Image operator learning based on local features

• Grant number: 17/09137-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: July 01, 2017
• End date: December 31, 2017
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Nina Sumiko Tomita Hirata
• Grantee: Augusto César Monteiro Silva
• Host Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil

Abstract

Image operators that are translation-invariant and locally defined with respect to a neighborhood delimited by a window W can be characterized by a function whose domain is the set of all images restricted to W. The result of an image processing transformation through this type of operator can be computed by applying, on each pixel in the image, the corresponding function. These operators can be designed from samples of input-output pairs of images that illustrate the desired transformation, using machine-learning techniques. In this project we investigate image operator learning based on local features of the images. Local feature extractors will be implemented within TRIOS lib (a library developed and maintained by our group) and experiments will be carried out to compare operators designed from the extracted features and those designed from methods already in TRIOS lib. Experimental evaluation will be performed on known image processing related public datasets. This project is related to the works on image operator learning conducted at the e Science Lab at IME-USP. (AU)