Non local Markov Random Fields: a new model and its application in image denoising

• Grant number: 13/25595-7
• Support Opportunities: Regular Research Grants
• Duration: April 01, 2014 - March 31, 2016
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Denis Henrique Pinheiro Salvadeo
• Grantee: Denis Henrique Pinheiro Salvadeo
• Host Institution: Instituto de Geociências e Ciências Exatas (IGCE). Universidade Estadual Paulista (UNESP). Campus de Rio Claro. Rio Claro , SP, Brazil
• Associated researchers: Alexandre Luís Magalhães Levada ; Nelson Delfino d'Ávila Mascarenhas

Abstract

In general, in the several modalities of imaging systems, researchers are very much interested in reducing noise of images (that depends on the modality), especially keeping structural details and resolution. Still, several of these modalities as Breast Tomosynthesis and Computed Tomography (CT) use the dangerous ionizing radiation for imaging. Thus, researchers in this area follow the ALARA principle (As Low As Reasonably Achievable), in order to use the lowest possible dose of radiation. However, reducing the dose in these systems implies increased noise in the acquired image. Based on these reasons, this work plan aims at investigating a new Markov Random Fields (MRF) model regarding, especially, the state of the art approach, which uses the redundant information contained in different patches (non local) of an image. This new model will be named Non Local MRF (NLMRF). As main advantages of this new model, we highlight: 1) the use of radiometric information in the selection of MRF cliques by considering the weighting of the energies from the model according to the similarity between the patches, 2) the generation of an adaptive algorithm to the noise type by considering a Bayesian Framework for obtaining the similarity measures between patches more suitable to each noise type and 3) to the local noise level provided by a parameter defined in terms of the Fisher Information. Thus, it is expected that, by using the NLMRF model to image denoising, especially of CT, Breast Tomosynthesis and Ultrasound, we may achieve a better balance between noise reduction and details preservation such as edges and other structures, when compared to existing methods, besides the possibility of reducing the radiation dose in systems which use ionizing radiation. (AU)

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

SALVADEO, DENIS H. P.; MASCARENHAS, NELSON D. A.; LEVADA, ALEXANDRE L. M.. Nonlocal Markovian models for image denoising. JOURNAL OF ELECTRONIC IMAGING, v. 25, n. 1, JAN 2016. (10/09248-7, 13/25595-7)