Image deblurring algorithm based on sparse positive source separation model

Patent metadata

Field	Value
Publication number	US-10937131-B2
Application number	US-201716341841-A
Country	US
Kind code	B2
Filing date	Jun 15, 2017
Priority date	Jun 15, 2017
Publication date	Mar 2, 2021
Grant date	Mar 2, 2021

Abstract

Provided is an image deblurring algorithm based on a sparse positive source separation model. The image deblurring algorithm is used for performing processing on a blurry image collected by an optical microscopic imaging system and generated due to diffraction effect and optical deviation; under a condition of single light sensitive imaging and of not increasing an external imaging device, a spatial resolution of the optical microscopic system may be improved to a nanoscale. In the disclosure, a blurring process of microscopic imaging is expressed as a linear combination of a Point Spread Function of the imaging system; by embedding the process into a positive separation optimized frame, adding a sparsity constraint and solving to remove blurring, the high-resolution microscopic imaging is implemented.

First claim

The invention claimed is:

1. An image deblurring method based on a sparse positive source separation model, comprising:
S1: performing calibration on an optical system;
S2: performing sparsification on an image content; and
S3: establishing an imaging matrix A and solving a sparse positive source separation optimized model;
wherein in the step S2, when multiple pieces of continuous microscopic images are available, the main purpose of the sparsification on the image content is to remove background interference, an interested sparse structure is viewed as a foreground, the background interference is viewed as a background, and the sparsification is performed on the images using sparse low-rank matrix decomposition; and when only one piece of blurry microscopic image is available, the main purpose of the sparsification on the image content is to remove a direct-current component in the image, comprising a spontaneous duration signal from an imaging device during an exposure process, or a smooth image content obtained because the background reflects light/emits the light, and the sparsification is performed on the image using a direct-current component elimination method.

2. The image deblurring method based on the sparse positive source separation model according to claim 1 , wherein the specific process of the step S1 is as follows:
for a microscopic system with an amplification factor being a, a pixel size being b and a diffraction limit being d, if a point light source with a physical size not greater than b/a is present in an observation range and no other point light sources are present within a range at a distance being d, a light spot generated by the point light source in an imaging plane of the microscopic system is considered as an effective observation of a PSF (Point Spread Function) of the system; three times or more of effective observations of the PSF are collected and are averaged, and then a w 1 *w 2 image block where the light spot is located after the averaging is taken as an estimation value of the PSF of the optical system, wherein both the w 1 and the w 2 are an integral value slightly greater than d/b; and when a point light source with a suitable physical size is not present in the observation range, an image that is scaled and clipped properly using an existing PSF is taken as the estimation value of the PSF to establish and solve the sparse positive source separation optimized model.

3. The image deblurring method based on the sparse positive source separation model according to claim 2 , wherein a process of establishing the imaging matrix in the step S3 is as follows:
S31: placing a PSF of one optical system at an i th pixel of an H*W spatial image;
S32: rearranging pixels of the spatial image with the PSF into HW*1 column vectors; and
S33: performing normalization on the column vectors so that a Euclidean norm is 1;
the sparse positive source separation optimized model solved in the step S3 is as follows:

min
x

⁢



y
-
Ax



1

+

λ
⁢


x


1

⁢

⁢

s
.
t
.

⁢
x

≥
0

through solving the above model, HW*1 column vectors x corresponding to a clear image may be obtained, wherein the y is column vectors of HW*1 rearranged by the pixels of the blurry image, the x is the HW*1 column vectors corresponding to the clear image, the λ is an equilibrium parameter, and the value of the λ should be adjusted adaptively according to a noise level and an energy level of the y.

4. The image deblurring method based on the sparse positive source separation model according to claim 2 , wherein if a to-be-processed image has a large resolution, the image is divided into a plurality of overlapped image blocks for processing; to divide the image blocks, two principles need to be met: 1) the overlapped length on a height direction is not smaller than w 1 , and the overlapped length on a width direction is not smaller than w 2 ; and 2) the image blocks should be large enough in size so that the image content meets a sparse hypothesis of the sparse positive source separation optimized model.

Assignees

• Univ Sun Yat Sen

Inventors

• Chen Chongyu
• Lin Liang

Classifications

• G02B21/008 Primary

  Details of detection or image processing, including general computer control · CPC title

• G06T2207/10056

  Microscopic image · CPC title

• G06T2207/20021

  Dividing image into blocks, subimages or windows · CPC title

• G06T5/20

  using local operators · CPC title

• G02B21/365

  Control or image processing arrangements for digital or video microscopes (G02B21/361, G02B21/362 take precedence) · CPC title