Image separation and geometric characterisation of mud flocs

This paper presents an approach to process and analyse digital images of mud flocs recorded via optical techniques. Grey-scale images of mud flocs consist of bright spots on a dark background, the recordings being characterised by a bimodal intensity (luminosity) distribution. First, we propose a fast algorithm to separate such two-intensity regions to obtain black and white images via Gaussian-curve fitting. This algorithm is evaluated in its capability to recognise the mass distribution and perimeter segmentation of mud flocs. Next, we characterise these images of mud flocs by using the concepts of multifractality, entropy and symmetry. This approach gives indication that a complex behaviour rises in a population of mud flocs, which evolves far from equilibrium. Multifractal scaling and differentiation into dominant and rare geometrically featured flocs are inferred to mark the evolution of a population of cohesive sediment flocs. The mathematical tools and method of analysis presented here are applied to a small set of flocs. Despite this limitation, the results obtained here seem promising, and application of this approach to a wider data set is expected to advance the interpretation of flocculation and floc morphodynamics of suspended cohesive sediment.