CN107592537A - A kind of self-adapting compressing towards Aerial Images collection samples distribution method - Google Patents
A kind of self-adapting compressing towards Aerial Images collection samples distribution method Download PDFInfo
- Publication number
- CN107592537A CN107592537A CN201710985612.4A CN201710985612A CN107592537A CN 107592537 A CN107592537 A CN 107592537A CN 201710985612 A CN201710985612 A CN 201710985612A CN 107592537 A CN107592537 A CN 107592537A
- Authority
- CN
- China
- Prior art keywords
- image
- premeasuring
- image set
- psnr
- variance
- 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
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention relates to a kind of self-adapting compressing towards Aerial Images collection to sample distribution method, comprises the following steps:Premeasuring process:Concentrated in testing image and randomly select some width images as premeasuring image set, calculate the variance of each image in premeasuring image set, and optimum prediction amount parameter is calculated according to premeasuring image set composite quality;Image variance model is established for testing image collection;Sample rate is distributed;Sampling is compressed using gaussian random matrix.The present invention can more effectively lift the integrative reconstruction quality of image set by distributing corresponding sample rate for the different images in image set.
Description
Technical field
The present invention relates to technical field of image processing, is adopted more particularly to a kind of self-adapting compressing towards Aerial Images collection
Sample distributing method.
Background technology
With developing rapidly for new unmanned plane, imaging technique of taking photo by plane has obtained increasingly extensive concern.Based on unmanned plane
Aerial Images have the characteristics that object variations are fast, image detail is abundant, resource consumption is big.Traditional Aerial Images obtain system
Nyquist sampling law is typically based on, is sampled with the speed of preferably at least twice signal highest frequency, then passes through JPEG
Or the encoder such as JPEG2000 realizes effective compression storage.Recent studies have indicated that compressed sensing (CS) theoretical breakthrough how
The limitation of Qwest's Sampling Theorem, it still can effectively be recovered with owing the CS measured values of Nyquist rate sampling, substantially reduced
IMAQ cost, for the processing of Aerial Images provides new selection.The measurement end of compression perceptual system is fairly simple, weight
Structure end is then complex, and such characteristic meets the demand for imaging applications of taking photo by plane.
Image set is the set of one group of image, and the adaptive measuring mechanism of existing compression perceptual system mainly utilizes a width figure
As the spatial correlation of interior each piecemeal, and lack this correlation between different images, therefore, it is difficult to lift the whole of image set
Body reconstruction quality.When obtaining image set, existing system generally only uses presetting average sample rate (baseline methods)
Or the method (TCS methods) classified on a small quantity to image is shot to target image, the image set being achieved in that is usually
Because the quality gap of different images is larger, follow-up application demand can not be met.
The content of the invention
The technical problems to be solved by the invention are to provide a kind of self-adapting compressing towards Aerial Images collection and sample distribution
Method, to be that different images distribute suitable sample rate in image set under total sampling resource constraint.
The technical solution adopted for the present invention to solve the technical problems is:There is provided a kind of towards the adaptive of Aerial Images collection
Compression sampling distribution method, comprises the following steps:
(1) premeasuring process:Concentrated in testing image and randomly select some width images as premeasuring image set, calculated pre-
The variance of each image in image set is measured, and optimum prediction amount parameter is calculated according to premeasuring image set composite quality;
(2) image variance model is established for testing image collection, the distribution coefficient for obtaining current i-th width image is:Wherein, DiPresent sample image i variance, D are concentrated for testing imageminRepresent
The minimum variance of image, D in premeasuring image setmaxThe maximum variance of image in premeasuring image set is represented, β joins for premeasuring
Number;
(3) sample rate is distributed, and the sample rate of current i-th width image is assigned as η 'i=η * (α+δi* (1- α)), wherein, η is
Presetting average sample rate, α are the optimum value chosen after being sampled to premeasuring image set according to image composite quality;
(4) sampling is compressed using gaussian random matrix.
Image set composite quality in the step (1) is defined as follows:Ph=λ PSNRa+(1-λ)·PSNRm, in formula,
PSNRmRepresent the Y-PSNR of the worst image of reconstruction quality in image set, PSNRm=min PSNR (1) ..., PSNR
(i) ... }, PSNR (i) represents the Y-PSNR between the i-th width original image and its reconstructed image, PSNRaRepresent image set
Average image quality, Represent all images and its reconstructed image in image set
Between mean square error, n represent pixel number of bits, λ is weight coefficient.
The variance of each image in the premeasuring image setWherein, pj
The pixel of image is represented, N represents picture size;Premeasuring parameterWherein, K is premeasuring image set
The quantity of middle image.
Distribution coefficient δ in the step (2)iThe image i concentrated with testing image is one-to-one.
α is less than 0.8 in the step (3).
Beneficial effect
As a result of above-mentioned technical scheme, the present invention compared with prior art, has the following advantages that and actively imitated
Fruit:The present invention proposes the evaluation index of image set composite quality, is established by way of data-driven based on image variance
The compression sampling distribution method of model.Compared to conventional method, institute's extracting method for the different images in image set by distributing phase
The sample rate answered, the integrative reconstruction quality of image set can be more effectively lifted, there is certain application value.
Brief description of the drawings
Fig. 1 is the flow chart of the inventive method;
Fig. 2 is premeasuring procedure chart.
Embodiment
With reference to specific embodiment, the present invention is expanded on further.It should be understood that these embodiments are merely to illustrate the present invention
Rather than limitation the scope of the present invention.In addition, it is to be understood that after the content of the invention lectured has been read, people in the art
Member can make various changes or modifications to the present invention, and these equivalent form of values equally fall within the application appended claims and limited
Scope.
Embodiments of the present invention are related to a kind of self-adapting compressing towards Aerial Images collection and sample distribution method, such as Fig. 1
It is shown, comprise the following steps:Premeasuring process:Concentrated in testing image and randomly select some width images as premeasuring image
Collection, the variance of each image in premeasuring image set is calculated, and optimum prediction amount is calculated according to premeasuring image set composite quality
Parameter;Image variance model is established for testing image collection;Sample rate is distributed;Sampling is compressed using gaussian random matrix.Tool
Body is as follows:
In present embodiment, the average image quality PSNR of image setaIt is defined as follows:
Wherein n is the number of bits for representing pixel,Represent the mean square error between all images and its reconstructed image in image set
Difference.PSNR (i) represents the Y-PSNR between the i-th width original image and its reconstructed image.The worst picture quality of image set
PSNRmIt is defined as follows:PSNRm=min PSNR (1) ..., PSNR (i) ... }, the composite quality P of image sethIt is defined as follows:
Ph=λ PSNRa+(1-λ)·PSNRm.Present embodiment have chosen in NWPU-RESISC45 image data bases Bridge,
For tri- Aerial Images collection of Island and Forest as testing image collection, each image set is interior to be unified for 256 comprising 700 width sizes
× 256 image, present embodiment sample distribution using these three image sets to test the self-adapting compressing that the present invention is put forward
(VSA) performance of method, evaluation metricses are the composite quality P of reconstruct image image seth.The block size of compressed sensing choose performance compared with
For 32 × 32 pixels of equilibrium.As shown in Fig. 2 in testing image collection X1、X2、…、Xi、…、XIIn randomly select 10 width images work
For premeasuring image set, pass through variance calculation formulaPredicted
The variance D' of spirogram image set each imagei, minimum variance DminWith maximum variance Dmax, so as to calculate
Obtain premeasuring parameter beta corresponding with each image set.
The premeasuring parameter list of 1 three atlas of taking photo by plane of table
Then, image variance model is established for testing image collection, the distribution coefficient of the i-th width image is:Wherein, DiPresent sample image i variance is concentrated for testing image, can be by above-mentioned
Variance calculation formula obtains, and resulting in distribution coefficient δ corresponding with each image in the image seti。
Subsequent VSA methods carry out sample rate distribution, and η is presetting average sample rate, the sample rate of current i-th width image
For η 'i=η * (α+δi*(1-α)).This example is tested using 10 width images of premeasuring image set, by α respectively with 0.1,
0.2 ... 0.7 substitution institute extracting method is tested, and takes PhObtain α during maximum.Based on carried VSA methods, different alpha parameters pair
In the influence of image set composite quality, as shown in table 2 below~table 4.
The composite quality P of the Bridge image set difference alpha parameters of table 2h(dB)
The composite quality P of the Island image set difference alpha parameters of table 3h(dB)
The composite quality P of the Forest image set difference alpha parameters of table 4h(dB)
It can be drawn from above-mentioned table, optimal image can be obtained when α takes 0.1 in Bridge and Forest image sets
Collect composite quality, and for Island image sets, α obtains optimal image set composite quality when taking 0.7.Therefore, VSA methods
The model parameter selected for above-mentioned Aerial Images collection is as shown in table 5.
The model parameter α tables of 5 three atlas of taking photo by plane of table
Sample rate using gaussian random matrix is compressed sampling after being assigned.In compression perceptual system, measurement end
The distribution that proposed VSA methods, baseline methods, TCS methods carry out sample rate is respectively adopted, reconstruction end is used uniformly smooth orchid
Moral weber projection algorithm is rebuild to all images, and image set composite quality result is as shown in table 6.
The P of 6 different distribution methods of tableh(dB) contrast table
It is seen that general requirment of the present invention according to image set, employs the evaluation index of image set composite quality, uses
To calculate the integrative reconstruction quality of image set.This method establishes a figure according to the relative complexity of different images in image set
As Tobin's mean variance model, and based on the compression sampling distribution method of model execution image set.Test result indicates that compared to existing side
Method, in the case where identical samples resource constraint, this method effectively improves the integrative reconstruction quality of Aerial Images collection.
Claims (5)
1. a kind of self-adapting compressing towards Aerial Images collection samples distribution method, it is characterised in that comprises the following steps:
(1) premeasuring process:Concentrated in testing image and randomly select some width images as premeasuring image set, calculate premeasuring
The variance of each image in image set, and optimum prediction amount parameter is calculated according to premeasuring image set composite quality;
(2) image variance model is established for testing image collection, the distribution coefficient for obtaining current i-th width image is:Wherein, DiPresent sample image i variance, D are concentrated for testing imageminRepresent
The minimum variance of image, D in premeasuring image setmaxThe maximum variance of image in premeasuring image set is represented, β joins for premeasuring
Number;
(3) sample rate is distributed, and the sample rate of current i-th width image is assigned as η 'i=η * (α+δi* (1- α)), wherein, η is default
Fixed average sample rate, α are the optimum value chosen after being sampled to premeasuring image set according to image composite quality;
(4) sampling is compressed using gaussian random matrix.
2. the self-adapting compressing according to claim 1 towards Aerial Images collection samples distribution method, it is characterised in that institute
The image set composite quality stated in step (1) is defined as follows:Ph=λ PSNRa+(1-λ)·PSNRm, in formula, PSNRmRepresent
The Y-PSNR of the worst image of reconstruction quality, PSNR in image setm=min PSNR (1) ..., PSNR (i) ... }, PSNR
(i) Y-PSNR between the i-th width original image and its reconstructed image, PSNR are representedaRepresent the average image matter of image set
Amount, Represent the mean square error between all images and its reconstructed image in image set
Difference, n represent the number of bits of pixel, and λ is weight coefficient.
3. the self-adapting compressing according to claim 1 towards Aerial Images collection samples distribution method, it is characterised in that institute
State the variance of each image in premeasuring image setWherein, pjRepresent image
Pixel, N represent picture size;Premeasuring parameterWherein, K is the number of image in premeasuring image set
Amount.
4. the self-adapting compressing according to claim 1 towards Aerial Images collection samples distribution method, it is characterised in that institute
State the distribution coefficient δ in step (2)iThe image i concentrated with testing image is one-to-one.
5. the self-adapting compressing according to claim 1 towards Aerial Images collection samples distribution method, it is characterised in that institute
State α in step (3) and be less than 0.8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710985612.4A CN107592537B (en) | 2017-10-20 | 2017-10-20 | A kind of self-adapting compressing sampling distribution method towards Aerial Images collection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710985612.4A CN107592537B (en) | 2017-10-20 | 2017-10-20 | A kind of self-adapting compressing sampling distribution method towards Aerial Images collection |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107592537A true CN107592537A (en) | 2018-01-16 |
CN107592537B CN107592537B (en) | 2019-11-08 |
Family
ID=61052680
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710985612.4A Active CN107592537B (en) | 2017-10-20 | 2017-10-20 | A kind of self-adapting compressing sampling distribution method towards Aerial Images collection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107592537B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108805944A (en) * | 2018-05-29 | 2018-11-13 | 东华大学 | A kind of online image set compression method sorted out precision and kept |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110176743A1 (en) * | 2010-01-21 | 2011-07-21 | Sankar Pathamadi V | Data compression methods |
WO2013047335A1 (en) * | 2011-09-28 | 2013-04-04 | ソニー株式会社 | Image processing device and method |
CN104243987A (en) * | 2014-09-29 | 2014-12-24 | 刘鹏 | Self-adaptive sampling rate based image sampling method |
CN106385584A (en) * | 2016-09-28 | 2017-02-08 | 江苏亿通高科技股份有限公司 | Spatial correlation-based distributed video compressive sensing adaptive sampling and coding method |
CN106851283A (en) * | 2016-12-06 | 2017-06-13 | 广西大学 | A kind of method and device of the image adaptive compressed sensing sampling based on standard deviation |
CN106941609A (en) * | 2017-02-15 | 2017-07-11 | 浙江工业大学 | The video-frequency compression method perceived based on self adaptation splits' positions |
CN107146260A (en) * | 2017-04-14 | 2017-09-08 | 电子科技大学 | A kind of compression of images based on mean square error perceives the method for sampling |
CN107211128A (en) * | 2015-03-10 | 2017-09-26 | 苹果公司 | Adaptive chroma down-sampling and color space switch technology |
-
2017
- 2017-10-20 CN CN201710985612.4A patent/CN107592537B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110176743A1 (en) * | 2010-01-21 | 2011-07-21 | Sankar Pathamadi V | Data compression methods |
WO2013047335A1 (en) * | 2011-09-28 | 2013-04-04 | ソニー株式会社 | Image processing device and method |
CN104243987A (en) * | 2014-09-29 | 2014-12-24 | 刘鹏 | Self-adaptive sampling rate based image sampling method |
CN107211128A (en) * | 2015-03-10 | 2017-09-26 | 苹果公司 | Adaptive chroma down-sampling and color space switch technology |
CN106385584A (en) * | 2016-09-28 | 2017-02-08 | 江苏亿通高科技股份有限公司 | Spatial correlation-based distributed video compressive sensing adaptive sampling and coding method |
CN106851283A (en) * | 2016-12-06 | 2017-06-13 | 广西大学 | A kind of method and device of the image adaptive compressed sensing sampling based on standard deviation |
CN106941609A (en) * | 2017-02-15 | 2017-07-11 | 浙江工业大学 | The video-frequency compression method perceived based on self adaptation splits' positions |
CN107146260A (en) * | 2017-04-14 | 2017-09-08 | 电子科技大学 | A kind of compression of images based on mean square error perceives the method for sampling |
Non-Patent Citations (3)
Title |
---|
刘浩等: "Edge-aware spatial-frequency extrapolation for consecutive block loss", 《SPRINGERPLUS》 * |
刘浩等: "Entropy-aware projected Landweber reconstruction for quantized block compressive sensing of aerial imagery", 《JOURNAL OF APPLIED REMOTE SENSING》 * |
曹玉强等: "图像自适应分块的压缩感知采样算法", 《中国图象图形学报》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108805944A (en) * | 2018-05-29 | 2018-11-13 | 东华大学 | A kind of online image set compression method sorted out precision and kept |
CN108805944B (en) * | 2018-05-29 | 2022-05-06 | 东华大学 | Online image set compression method with maintained classification precision |
Also Published As
Publication number | Publication date |
---|---|
CN107592537B (en) | 2019-11-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106910161B (en) | Single image super-resolution reconstruction method based on deep convolutional neural network | |
WO2020108358A1 (en) | Image inpainting method and apparatus, computer device, and storage medium | |
CN107909556B (en) | Video image rain removing method based on convolutional neural network | |
CN108632625A (en) | A kind of method for video coding, video encoding/decoding method and relevant device | |
CN110717868B (en) | Video high dynamic range inverse tone mapping model construction and mapping method and device | |
CN108805839A (en) | Combined estimator image defogging method based on convolutional neural networks | |
CN103927734B (en) | A kind of based on the broad image quality evaluating method without reference | |
CN109120937A (en) | A kind of method for video coding, coding/decoding method, device and electronic equipment | |
CN112116601A (en) | Compressive sensing sampling reconstruction method and system based on linear sampling network and generation countermeasure residual error network | |
CN109922346A (en) | A kind of convolutional neural networks for the reconstruct of compressed sensing picture signal | |
Yang et al. | Optimized-SSIM based quantization in optical remote sensing image compression | |
CN103761724A (en) | Visible light and infrared video fusion method based on surreal luminance contrast pass algorithm | |
CN114998145A (en) | Low-illumination image enhancement method based on multi-scale and context learning network | |
CN115052187B (en) | Super-resolution live broadcast system based on online training | |
CN110910347A (en) | Image segmentation-based tone mapping image no-reference quality evaluation method | |
CN107592537A (en) | A kind of self-adapting compressing towards Aerial Images collection samples distribution method | |
CN112150356A (en) | Single compressed image super-resolution reconstruction method based on cascade framework | |
CN106454386B (en) | A kind of method and apparatus of the Video coding based on JND | |
CN111062926B (en) | Video data processing method, device and storage medium | |
CN113112557A (en) | Block self-adaptive carton image compression method based on dictionary learning | |
CN108573509A (en) | Self-adapting compressing perceptual image coding method based on the notable detection model of measurement field | |
CN116468625A (en) | Single image defogging method and system based on pyramid efficient channel attention mechanism | |
CN103475875A (en) | Image adaptive measuring method based on compressed sensing | |
CN115147311A (en) | Image enhancement method based on HSV and AM-RetinexNet | |
CN111127587B (en) | Reference-free image quality map generation method based on countermeasure generation network |
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 |