The most suitable image for PIV analyses constitutes homogeneously distributed bright particle images on a completely dark background.

Images are subsequently interrogated by means of statistical operators, such as cross correlation, to infer flow velocities representative of the particle image displacements captured within the image sub-sections (Theunissen 2012). Particle image velocimetry (PIV) is a non-intrusive technique for measuring flow velocities by analysing successive images of seeding particles captured by specialised cameras. The method is assessed through a Monte Carlo simulation with synthetic images and its performance under realistic imaging conditions is proven based on three experimental test cases.

Based on the observation that the temporal variation in light intensity follows a completely different distribution for flow regions and object regions, the method utilizes a normality test and an automatic thresholding method on the retrieved probability to identify regions to be masked. The method does not require any a priori knowledge of the static objects (i.e., contrast, brightness, or strong features) as it exploits statistical information from multiple PIV images. In this work, the authors propose a novel method for the automatic detection of static image regions which do not contain relevant information for the estimation of particle image displacements and can consequently be excluded or masked out. These are, however, not always present in experimental images necessitating a more robust and general approach. The automatic detection of masks usually necessitates special features or reference points such as bright lines, high contrast objects, and sufficiently observable coherence between pixels. These data typically come in the form of a logical mask, generated manually or through automatic algorithms. The measurement of displacements near the vicinity of surfaces involves advanced PIV algorithms requiring accurate knowledge of object boundaries.

Imagej piv