CN105389785A - Processing method of point spread function - Google Patents

Abstract

The invention discloses a processing method of a point spread function in the technical field of microscopes. A method for processing a measurement matrix composed of point spread functions and low-resolution images acquired and obtained by a microscope to obtain a high-resolution reconstructed image is provided. The method comprises the steps of performing one-dimensional processing for point spread functions to form vectors; composing the measurement matrix by using the point spread function vector at each high-resolution grid center in order; then processing the measurement matrix and the low-resolution images recorded by a recording camera based on the microscope; and finally reconstructing through a compression sensing reconstructing algorithm to obtain high-resolution images. The processing method of the point spread function lays a foundation for compression sensing based optical super-resolution micro-imaging of super-diffraction limit to develop from theoretical research to functionization.

Description

A kind of disposal route of point spread function

Technical field

The invention belongs to microscopy field, specifically provide a kind of disposal route of point spread function.

Background technology

Super-resolution micro-imaging can make human eye directly observe microprocess and the structure of amplification.Super-resolution micro-imaging is widely used and demand in fields such as biology, medical science, material, precision optical machinery and microelectronics, and the solution for each subject engineering and key scientific issues is significant.The optical microscope of visible light wave range has the advantage of noncontact, not damaged, detectable sample interior, is the important tool of research organ, tissue and cell.Due to the existence of diffraction limit, horizontal and vertical resolution only has 200nm and 500nm respectively.The optical ultra-discrimination micro-imaging theory and technology of super diffraction limit is the focus of research always.

The optical ultra-discrimination micro-imaging realizing super diffraction limit that appears as of compressed sensing technology provides possibility.The optical ultra-discrimination micro-imaging of current super diffraction limit can be avoided one or minority fluorescence molecule once can only to make the repetition tedious work of image acquisition by compressed sensing binding site spread function, thus save huge human and material resources, shorten the collection reconstruct cycle of super-resolution imaging, greatly improve temporal resolution, simplify history image file administration, reduce data volume, for noise process and control also can more simplify.

Summary of the invention

The present invention is in order to the low problem of MIcrosope image resolution, and spy provides a kind of disposal route of point spread function.

The present invention is achieved by following proposal: a kind of disposal route of point spread function, and the process of described method is:

Step one: based on experiment by Gaussian function or the microscopical point spread function of Bessel function fitting, the fine-resolution meshes that the low-resolution image y of microscopical image record cameras record divides pixel less;

Step 2: the source point being point spread function with each grid element center, the numerical value of measuring point spread function in each pixel of low-resolution image, then by end to end for the numerical value row or column in each pixel, formation one is based on the column vector Φ of point spread function _j;

Step 3: based on the column vector Φ of the point spread function of all grid mid points of fine-resolution meshes _jaccording to the matrix Φ of the end to end formation of all grid mid point row or column based on point spread function, wherein Φ ∈ R ^{m × N}, M < N, M and N are natural numbers, Φ _jthe jth row of Φ;

The each row orthogonalization of step 4: Φ, then each row is unitization, and last Φ respectively arranges unitization, obtains matrix Φ _p; Make Q=Φ _pΦ ' (Φ Φ ') ^-1, matrix Φ _s=Q Φ;

Step 5: by microscopical image record collected by camera record low-resolution image y;

Step 6: the low-resolution image y that microscope collects is processed by matrix Q:

y _S＝Qy；

Step 7: reconstruct real super-resolution image x by traditional signal reconstruction algorithm:

min||x|| ₀s.t.y _S＝Φ _Sx。

The present invention, by point spread function of one-dimensional, forms calculation matrix.Then by calculation matrix and low resolution image process, high-resolution reconstructed image is finally obtained.The method is greatly high signal reconstruction ability and resolution.Method of the present invention not only simplifies microscopical hardware design and realization, and improves signal reconstruction effect, has a wide range of applications in fields such as microscopical image procossing.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the disposal route of a kind of point spread function described in embodiment one; Fig. 2 applies embodiment based on OMP restructing algorithm to matrix Φ ^{169 × 4096}, matrix size 169 × 4096, the Accurate Reconstruction probability before and after the process that process obtains and the graph of a relation of degree of rarefication; Fig. 3 applies embodiment based on BP restructing algorithm to matrix Φ ^{169 × 4096}, matrix size 169 × 4096, the Accurate Reconstruction probability before and after the process that process obtains and the graph of a relation of degree of rarefication; Fig. 4 applies embodiment based on CVX restructing algorithm to matrix Φ ^{169 × 4096}, matrix size 169 × 4096, the Accurate Reconstruction probability before and after the process that process obtains and the graph of a relation of degree of rarefication.

Embodiment

Embodiment one: illustrate present embodiment according to Figure of description 1.A disposal route for point spread function, the process of described method is:

Step one: based on experiment by Gaussian function or the microscopical point spread function of Bessel function fitting, the fine-resolution meshes that the low-resolution image y of microscopical image record cameras record divides pixel less;

Step 2: the source point being point spread function with each grid element center, the numerical value of measuring point spread function in each pixel of low-resolution image, then by end to end for the numerical value row or column in each pixel, formation one is based on the column vector Φ of point spread function _j;

Step 3: based on the column vector Φ of the point spread function of all grid mid points of fine-resolution meshes _jaccording to the matrix Φ of the end to end formation of all grid mid point row or column based on point spread function, wherein Φ ∈ R ^{m × N}, M < N, M and N are natural numbers, Φ _jthe jth row of Φ;

The each row orthogonalization of step 4: Φ, then each row is unitization, and last Φ respectively arranges unitization, obtains matrix Φ _p; Make Q=Φ _pΦ ' (Φ Φ ') ^-1, matrix Φ _s=Q Φ;

Step 5: by microscopical image record collected by camera record low-resolution image y;

Step 6: the low-resolution image y that microscope collects is processed by matrix Q:

y _S＝Qy；

Step 7: reconstruct real super-resolution image x by traditional signal reconstruction algorithm:

min||x|| ₀s.t.y _S＝Φ _Sx。

Embodiment two: this embodiment is further illustrating the disposal route of a kind of point spread function described in embodiment one, low-resolution image y large in step one can be the image of the arbitrary resolution that existing microscope adopts arbitrary resolution record collected by camera to obtain, y ∈ R ^m, x ∈ R ⁿ.

Embodiment three: illustrate present embodiment below in conjunction with Fig. 2-Fig. 4.Present embodiment adopts the gaussian signal of the different degree of rarefication corresponding matrix Φ for the treatment of front and back respectively ^{169 × 4096}, matrix size 169 × 4096, the reconstruct SBR compared after its each 500 experiments is greater than the reconstruct probability of 90dB.Be with in Fig. 2 what mark is curve before process; Band what mark is curve after process.Fig. 2 represents the result of calculation of OMP restructing algorithm; The result of calculation of BP restructing algorithm is represented in Fig. 3; Fig. 4 represents the result of calculation of CVX restructing algorithm.

Experimental result as shown in figs 2-4.As seen from Figure 2, based on OMP restructing algorithm, the quality reconstruction curve after process is positioned at top before treatment, and quality reconstruction is better; Based on BP restructing algorithm, the quality reconstruction curve after process is positioned at top before treatment, and quality reconstruction is better; As seen from Figure 4, based on CVX restructing algorithm, the quality reconstruction curve after process is positioned at top before treatment, and quality reconstruction is better.

Claims (1)

1. a disposal route for point spread function, is characterized in that: the process of described method is:

Step one: based on experiment by Gaussian function or the microscopical point spread function of Bessel function fitting, the fine-resolution meshes that the low-resolution image y of microscopical image record cameras record divides pixel less;

Step 2: the source point being point spread function with each grid element center, the numerical value of measuring point spread function in each pixel of low-resolution image, then by end to end for the numerical value row or column in each pixel, formation one is based on the column vector Φ of point spread function _j;

Step 3: based on the column vector Φ of the point spread function of all grid mid points of fine-resolution meshes _jaccording to the matrix Φ of the end to end formation of all grid mid point row or column based on point spread function, wherein Φ ∈ R ^{m × N}, M < N, M and N are natural numbers, Φ _jthe jth row of Φ;

The each row orthogonalization of step 4: Φ, then each row is unitization, and last Φ respectively arranges unitization, obtains matrix Φ _p; Make Q=Φ _pΦ ' (Φ Φ ') ^-1, matrix Φ _s=Q Φ;

Step 5: by microscopical image record collected by camera record low-resolution image y;

Step 6: the low-resolution image y that microscope collects is processed by matrix Q:

y _S＝Qy；

Step 7: reconstruct real super-resolution image x by traditional signal reconstruction algorithm:

min||x|| ₀s.t.y _S＝Φ _Sx。