Técnicas computacionales para la mejora de la resolución y análisis de imágenes de microscopía electrónica

  1. Bárcena González, Guillermo
Supervised by:
  1. María de la Paz Guerrero Lebrero Director
  2. Elisa Guerrero Vázquez Director

Defence university: Universidad de Cádiz

Fecha de defensa: 26 June 2017

Committee:
  1. Nicolai Petkov Chair
  2. Andrés Yáñez Escolano Secretary
  3. Ana María Beltrán Custodio Committee member
Department:
  1. Ingeniería Informática

Type: Thesis

Teseo: 481282 DIALNET

Abstract

ABSTRACT Nanotechnology encompasses the understanding of the fundamental physics of nanometer-scale objects, and in this mater Image processing plays an important role. In particular, quantitative analysis of nanostructures is based on analysis of images obtained from electron microscopes. Nowadays, in spite of having sophisticated devices for image capture, some limitations (blur, lens aberrations, lens aperture, diffraction, brightness, etc.) during the acquisition process can affect the final resolution of the images, where resolution is defined as the degree of detail and quality of an image, and not only the number of pixels of it. In order to overcome these limitations, we focus on the study and development of advanced computational tools for enhancing the resolution and analysis of electron microscopy images to enable quantitative characterization and dynamic visualization. The drift effect inherent in the images obtained from STEM microscopes (HAADF images), is one of the main distortions dealt in the literature. Different methods have been studied and an alternative that improves the computational time has been proposed in this PHD thesis. Our proposal is based on the analysis of the second harmonics angle of the Fourier transform in order to calculate the drift effect. Successful validations with simulated and experimental images have been carried out. Given that Fourier transform allows improving the computational efficiency of certain calculations, and also facilitates the observation and analysis of different properties of the images that are not possible to realize from the real space, a new approach for strain mapping, using the inter-dumbbell distance is introduced. Although not validated experimentally, its potentiality has been highlighted, and this novelty opens a new way for strain mapping calculations, where images with higher resolution will be needed. In this context the application of Super-Resolution techniques to the Electron Microscopy field has been addressed. A customized Super-Resolution methodology to enhance HAADF-images has been proposed. This methodology includes, in the alignment stage, a new method to align the images and an algorithm for selecting the reference image. The application of this algorithm guarantees a minimization of the drift effect in the final image. The complete methodology has been evaluated experimentally with different materials, showing the improvement in the images quality using different qualitative and quantitative measures. In addition, different modifications to optimize the computational time have been adopted as well as new proposals have been introduced to adapt the restoration algorithms to images from materials with crystalline structures. A correspondence between the material lattice and the algorithms parameters has been found. The present research work allows the improvement of the quality of electron microscopes images, providing higher accuracy in the analysis of materials.