Models, Inference and Synthesis for Texture In Color – MISTIC

This project intend to develop new approaches and conceptual methodology to set up a theoretical sound framework for texture
modeling . Texture analysis is a fundamental problem in image processing with numerous fields of applications in medical imaging,
computer graphics or data based indexation and classification. The original proposed work forms an interface between different
fields of expertise : probability and statistics, image and signal processing, computer science and automation. We will mainly focus
on statistical descriptors of textures with a specificity for vectorial data treatment. One of the project's main objectives is that the
features statistics computed from the texture may be described and connected to statistical properties of vector-valued parametric
random fields for color imaging in both continuous and discrete setting. The questions about the characterizations of the features as
well as the synthesis process from the identified models are core issues.