30 de mayo de 2023 a 2 de junio de 2023 Ciencias Naturales, Exactas y Ténicas
Facultad de Matemática y Computación
America/Havana zona horaria

Nonlinear diffusion by regions. KMLS a new version, comparison with other gradient threshold estimations for the diffusion coefficient

No programado
20m
Facultad de Matemática y Computación

Facultad de Matemática y Computación

Ponente

Sr. Reinaldo Barrera Travieso (Instituto de Geofisica y Astronomia)

Descripción

In this paper, the nonlinear diffusion model applied to image smoothing (ADIMP), known as the Perona-Malik anisotropic model, is studied. This model is characterized by requirements on the diffusion coefficient, hence the changes that occur in the resulting family of images, depend on the gradient based coefficient expression’s and on the way of estimating the gradient threshold. In this work it is proposed to carry out the non-linear diffusion by regions, using the SLIC method for segmentation. A modification to the KMLS method for estimating the parameter 𝜅 (gradient threshold) is proposed and a comparison with other coefficients expressions is done. The influence of ADIMP on edge detection in mammography images from the Inbreast database is proved.

Autor primario

Sr. Reinaldo Barrera Travieso (Instituto de Geofisica y Astronomia)

Coautores

Angela León Mecías (Universidad de La Habana) Sr. Richard Mendez Castillo (Universidad de La Habana)

Materiales de la presentación

Todavía no hay materiales.