Abstract: A method for the temporal filtering of a sequence of digitized noisy images includes the evaluation of an anti-causal filtered sample (P.sub.t.sup.A) in order to reconstruct a present noisy sample (I.sub.t.sup.P) of a given pixel in a present image by an anti-causal linear combination of a causal filtered sample (P.sub.t.sup.C) obtained by preliminary causal linear temporal filtering in association with coefficients (b.sub.j.sup.C) and an anti-causal noisy sample (I.sub.t+1.sup.A). To the samples (P.sub.t.sup.C, I.sub.t+1.sup.A) there are assigned weights calculated as functions of a causal gain factor (K.sub.t.sup.C), equal to the inverse of the sum of the coefficients of the causal linear filtering, and an anti-causal continuity coefficient (.alpha..sub.t.sup.A), the weight of the anti-causal sample (I.sub.t+1.sup.A) being equal to the probability of intensity continuity between the anti-causal sample (I.sub.t+1.sup.A) and a previous filtered sample (P.sub.t.sup.C, P.sub.t-1.sup.C) in the sequence.