Publication: Problemas de tratamiento de imágenes basados en la incorporación de características humanas
Loading...
Official URL
Full text at PDC
Publication Date
2020-03-13
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad Complutense de Madrid
Citations
Exportar
Abstract
Esta memoria se enmarca dentro del tratamiento de imágenes mediante ordenador -image processing-, disciplina que ha ido encontrando cada vez más aplicaciones durante las últimas décadas. Los algoritmos desarrollados para trabajar con imágenes digitales pueden realizar tareas que oscilan desde un bajo nivel de procesamiento -donde se suele enmarcar el problema de detección de bordes- a un nivel alto -reconocimiento de objetos-. Al estar en la base del tratamiento de imágenes -junto con técnicas como el suavizado-, la detección de bordes juega un papel fundamental como paso inicial de muchas de estas técnicas y aplicaciones. Esto se debe a que, por ejemplo, reconocer un objeto en una imagen requiere en general el hallar previamente las diferencias importantes en la información espectral entre los píxeles. Para encontrar estas diferencias hace falta pasar por una serie de fases, la última de las cuales es el escalado. En esta fase final, se establece un método que permite decidir si estas diferencias -edginess- significan que dicho píxel tiene que ser declarado un borde o no. El enfoque clásico durante esta fase de escalado consiste en tomar la decisión de si hay un borde o no estudiando píxel a píxel. Esto puede considerarse como un método de evaluación local, el cual presenta problemas prácticos, ya que los píxeles candidatos a borde no deben considerarse independientes...
This memoir is framed into digital image processing, field that has found more and more applications along the last decades. The algorithms developed for working on digital images can range from low-level processing tasks –where the edge detection problem is localised- to a high-level processing -as object recognition-. As it stays in the base of other important techniques, edge detection plays an important role in image processing field. This is due to for recognizing an object in an image typically requires finding the important differences in spectral information between pixels. In order to find these differences is required going through several phases, being the last one of them the scaling phase. During this last phase is decided if these differences (i.e edginess) mean that the pixel should be declared as an edge or not. The classic approach in the scaling phase is based in taken the decision if the pixel is an edge or not just considering adjacent pixels of a certain one. This is considered as a local evaluation method, and this approach presents important limitations because the pixels that are candidate to be edge should not been considered as independents...
This memoir is framed into digital image processing, field that has found more and more applications along the last decades. The algorithms developed for working on digital images can range from low-level processing tasks –where the edge detection problem is localised- to a high-level processing -as object recognition-. As it stays in the base of other important techniques, edge detection plays an important role in image processing field. This is due to for recognizing an object in an image typically requires finding the important differences in spectral information between pixels. In order to find these differences is required going through several phases, being the last one of them the scaling phase. During this last phase is decided if these differences (i.e edginess) mean that the pixel should be declared as an edge or not. The classic approach in the scaling phase is based in taken the decision if the pixel is an edge or not just considering adjacent pixels of a certain one. This is considered as a local evaluation method, and this approach presents important limitations because the pixels that are candidate to be edge should not been considered as independents...
Description
Tesis inédita de la Universidad Complutense de Madrid, Facultad de Estudios Estdísticos, leída el 08/11/2019