« Distance-based pattern complexity for morphology quantification and image quality assessment » by Johan Chaniot

Seminar by Johan Chaniot, Chargé d'enseignement et de recherche, Ecole des Mines de Saint-Etienne

Abstract

Image complexity is a multifaceted concept of paramount interest to various domains of image processing and analysis. Consequently, complexity manifests in numerous forms as image quality and structure morphology, encompassed within the term “pattern complexity”. Recent literature has exposed certain limitations related to pattern complexity quantification either for morphological characterisation (MC) in complex microstructure analysis or for image quality assessment (IQA) in live cells imaging. The objective pursued here is how to improve the quantification of pattern complexity to better adapt it to the characteristics of images of real structures? The synergy of IQA (which needs to more accurately reflect structural features) and  MC (which needs to be extended to more complex scenarios) will be explored, in particular through distance fields, a versatile representation of digital images in a multidimensional vectorial space. Indeed, valuable information such as connectivity and tortuosity measures can be extracted from distance maps as they encode local and global information. Various application domains as neuroscience, materials science and process engineering could benefit from such work.