A new two-pass switching rank-ordered arithmetic mean (TSRAM) filter controlled by evidence fusion is proposed for fuzzy image restoration. The new filter mechanism is composed of an efficient impulse detector based on Dempster-Shafer evidence theory and a rank-ordered arithmetic mean filter that works by estimating the noise-free ordered mean values. A modified Dempster’s combination rule is proposed and applied to the impulse detector. In addition, to improve the filtering performance, a second pass filtering is performed using a simple switching median filter. Experimental results have demonstrated that the proposed filter outperforms other switching-based median filters in terms of both noise suppression and detail preservation. The proposed new filter also provides excellent robustness at various percentages of impulsive noises.