Infrared image processing for local convective heat transfer measurements in rib-enhanced channels

In the context of an experimental study designed to retrieve the local heat transfer characteristics in closed channels with ribs in several configurations by IR thermography, a filtering strategy must be adopted to calculate the laplacian of the temperature field, to remove not only Gaussian noise, but also unwanted local features due to the pattern of the conductors of the heater. In this paper, these issues are addressed with a comparative analysis on filtering and interpolation techniques, in particular mean filtering, local surface fitting and cubic smoothing splines, which are carried out by showing their influence on the smoothed thermogram and on the laplacian of the temperature field. The use of cubic smoothing splines gives the best approximation of the temperature field with respect to the mean filter, which alters the temperature field in proximity of the ribs, and to surface interpolation. However, the only filter among those tried that gives meaningful derivatives on the whole image is the mean filter, with a kernel size equal or bigger to that of the spatial feature to be filtered. The resulting derivative also need further filtering to reduce local spikes induced by the filtering operations.