Papers
Compressive Spectral Image Reconstruction by using a Deep Image Prior with a Mismatch Regularizer

Compressive Spectral Image Reconstruction by using a Deep Image Prior with a Mismatch Regularizer

Jul 11, 2022·
Romario Gualdrón-Hurtado
Romario Gualdrón-Hurtado
Jorge Bacca
Jorge Bacca
Henry Arguello
Henry Arguello
Abstract
A compressive image recovery method that includes a regularizer in the baseline deep image prior is proposed to consider the calibration sensing model mismatch.
Type
Publication
Computational Optical Sensing and Imaging 2022

Proposed Method

screen reader text
Visual representation of the proposed method, where the upper branch obtains the spectral image, and the other branch considers the model mismatch.

Qualitative results

screen reader text
(a) Testbed CASSI implementation. (b) RGB scene, visual representation of the scene obtained with DIP, and the proposed method. (c) Normalized spectral signatures of the recovered scenes in the two selected points.

Romario Gualdrón-Hurtado
Authors
Romario Gualdrón-Hurtado
M.Sc. (s) Systems Engineering
Jorge Bacca
Authors
Jorge Bacca
Assistant professor at Universidad Industrial de Santander
Henry Arguello
Authors
Henry Arguello
Titular professor at Universidad Industrial de Santander