Browsing by Author "Barrios E.M."

Now showing 1 - 2 of 2

LRemoving Dust Artifacts in Retinal Images via Dictionary Learning and Sparse-Based Inpainting
(Institute of Electrical and Electronics Engineers Inc., 2019) Barrios E.M.; Marrugo A.G.; Millán M.S.
In the field of ophthalmology, retinal images are essential for the diagnosis of many diseases. These images are acquired with a device called the retinal camera. However, often small dust particles in the sensor produce image artifacts that can be confused with small lesions, such as micro-aneurysms. The digital removal of artifacts can be understood as an inpainting process in which a set of pixels are replaced with a value obtained from the surrounding area. In this paper, we propose a methodology based on the sparse representations and dictionary learning for the removal of artifacts in retinal images. We test our method on real retinal images coming from the clinical setting with actual dust artifacts. We compare our restoration results with a diffusion-based inpainting technique. Encouraging experimental results show that our method can successfully remove the artifacts, while assuring the continuity of the retinal structures, like blood vessels. © 2019 IEEE.
Noise-Robust Processing of Phase Dislocations using Combined Unwrapping and Sparse Inpainting with Dictionary Learning
(Institute of Electrical and Electronics Engineers Inc., 2019) Pineda J.; Meza J.; Barrios E.M.; Romero L.A.; Marrugo A.G.
The problem of phase unwrapping from a noisy and also incomplete wrapped phase map arises in many optics and image processing applications. In this work, we propose a noise-robust approach for processing regional phase dislocations. Our approach combines phase unwrapping and sparse-based inpainting with dictionary learning to recover the continuous phase map. The method is validated both using numerically simulated data with strong additive white Gaussian noise and phase dislocations; and experimental data from fringe projection profilometry. Comparisons with other phase inpainting method referred to as PULSI+INTERP, show the suitability of the proposed method for phase restoration even in extremely noisy phases. The error given by the proposed method on the highest level of noise (RMSE=0.0269 Rad) remains the smallest compared to the error given by PULSI+INTERP for noise-free data (RMSE=0.0332 Rad). © 2019 IEEE.