Deep frequency-recurrent priors for inverse imaging reconstruction

•A multi-profile high-frequency transform-guided denoising autoencoder for attainting deep frequency recurrent prior.•We extract a set of multi-profile high-frequency components via a specific transformation.•The learned high-frequency prior information is incorporated into classical iterative recon...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Signal processing Ročník 190; s. 108320
Hlavní autoři: He, Zhuonan, Hong, Kai, Zhou, Jinjie, Liang, Dong, Wang, Yuhao, Liu, Qiegen
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier B.V 01.01.2022
Témata:
Denoising autoencoder network
High-frequency component
Imaging reconstruction
MRI reconstruction
Proximal gradient descent
Sparse-view CT
Sparse-view CT
Denoising autoencoder network
Imaging reconstruction
Proximal gradient descent
High-frequency component
MRI reconstruction
ISSN:0165-1684, 1872-7557
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:•A multi-profile high-frequency transform-guided denoising autoencoder for attainting deep frequency recurrent prior.•We extract a set of multi-profile high-frequency components via a specific transformation.•The learned high-frequency prior information is incorporated into classical iterative reconstruction. Ill-posed inverse problems in imaging remain an active research topic in several decades, with new approaches constantly emerging. Recognizing that the popular dictionary learning and convolutional sparse coding are both essentially modeling the high-frequency component of an image, which convey most of the semantic information such as texture details, in this work we propose a novel multi-profile high-frequency transform-guided denoising autoencoder for attaining deep frequency-recurrent prior (DFRP). To achieve this goal, we first extract a set of multi-profile high-frequency components via a specific transformation and add artificial Gaussian noise to these high-frequency components as training samples. Then, as the high-frequency prior information is learned, we incorporate it into classical iterative reconstruction by proximal gradient descent. Preliminary results on highly under-sampled magnetic resonance imaging and sparse-view computed tomography reconstruction demonstrated that the proposed method can efficiently reconstruct feature details and presented advantages over state-of-the-arts. The network architecture at the training procedure and one-iteration diagram at the iterative reconstruction of DFRP. (a) The DAE network used in DFRP. (b) The illustration of DFRP at iterative reconstruction phase. The Conv layer is denoted as “C”. Additionally, the Concatenation layer is denoted as “Cat”. Here the example of MRI reconstruction is visualized. [Display omitted]
ISSN:0165-1684
1872-7557
DOI:10.1016/j.sigpro.2021.108320