When it comes to cooking, salt and pepper are essential to the chef; but when it comes to imagery, researchers would be better off without it. Speckle noise, otherwise known as the salt and pepper effect, corrupts the ground truth when it comes to images obtained with coherent illumination. Of all the approaches to extracting information contained in speckled imagery, the statistical one seems to provide the best models and tools; the authors succeed in presenting an appealing road map for the delivery of more accurate inference using this approach.
Having very carefully selected their model to explain the behavior of data obtained with coherent illumination, they proceed to identify maximum likelihood (ML) as the best choice out of a variety of estimation techniques when it comes to finite samples. However, maximum likelihood estimators (MLE) can be quite biased, given small or moderate sample sizes. Based on the bootstrap method proposed by Efron [1], where resampling is utilized to obtain additional pseudo-samples and then to extract information from these samples to improve inference, the authors propose a particular bootstrap scheme. Since MLE does not have a closed form, the authors adapt the original idea by Efron in a way that does not require the quantity of interest to have a closed form.
Presenting the ideas in a progressive, well-paced manner, and supporting them when necessary from authors with a clear understanding of the field, this paper is a joy to read.