The Impact of Different Image Processing Techniques on the Modal Decomposition of Methane Jets Exposed to Acoustic Excitation

Currently, modal decomposition is a technique widely used in applications aimed at identifying and extracting dominant features of flows. The discovery of models based on the extraction of information from data is constantly evolving, transforming the process of modeling, predicting and controlling complex dynamic systems. A commonly used dataset consists of temporally evolving experimental digital images, which have a large degree of noise. The objective of this study is to explore the effect of various image processing methods on the dominant modes in an acoustically-coupled combustion problem extracted via Proper Orthogonal Decomposition (POD). Commonly used image processing approaches (median filters, block-matching and 3D filtering or BM3D, threshold filters, band-pass filters, Wavelet transforms, Fredholm transforms and Kalman filters) are investigated. For each approach, an optimal image enhancement parameter is found, which removes much of the noise and improves the experimental input data of the modal decomposition. Consequently, there is an increase in the probability of obtaining a greater total number of resulting modes, making it possible to define more precisely the dynamical behavior of acoustically forced diffusion flames.