Fuzzy sets which capture the meaning representation of linguistic variables have been widely used in image processing in the last decades. Fuzzy sets are associated with vagueness which is type 1 uncertainty. Interval-valued fuzzy sets (IVFS) are associated with type 2 semantic uncertainty. Indeed, the length of the interval provides the "non-specifity" measure for IVFS. We investigate this particular information measure applied to low-level image processing. This method can be used for both smoothing and noise filtering and applications in speckle noise reduction show the interest of this concept.