Reinitialize seed parameter in image similarity metrics



Hello. In the similarity metrics of ITK, we can use a smaller number of spatial samples used for estimation of similarity, hoping to reduce the time needed for a single estimation. In case of not using the UseAllPixelOn() function, we use a randomly sampled part of the pixels in the images. My questions are the following ones:

  1. If the ReinitilizeSeed(int seed) is executed once, will the metric use the same sample of pixels in order to estimate the similarity between the images for each given transformation?
  2. If we use the ReinitilizeSeed() (without the argument), will it use different random sample at each similarity estimation? Do we have to call ReinitilizeSeed() at each estimation in order to increase the non-deterministic behaviour of the metric?