CN103871387B - A kind of image sampling method based on liquid crystal display optical modulation - Google Patents

Abstract

The invention provides the compressive sampling method that a kind of liquid crystal display carries out optical modulation, element used comprises: liquid crystal display, liquid crystal display Drive and Control Circuit, fourier lense, photodiode.The method comprises the following steps: the light that object being measured sends is imaged on liquid crystal display through optical imagery parts.Each pixel of liquid crystal display is equivalent to a photoswitch, and liquid crystal display driving circuit controls the torsion of liquid crystal, to change its transmittance, makes light selectively through liquid crystal display.Changed by the transmittance of liquid crystal display, physically realize a programmable observing matrix.Light through liquid crystal display injects fourier lense.Photodiode accepts light intensity at lens focus place, and carries out integration to light intensity, and the result obtained is equivalent to the observed reading in compressed sensing theory.Structure of the present invention is simple, cost is lower, has good use value and wide application prospect at optical imagery and image acquisition field.

Description

A kind of image sampling method based on liquid crystal display optical modulation

(1) technical field

The present invention relates to technical field compressed sensing theory being applied to image acquisition, particularly a kind of using the new type of compression method of sampling of liquid crystal display as the Physical realization of observing matrix.

(2) background technology

Traditional signals collecting is based on nyquist sampling theorem, and when obtaining signal, sample frequency must be greater than the twice of highest frequency in signal, could Accurate Reconstruction signal.But along with developing rapidly of science and technology, high-resolution digital equipment sampling creates huge data, also saving storage and transmission cost are a great problems to greatest extent how more efficiently to process these data, propose requirements at the higher level to signal handling capacity, bring great challenge also to corresponding hardware device.In fact, most of data that traditional sampling obtains are unessential, and in the processing procedure of signal or image, only retain the data that some is important, give up a large amount of redundant datas, the signal after reconstruct or image can't cause visual difference.Because the data major part collected is all unessential, can be dropped, so can directly gather part important, the data that are not finally dropped, and can accurately reconstruct original signal or image, this is the thought of compressed sensing theory.

Compressed sensing theory was formally proposed in 2006 on the basis of correlative study by Cand é s and Donoho.Its core concept is carried out compression and sampling merging.First the non-self-adapting linear projection (measured value) of collection signal, then reconstructs original signal according to corresponding restructing algorithm by measured value.The advantage of compressed sensing is the data volume that the projection measurements amount of signal is far smaller than traditional sampling method and obtains, and can suppress random noise.The first compression sampling of this principle, measured signal being mapped to low-dimensional by higher-dimension and sample to it, choose suitable rarefaction representation base Ψ, it is sparse for making original signal f convert gained vector x through Ψ.Then according to observation data y, observing matrix Φ and rarefaction representation base Ψ, choose special algorithm and solve y=Φ Ψ x, be finally finally inversed by original signal f by f=Ψ x.

Since compressed sensing proposes, scholars have carried out applied research widely in optical imaging field to it.As grain graininess measurement, single pixel camera, ultra-thin imaging system, frequency multiplexing technique intelligence imaging, multispectral imaging, the imaging of CMOS low data rate, Magnetic resonance imaging, astronomical sight etc.Such as, patent (numbering: 201210058483.1) " photon counting compression sampling phased-array laser three-D imaging method " modulates laser illuminator with liquid crystal optical phased array, the calculation matrix used according to compression sampling irradiates target, the photon number returned is converted in time, finally by compression sampling recovery algorithms reconstruction of three-dimensional images by geiger mode avalanche photodiodes (APD) receiving target echoed signal and Single Photon Counting device.Relate to the modulation that calculation matrix only realizes the signal to transmitting terminal in above-mentioned patent, do not consider the compressed sensing implementation of receiving end signal.

Liquid crystal display with liquid crystalmaterial is basic module, filling liquid crystal material between two pieces of parallel plates, is changed the arrangement situation of liquid crystal material interior molecules by voltage, to reach the object of shading and printing opacity, shows deep mixed, in picturesque disorder image.As long as add the filter layer of three-primary colours between two pieces of flat boards, just color display can be realized.Liquid crystal display power consumption is very low, therefore extremely slip-stick artist's favor, is applicable to the electronic equipment using battery.

The present invention adopts the liquid crystal display removing the adjustable gray scale of monochrome of backlight.Each pixel of liquid crystal display is equivalent to a photoswitch, and liquid crystal display driving circuit controls the torsion of liquid crystal, to change its transmittance, makes light selectively through liquid crystal display.Changed by the transmittance of liquid crystal display, physically realize a programmable observing matrix.

(3) summary of the invention

How by compressed sensing Successful utilization in optical imagery, key is the compression sampling how correctly obtaining image.From each formation method, except cmos imaging, all the other have all related to coding mode.Coding mode great majority are all realize by optical system, once mask fabrication completes, its coding mode is also just fixing, cannot change.Picture digital micromirror array (DigitalMirrorDevice, DMD) this programmable coding implementation, although flexibly and easily, is applied to the complicacy that imaging system can increase light path, influential system reliability.And DMD can only realize binary observing matrix.

The present invention proposes a kind of new compression sampling mode, adopts liquid crystal display as the Physical realization of observing matrix Φ.

Element used comprises: liquid crystal display, liquid crystal display Drive and Control Circuit, fourier lense, photodiode.

The technical solution used in the present invention is: the light that object being measured sends is imaged on liquid crystal display through optical imagery parts.Each pixel of liquid crystal display is equivalent to a photoswitch, and liquid crystal display driving circuit controls the torsion of liquid crystal, to change its transmittance, makes light selectively through liquid crystal display.Changed by the transmittance of liquid crystal display, physically realize a programmable observing matrix.Light through liquid crystal display injects fourier lense.Photodiode is at lens focus place receiving light power, and by light intensity integration, the result obtained is equivalent to the observed reading y in compressed sensing theory.

The invention has the advantages that: the programmability with similar DMD element.And, be through liquid crystal display due to light instead of enter lens after DMD element reflects, therefore can each for system element be positioned on an optical axis, directly fourier lense can be positioned over the liquid crystal display back side.The selection of such focal length of lens, just no longer by the restriction of light reflection angle, reduces the complexity of light path.Meanwhile, the shortcoming that DMD can only realize scale-of-two observing matrix is overcome.And the price of liquid crystal display is comparatively cheap, reduce the cost of imaging system.

(4) accompanying drawing explanation

Below in conjunction with the drawings and specific embodiments, the present invention will be further described.

Fig. 1 is structural representation of the present invention.

Fig. 2 is the simulated experiment result of the embodiment of the present invention.

Accompanying drawing indicates

1, liquid crystal display 2, fourier lense 3, photodiode

(5) embodiment

In FIG, liquid crystal display (1), fourier lense (2), photodiode (3) are positioned on an optical axis, fourier lense (2) is positioned between liquid crystal display (1) and photodiode (3), and photodiode (3) is positioned at the focus place of fourier lense (2).

Step 1: the light that object being measured sends is imaged on liquid crystal display through optical imagery parts.Picture now on liquid crystal display is original signal X _{n × 1}.

Step 2: liquid crystal display driving circuit controls the torsion of liquid crystal, the pixel on liquid crystal display is arranged and is in pseudo-random state i, wherein 1≤i≤M, their state constitutes the i-th row h of observing matrix Φ _i(size is N).Will be now that X is at h through the signal of liquid crystal display _iunder value.

Step 3: the light through liquid crystal display injects fourier lense.Photodiode, at lens focus place receiving light power, by light intensity integration, obtains result y _i=h _ix _{n × 1}.

Step 4: repeat step above M time, then M liquid crystal screen image element ordered state constitutes observing matrix Φ _{m × N}, M time measurement result forms observation data matrix Y=Φ X.

The mathematical model of compression sampling part is as follows:

One-dimensional discrete signal x can be expressed as

$x=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{s}_i{\mathrm{\Ψ}}_i$ Or x=ψ s (1)

Wherein Ψ=[ψ ₁| ψ ₂| ψ _n], s is weighting coefficient s _i=<x, ψ _i>=ψ _i ^tn × 1 column vector of x.X and s is the equivalent representation of signal.If x is only the linear combination (K<N) of K base vector, then signal x is claimed to be that K is sparse.As K ∈ N, above-mentioned formula only has the large coefficient of minority and a large amount of little coefficients, and now signal x becomes compressible.

Consider a linear measurement process, calculate M x(M<N) and vector set inner product, that is: y _i=<x, φ _j>.If measured value y _im × 1 vector formed is y, by m × N the matrix formed as vector is Φ, then have

y＝Φx＝ΦΨ＝Θs(2)

Wherein Θ=Φ Ψ is M × N matrix.Matrix Θ will meet constraint equidistant characteristics (restrictedisometryproperty (RIP))

(1-ε)||v|| ₂≤||Θv|| ₂≤(1+ε)||v|| ₂(3)

Namely matrix Θ must keep the length of specific K sparse spike.Because Ψ is fixing, Θ=Φ Ψ be made to meet constraint equidistant characteristics, can be solved by design observing matrix Φ.The people such as Cand é s demonstrate when Φ is gaussian random matrix, and matrix Θ can meet constraint equidistant characteristics with greater probability.Gauss's observing matrix that can be therefore M × N by selection size obtains.

Fig. 2 is the simulated experiment result of the embodiment of the present invention.Fig. 2 (a) is original image, and size is 100 × 100, is gray level image.The energy of each frequency range of this pictures is all very abundant, appropriateness be mixed with details, smooth region, shade and texture, thus can well test various image processing algorithm, be the standardized test chart of most widespread use in Research on Image Compression Algorithm.Fig. 2 (a) is equivalent to the original signal X imaged in step 1 on liquid crystal display _{10000 × 1}.Observing matrix Φ chooses the scale-of-two stochastic matrix that size is M × 10000.Such Y=Φ X just constitutes the observation data matrix obtained after M sampling in step 4.In the simulated experiment of the present embodiment, choose minimum full variational method solving equation Y=Φ X.The result X ' gone out through this Algorithm for Solving forms reconstruction image.The image of Fig. 2 (b) for reconstructing when sampling number M gets 4000.

The minimum full variational method is Cand é s etc. is all sparse angle from the discrete gradient of a large amount of natural image, the image reconstruction algorithm being more suitable for reconstruction of two-dimensional images put forward.The Total Variation of compression of images is as follows

minTV(f)s.t.y＝Φf(4)

Objective function TV (f) is image discrete gradient sum, namely

$\mathrm{TV}\left(f\right)={\mathrm{\&Sigma;}}_{\mathrm{ij}}\sqrt{{(f_{i+1}-f_{i,j})}^2+{(f_{i,j+1}-f_{\mathrm{ij}})}^2}---\left(5\right)$

Claims (3)

1. one kind is carried out the compressive sampling method of optical modulation with liquid crystal display, it is characterized in that: the light that object being measured sends is imaged on liquid crystal display through optical imagery parts, each pixel of liquid crystal display is equivalent to a photoswitch, liquid crystal display driving circuit controls the torsion of liquid crystal, to change its transmittance, make light selectively through liquid crystal display, changed by the transmittance of liquid crystal display, physically realize a programmable observing matrix, light through liquid crystal display injects fourier lense, photodiode is at lens focus place receiving light power, by to light intensity integration, the result obtained is equivalent to the observed reading in compressed sensing theory, the method comprises the following steps:

Step 1: the light that object being measured sends is imaged on liquid crystal display through optical imagery parts, the picture now on liquid crystal display is original signal X _{n × 1};

Step 2: liquid crystal display driving circuit controls the torsion of liquid crystal, the pixel on liquid crystal display is arranged and is in pseudo-random state i, wherein 1≤i≤M, pseudo-random state i constitute the i-th row h of observing matrix Φ _i, will be now that X is at h through the signal of liquid crystal display _iunder value;

Step 3: the light through liquid crystal display injects fourier lense, photodiode, at lens focus place receiving light power, by light intensity integration, obtains result y _i=h _ix _{n × 1};

Step 4: repeat step above M time, then M liquid crystal screen image element ordered state constitutes observing matrix Φ _{m × N}, M time measurement result forms observation data matrix Y=Φ X.

2. liquid crystal display as claimed in claim 1 carries out the compressive sampling method of optical modulation, it is characterized in that: adopt the liquid crystal display without the monochromatic adjustable gray scale of backlight to carry out optical modulation to original image signal, by the Physical realization of liquid crystal display as observing matrix.

3. liquid crystal display as claimed in claim 1 carries out the compressive sampling method of optical modulation, it is characterized in that: photodiode at fourier lense focus place receiving light power, and carries out integration to the light signal after liquid crystal display modulation.