Author
Mohsen Shahrezaei
;
Alireza Shujaei Fard
;
Hamidreza Yazdani
Author Original
محسن شاهرضایی علیرضا شجاعی فرد حمیدرضا یزدانی
Publication Date
1400-10
Subject
Digital image recovery, tensor, high order tensor, tensor completion
Type
Periodical
Language
Persian
Digital
Yes
Manuscript
No
Library
University of Toronto
Library Asset ID
ISSN: 2251-8088, EISSN: 2645-6141, DOI: 10.22055/jamm.2021.38887.1971
Record ID
cdi_doaj_primary_oai_doaj_org_article_7986add877f34686a453b5e3b7bd6b01
Library Location
DOAJ Directory of Open Access Journals
Date
1400-10
Notes
High-order tensorial structural data are used in many imaging scenarios such as hyperspectral imaging and color videos. Retrieving a tensor from an incomplete set of elements, which is called tensor completion, has many applications in fields such as digital image processing and compression. In tensorial completion, in addition to the incompleteness of the observed data, another problem is the quantification of the verses. Quantization is an important step for the transmission and storage of high-dimensional data in order to reduce the need for storage and save energy. Here, a new method for recovering low-rank tensors from a small number of binary (single-bit) measurements is presented. The single-bit tensor completion method relies on the application of tensor completion in matrixed versions with binary data in the contextual tensor of the data. Experimental results in hyperspectral images show that direct operation with binary measurement, instead of their true values, leads to less recovery error. Here, a given third-order tensor is recovered with binary terms. In practice, we open the tensor as a ternary hypermatrix and apply the quantified tensorial completion algorithm to all models of the said tensor matrix. The data space here is hyperspectral satellite images with the aim of recovering distorted images.
Erişim bilgileri
Available Online
