Search

In this paper we focus on learning optimized partial differential equation (PDE) models for image filtering. In this approach, the grey-scaled images are represented by a vector field of two real-valued functions and the image restoration problem is modelled by an evolutionary process such that the restored image at any time satisfies an initial-boundary-value problem of cross-diffusion with reaction type. The coupled evolution of the two components of the image is determined by a nondiagonal matrix that depends on those components. A critical question when designing a good-performing filter lies in the selection of the optimal coefficients and influence functions which define the cross-diffusion matrix. We propose the use of deep learning techniques in order to optimize the parameters of the model. In particular, we use a back propagation technique in order to minimize a cost function related to the quality of the denoising processe, while we ensure stability during the learning procedure. Consequently, we obtain improved image restoration models with solid mathematical foundations. The learning framework and resulting models are presented along with related numerical results and image comparisons.

Dissertação de Mestrado em Matemática, área de Especialização em Análise Aplicada e Computação, apresentada à Faculdade de Ciências e Tecnologia da Universidade de Coimbra. A resolução numérica da equação fraccionária do calor ut + (-Δ) α/2u = 0, tem sido um área de investigação muito activa nas últimas décadas. Na literatura, a mesma notação para o operador de difusão aparece associada a definições diferentes, não sendo claro se estas são equivalentes. Neste trabalho estudamos maioritariamente duas definições para este operador: o operador de Riesz na forma integral e o operador de Riesz espectral. São apresentados métodos numéricos que convergem para a solução do problema de difusão fraccionário respectivo a cada definição. Por último, usando estes métodos comparamos as soluções numéricas dos problemas associados a cada definição. The numerical resolution of the fractional heat equation ut + (-Δ) α/2u = 0, has been a topic of great interest in the last decades. In the literature the same notation for the diffusion operator is associated with different definitions, and it is unclear if they are equivalent. In this work we study two of the major definitions for this operator: the integral Riesz operator and the spectral Riesz operator. We introduce numerical methods that converge for the solution of the fractional diffusion problem associated with each definition. We finish by using this methods to compare the numerical solutions of each problem.