This paper presents a new filter

called Averaging-Based Adaptive Filter (ABA Filter)

whose objective is to overcome the deficiency of current filters that they can not simultaneously possess the merits of both efficiency and simplicity when dealing with mixed noise. Although adaptive filters are robust owing to their effectual structures

the complexity of these filters is generally too great to be realized in real time. However

for video applications

it is very important to design a fast filter. Compared with other adaptive filters

ABA filter uses averaging as its major operation and the averaging value is utilized repeatedly in its adaptive processing so that it needs much less processing time. The adaptive structure makes ABA filter able to adopt different algorithms according to different noise and different image areas. The experimental results show clearly that

when the density of impulse noise is lower than 10%

ABA filter is capable of eliminating the mixed Gaussian noise and impulse noise with much fewer operations.