Non linear Temporal Textures Synthesis: a Monte Carlo approach ⋆ Andrea Masiero and Alessandro Chiuso Department of Information Engineering University of Padova Via Gradenigo, 6/b, 35100 Padova, Italy {andrea.masiero, chiuso}@dei.unipd.it Abstract. In this paper we consider the problem of temporal texture modeling and synthesis. A temporal texture (or dynamic texture) is seen as the output of a dynamical system driven by white noise. Experimental evidence shows that linear models such as those introduced in earlier work are sometimes inadequate to fully describe the time evolution of the dynamic scene. Extending upon recent work which is available in the literature, we tackle the synthesis using non-linear dynamical models. The non-linear model is never given explicitly but rather we describe a methodology to generate samples from the model. The method requires estimating the “state” distribution and a linear dynamical model from the original clip which are then used respectively as target distribution and proposal mechanism in a rejection sampling step. We also report extensive experimental results comparing the proposed approach with the results obtained using linear models (Doretto et al.) and the “closed- loop” approach presented at ECCV 2004 by Yuan et al.. 1 Introduction Modeling of complex scenes such as texture has been subject of intensive research in the past years. Models of physical phenomena are widely used in a number of field such as Control, Econometrics, Bioengineering and so on; in computer vision several tasks - such as synthesis, recognition, classification, segmentation - connected to video sequences are facilitated when a model is available (see for instance [3] for a specific example related to textures). Statistical-based models appeared soon to be the useful tool to tackle the problem; this line of work was pioneered by Julesz (see [7]). After that, much work has been done (see for instance [22, 19, 21, 6] just to cite a few references) which addresses the problem of modeling and synthesis of (static) textured im- ages. In this work we are interested instead in the “temporal” evolution of tex- tured images which we call temporal or dynamic textures. “Temporal” textures ⋆ This work has been supported in part by RECSYS project of the European Commu- nity and by the national project New methods and algorithms fr identification and adaptive control of technological systems funded by MIUR.