Image reconstruction method based on second-order L0 minimization and edge priori

Abstract

The invention discloses an image reconstruction method based on second-order L0 minimization and edge prior. The method comprises the following steps: wavelet transform is carried out on the originalimage, and then the original image is measured by using a multivariable sampling mechanism to obtain all low-frequency data CLL and desampled high-frequency data YLH, YHL, YHH; the low-frequency dataCLL and the zeroed high-frequency wavelet coefficients are transformed by inverse wavelet transform to obtain the low-frequency image XLL. The first-order L0 gradient minimization is used to extract the edge of low-frequency image XLL, and the multivariate prior edge information ELH, EHL, EHH is obtained. The second-order L0 minimization is used to solve the image sparse reconstruction problem under the constraint of multivariate edge priori information, and the reconstruction high-frequency coefficient matrices RecXLH, RecXHL, RecXHH are obtained. The inverse wavelet transform of CLL, RecXLH,RecXHL, RecXHH is used to get the reconstructed image RecX. This method can improve the precision of image reconstruction, and has the characteristics of high precision and good visual effect.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to an image reconstruction algorithm based on second-order L0 minimization and edge prior, which can be applied to practical engineering fields such as image compression and imaging. Background technique [0002] Under the compressed sensing framework, let the original image be The sparse basis of wavelet transform is ψ∈R N*N , the measurement matrix is ​​φ∈R M*N , then the measurement signal y=φ*ψ*x=Θ*x∈R can be obtained M*1 . Finally, the image reconstruction problem can be solved by the following equation. [0003] [0004] where ‖x‖ 0 is the L0 norm of x, representing the number of non-zero elements in x. [0005] Due to the downsampled measurement method, the results of image reconstruction are generally inaccurate, especially when the number of measurements is relatively small. Moreover, the image sparse reconstruction problem based on the above formula is ...

AI Technical Summary

Problems solved by technology

However, natural images do not have obvious edge structures, and contain a lot of low-frequency noise detail structure information

Therefore, how to extract accurate edge structures from complex natural images as prior information provides challenges for image reconstruction

[0007] Based on the above, under the compressed sensing framework, due to the downsampling measurement method, the image reconstruction accuracy is low, and the visual effect is poor, especially when the number of measurements is relatively small. In order to solve the above problems, this case arises

Images