A k-trimmed mean is a mean where the largest and smallest k% of the observations are removed before the mean is calculated.
A trimmed mean is a mean where a given percentage of the data from the largest and smallest observations are removed before the mean is calculated.
It is more robust than an ordinary mean; for example, a 25% trimmed mean (the average of the data between the upper and lower quartiles, sometimes called the interquartile mean) is able to tolerate up to 25% arbitrary contamination.
A median is (asymptotically) a 50% trimmed mean.
One alternative to a trimmed mean is a Winsorized mean.
Reference: Wikipedia - Trimmed mean
