out exists fuzzy existing
As and have granular sensation.Here it is an embodiments of Focus field emission array algorithm failure.Therefore, how to construct with noise robustness
Focus field emission array has important research significance and practical value.
Summary of the invention
Aiming at the problems existing in the prior art, the present invention provides a kind of, and the image with very noisy robustness focuses survey
Implementation method is spent, the technical solution used in the present invention is as follows:
A kind of image Focus field emission array implementation method based on piecemeal Hadamard transform, includes the following steps:
Step S1: the line number of original image pixels and columns are adjusted to 2nIntegral multiple, wherein n be positive integer, obtain
To image f (x, y), the line number and columns of image f (x, y) is indicated with M and N respectively；
Step S2: noise reduction process is carried out to image f (x, y), obtains image g (x, y)；
Step S3: carrying out piecemeal processing to image g (x, y), and obtaining size is 2n×2nThe subgraph S of pixeli(x, y),
Middle i=1,2 ..., M × N/22n；Subgraph SiThe variable-value of (x, y) are as follows: x=0,1 ..., 2n- 1, y=0,1 ..., 2n-
1；
Step S4: each width subgraph S is calculatediThe Hadamard transform of (x, y) obtains the frequency coefficient T of corresponding subgraphi
(u, v), wherein u=0,1 ..., 2n- 1, v=0,1 ..., 2n-1；
Step S5: each width subgraph frequency coefficient T is choseni(0,0), Ti(0,1) and Ti(1,0) constructs the spy of image
Reference ceases and is labeled as Fi, its calculation formula is: Fi=0.8Ti(0,0)+0.1Ti(0,1)+0.1Ti(1,0)；
Step S6: by all subgraph characteristic information FiThe set constituted is denoted as { Fi| i=1,2 ..., M × N/22n,
The variance of the set is sought, and using variance yields as the Focus field emission array value of entire image.
Preferably, noise reduction formula in the step S2 are as follows: g (x, y)=f (x, y) * ST；
Wherein ST is 4 × 4 symmetrical Filtering Templates, specifically:
Preferably, the Hadamard transform formula in the step S4 are as follows:
Wherein K (x, y, u, v) is the kernel function of Hadamard transform.
Preferably, Hadamard transform for ease of calculation, especially facilitates and realizes Hadamard transform on hardware circuit, adopt
The transformation, specific formula are realized with matrix multiplication are as follows:
WhereinIt is 2nRank hadamard matrix is calculated by recursive fashion, is comprised the concrete steps that by 2 rank hada
Ma matrixIt sets out, is calculated using recursive fashionRecurrence formula is as follows:
Compared with prior art, the invention has the advantages that: the principle of the invention is simple, using the side of image block
Formula and Hadamard transform extract image detail information, have the advantages that computation complexity is low, while passing through step S2~S6's
Implement, largely reduce the interference of noise on image detailed information so that the obtained Focus field emission array of this method have compared with
High noise robustness, the noise resisting ability being suitable under the passive imaging system of camera, especially low contrast image-forming condition
It is relatively strong, it is suitable for promoting the use of.
Detailed description of the invention
Fig. 1 is implementation steps block diagram of the invention.
Specific embodiment
Technical solution of the present invention is understood for the ease of technical staff, now in conjunction with Figure of description and embodiment to the present invention
Technical solution be described in further detail.
The invention proposes a kind of image Focus field emission array implementation method based on piecemeal Hadamard transform, implementation step frame
Figure is as shown in Figure 1, in the present embodiment, select n=3, then the specific steps of this method refine are as follows:
The line number of original image pixels and columns are adjusted to 8 integral multiple by step S1, can be by cutting out to image
It cuts or interpolation is realized, just obtain image f (x, y) in this way.Here, the line number of f (x, y) and columns are indicated with M and N respectively.
Why picturedeep and columns are adjusted be because method proposed by the invention be to be realized based on image block, when
When n=3, need image to be divided into several width subgraphs that size is 8 × 8 pixels.
In order to reduce the influence of picture noise (mainly Gaussian noise, salt-pepper noise and multiplying property impact noise), next
Step S2~S6 play key effect.
Step S2 carries out noise reduction process to the image f (x, y) of previous step, obtains image g (x, y).Specific formula for calculation
Are as follows: g (x, y)=f (x, y) * ST.Wherein ST is 4 × 4 symmetrical Filtering Templates, is specifically defined are as follows:
In fact, due to the symmetry characteristic of Filtering Template ST and it in value part " mean value " effect, with filtering
Template ST carries out handling to be equivalent to being extracted the intermediate frequency information of image to image f (x, y), and has filtered out most of high-frequency noise letter
Breath.So it lays a good foundation for the quantification treatment of subsequent step.
Preliminary anti-noise sonication is realized by step S2.
Step S3 carries out piecemeal processing to the obtained image g (x, y) of step S2, and obtaining several width sizes is 8 × 8
Subgraph Si(x, y), wherein i=1,2 ..., M × N/64.The size for paying attention to subgraph is 8 × 8, therefore subgraph Si(x, y)
Variable-value are as follows: x=0,1 ..., 7, y=0,1 ..., 7.Why carrying out piecemeal processing to image is calculated for reduction
The considerations of complexity.This in traditional airspace filter calculating process using pixel-by-pixel point handled by the way of compared with, significantly
Reduce computation complexity.In addition, based on the calculation method of piecemeal, there are also the effects of smothing filtering, can further decrease noise
Influence to image definition quantized result.
Step S4 calculates each width subgraph SiThe Hadamard transform of (x, y) obtains the frequency coefficient T of corresponding subgraphi
(u, v), wherein u=0,1 ..., 7, v=0,1 ..., 7.
Here, Hadamard transform formula becomes:
Wherein K (x, y, u, v) is the kernel function of Hadamard transform, is related in original definition to variable x, y, u and v
Value carries out Binary Conversion, and carries out multiplying to corresponding binary bit, this causes its computation complexity higher.To understand
Certainly this problem, the present invention realize Hadamard transform, specific formula using matrix multiplication are as follows:
Wherein H8It for 8 rank hadamard matrixs, is calculated by recursive fashion, comprised the concrete steps that by 2 rank hada
Ma matrixIt sets out, H is calculated using recursive fashion8, recurrence formula is as follows:
Calculate H8:
The present invention constructs Focus field emission array using Hadamard transform, and to be primarily due to its computation complexity very low.From 8 rank hada
Ma matrix H8Definition can see, the value of the matrix element only has " 1 " and " -1 " two values, it means that its matrix multiplication
Operation finally all becomes subgraph Si" addition " and " subtraction " operation of (x, y) itself pixel value, greatly reduces calculating
Complexity, while being easy to the realization of hardware circuit, this is vital for portable photographing system.
Step S5 chooses each width subgraph frequency coefficient Ti(0,0), Ti(0,1) and Ti(1,0) constructs the subgraph
Characteristic information and be labeled as Fi, its calculation formula is: Fi=0.8Ti(0,0)+0.1Ti(0,1)+0.1Ti(1,0).Why select
Select Ti(0,0), Ti(0,1) and Ti(1,0) these three frequency domain components are because Hadamard transform has energy accumulating characteristic, also
It is that the energy of image all concentrates on a small number of several low frequency coefficients.In fact, FiThe middle low-frequency information of image is represented, this
Information in a frequency range is most to be not readily susceptible to influence of noise.
Step S6, all subgraph characteristic information FiThe set constituted is denoted as { Fi| i=1,2 ..., M × N/64 }, it asks
The variance of the set, and using variance yields as the Focus field emission array value of entire image.Good image is focused for a width, is wrapped
The detailed information contained is more, specifically has marginal information or zone boundary information, and these detailed information are both present in image
Regional area in.And the extraction of this partial information is most important for the calculating of Focus field emission array.This is also that the present invention implements step
Another reason of piecemeal operation is carried out in rapid S3 to image g (x, y).In fact, image is more clear, the brightness change of image
It is more obvious, is exactly that the pixel value of clear image has biggish dispersion from the viewpoint of image pixel value.In statistics
It is upper to measure this discrete feature usually using variance.Therefore, the present invention passes through in step s 6 calculates all subgraph features
Information FiVariance obtain Focus field emission array.
The variance yields that step S6 is calculated means that more greatly sharpness information included in each width subgraph
Contrast it is bigger, i.e., the detailed information for including in image is more.
Apply in example at other, can the pixel value according to image to be processed and the real-time demand to focus method etc. because
Element selection n value, n value is bigger, and piecemeal quantity is fewer, and the real-time of method is better, but focusing effect is poorer；Conversely, n value is got over
Small, piecemeal quantity is more, and focusing effect is better, but the real-time of method is deteriorated, so the n value of one compromise of selection is very heavy
It wants.
It should be noted that above-described embodiment can be freely combined as needed.The above is only of the invention preferred
Embodiment, it is noted that for those skilled in the art, in the premise for not departing from the principle of the invention
Under, several improvements and modifications can also be made, these modifications and embellishments should also be considered as the scope of protection of the present invention.
Claims (4)
1. a kind of image Focus field emission array implementation method based on piecemeal Hadamard transform, it is characterised in that: specific step is as follows:
Step S1: the line number of original image pixels and columns are adjusted to 2nIntegral multiple, wherein n be positive integer, obtain image
The line number and columns of f (x, y), image f (x, y) are indicated with M and N respectively；
Step S2: noise reduction process is carried out to image f (x, y), obtains image g (x, y)；
Step S3: carrying out piecemeal processing to image g (x, y), and obtaining size is 2n×2nThe subgraph S of pixeli(x, y), wherein i
=1,2 ..., M × N/22n；Subgraph SiThe variable-value of (x, y) are as follows: x=0,1 ..., 2n- 1, y=0,1 ..., 2n-1；
Step S4: each width subgraph S is calculatediThe Hadamard transform of (x, y) obtains the frequency coefficient T of corresponding subgraphi(u,
V), wherein u=0,1 ..., 2n- 1, v=0,1 ..., 2n-1；
Step S5: each width subgraph frequency coefficient T is choseni(0,0), Ti(0,1) and Ti(1,0) believes to construct the feature of image
It ceases and is labeled as Fi, its calculation formula is: Fi=0.8Ti(0,0)+0.1Ti(0,1)+0.1Ti(1,0)；
Step S6: by all subgraph characteristic information FiThe set constituted is denoted as { Fi| i=1,2 ..., M × N/22n, ask this
The variance of set, and using variance yields as the Focus field emission array value of entire image.
2. the image Focus field emission array implementation method based on piecemeal Hadamard transform as described in claim 1, it is characterised in that: institute
Stating noise reduction formula in step S2 are as follows: g (x, y)=f (x, y) * ST, wherein ST is 4 × 4 symmetrical Filtering Templates, specifically:
3. the image Focus field emission array implementation method based on piecemeal Hadamard transform as described in claim 1, it is characterised in that: institute
State the Hadamard transform formula in step S4 are as follows:
Wherein K (x, y, u, v) is the kernel function of Hadamard transform.
4. the image Focus field emission array implementation method as claimed in claim 1 or 3 based on piecemeal Hadamard transform, feature exist
It is realized in: the Hadamard transform using matrix multiplication, specific formula are as follows:
WhereinIt is 2nRank hadamard matrix is calculated by recursive fashion, is comprised the concrete steps that by 2 rank hadamard matrixsIt sets out, is calculated using recursive fashionRecurrence formula is as follows:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910101960.XA CN109934801A (en) | 2019-01-25 | 2019-01-25 | A kind of image Focus field emission array implementation method based on piecemeal Hadamard transform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910101960.XA CN109934801A (en) | 2019-01-25 | 2019-01-25 | A kind of image Focus field emission array implementation method based on piecemeal Hadamard transform |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109934801A true CN109934801A (en) | 2019-06-25 |
Family
ID=66985467
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910101960.XA Pending CN109934801A (en) | 2019-01-25 | 2019-01-25 | A kind of image Focus field emission array implementation method based on piecemeal Hadamard transform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109934801A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1908801A (en) * | 2006-08-22 | 2007-02-07 | 宁波大学 | Hadamard transformation based digital micro imaging automatic focusing method |
| CN105354817A (en) * | 2015-09-25 | 2016-02-24 | 济南中维世纪科技有限公司 | Noise image automatic focusing method |
| CN105551029A (en) * | 2015-12-09 | 2016-05-04 | 云南大学 | Multi-spectral remote sensing image-based ship detection method |
-
2019
- 2019-01-25 CN CN201910101960.XA patent/CN109934801A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1908801A (en) * | 2006-08-22 | 2007-02-07 | 宁波大学 | Hadamard transformation based digital micro imaging automatic focusing method |
| CN105354817A (en) * | 2015-09-25 | 2016-02-24 | 济南中维世纪科技有限公司 | Noise image automatic focusing method |
| CN105551029A (en) * | 2015-12-09 | 2016-05-04 | 云南大学 | Multi-spectral remote sensing image-based ship detection method |
Non-Patent Citations (4)
| Title |
|---|
| GUO L ET AL.: "Focus Measure Based on the Image Moments", 《IEEE XPLORE》 * |
| SANG-YONG LEE ET AL.: "Enhanced Autofocus Algorithm Using Robust Focus Measure and Fuzzy Reasoning", 《IEEE TRANSACTIONS ON CIRCUITS & SYSTEMS FOR VIDEO》 * |
| 何东健等: "《数字图像处理》", 31 July 2003, 西安电子科技大学出版社 * |
| 张兆礼等: "《现代图像处理技术及Matlab实现》", 30 November 2001, 人民邮电出版社 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9633422B2 (en) | Method for image processing using local statistics convolution | |
| Zhang et al. | PCA-based spatially adaptive denoising of CFA images for single-sensor digital cameras | |
| Yang et al. | Textured image demoiréing via signal decomposition and guided filtering | |
| Cui et al. | Color image demosaicking using a 3-stage convolutional neural network structure | |
| CN111275655B (en) | Multi-focus multi-source image fusion method | |
| CN109859196A (en) | A kind of image Focus field emission array implementation method based on Modular PCA | |
| CN103997592B (en) | Vedio noise reduction method and system | |
| Ren et al. | Research of multi-focus image fusion based on M-band multi-wavelet transformation | |
| Akbari et al. | Adaptive saliency-based compressive sensing image reconstruction | |
| Junwu et al. | An infrared and visible image fusion algorithm based on LSWT-NSST | |
| CN109801288A (en) | A kind of image Focus field emission array implementation method based on directional statistics characteristic | |
| CN104732480B (en) | A kind of remote sensing images ultra-resolution method based on non local regularization model | |
| Zhu et al. | Multiscale channel attention network for infrared and visible image fusion | |
| CN109801289A (en) | A kind of image Focus field emission array implementation method based on Gabor transformation | |
| Tang et al. | Curvature-based real-time brightness adjustment for ultra hd video | |
| CN109785323A (en) | A kind of image Focus field emission array implementation method based on intermediate frequency filtering | |
| CN109886943A (en) | A kind of image Focus field emission array implementation method based on number theoretic transform | |
| CN109934801A (en) | A kind of image Focus field emission array implementation method based on piecemeal Hadamard transform | |
| CN109859194A (en) | A kind of image Focus field emission array implementation method based on Local Edge Detection | |
| CN109934876A (en) | A kind of image Focus field emission array implementation method based on second order moment function | |
| Kannan et al. | Optimal decomposition level of discrete wavelet transform for pixel based fusion of multi-focused images | |
| CN109886942A (en) | A kind of Focus field emission array implementation method based on image kurtosis | |
| Ren et al. | Illumination-invariant non-local means based video denoising | |
| Shamna | Satellite image resolution and brightness enhancement using discrete, stationary wavelet and singular value decomposition | |
| CN109859195A (en) | A kind of image Focus field emission array implementation method based on local phase feature |
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 |